The Parsons Game: Simulating Talcott Parsons’ …docshare01.docshare.tips/files/13087/130870575.pdfTHE PARSONS GAME: THE FIRST SIMULATION OF TALCOTT PARSONS' THEORY OF ACTION by

i

THE PARSONS GAME: THE FIRST SIMULATION OF TALCOTT PARSONS' THEORY OF ACTION

by Stan Rifkin

B.S., Business Administration (Quantitative Methods), School of Business & Economics, California State University at Northridge, 1968

M.S., Computer Science, School of Engineering & Applied Science, University of California at Los Angeles, 1972

A dissertation submitted to:

The Faculty of the

Graduate School of Education and Human Development of The George Washington University

in partial fulfillment of the requirements for the degree of Doctor of Education

January 30, 2005

Dissertation directed by:

Dr. David Schwandt, Professor of Human Resource Development, and

Director, Center for the Study of Learning and Executive Leadership Doctoral Program,

Graduate School of Education and Human Development, The George Washington University

Committee:

Dr. Walter Andre Brown, Assistant Professor of Higher Education Administration,

Graduate School of Education and Human Development, The George Washington University

Dr. Robert Hanneman, Professor of Sociology,

College of Humanities, Arts, and Social Sciences, University of California at Riverside

1.0 – 7 Dec 2004

Page ii

© Copyright 2004 Stan Rifkin

All rights reserved

Page iii

Contents

Abstract ............................................................................................................v Acknowledgments.......................................................................................... vi I. Introduction................................................................................................1

How this study differs from its predecessors.............................................4 Novelty of results.......................................................................................5 Conceptual framework...............................................................................7 Problem....................................................................................................16 Research question ....................................................................................17 Significance..............................................................................................17 Limitations ...............................................................................................21

II. Literature review.....................................................................................25 Theory, model, and simulation ................................................................25 The theory of action .................................................................................25 Tension management and learning ..........................................................26 Place of Parsons' theory of action in sociology .......................................27 The bad news ...........................................................................................29 Locating this work within all of Parsons'.................................................29 Models......................................................................................................30 Formalization ...........................................................................................32 Time .........................................................................................................34 Process .....................................................................................................36 Simulations of social systems ..................................................................38 Discrete event simulation.........................................................................39

III. Methods....................................................................................................43 Research overview...................................................................................43 Research methods ....................................................................................43 Delimitations............................................................................................52

IV. The model and simulation........................................................................55 Why simulate? .........................................................................................55 Model construction ..................................................................................56 Basic concept ...........................................................................................57 Model of tension and learning .................................................................58 Operation of the simulation .....................................................................63 Rules ........................................................................................................67 Assumptions.............................................................................................69 Mapping the model to theory...................................................................71

V. Results......................................................................................................76 Example ...................................................................................................76 Base cases ................................................................................................80 Extension..................................................................................................84

VI. Conclusions and recommendations for further study ..............................87 Review of purpose and research question................................................87 Review of findings...................................................................................87 Discussion................................................................................................88

Page iv

Implications..............................................................................................91 Recommendations for further study.........................................................93 In sum.......................................................................................................95

Epilogue .........................................................................................................96 References......................................................................................................99 Appendix – Attestation of an Expert ...........................................................114 Appendix – Simulation Program Listing .....................................................115

Figures Figure 1. Sastry’s "simplified causal diagram of the punctuated change theory." .............4 Figure 2. The components of action systems. ..................................................................12 Figure 3. The action system in relation to its environment. .............................................14 Figure 4. Interchange media (whose paths are represented by arrows) in the general

theory of action.. ..............................................................................................15 Figure 5. Phases in the relationship of a system to its situation. ......................................16 Figure 6. Flow from theory to action. ...............................................................................18 Figure 7. Relationship among process, event, and state (notional). ..................................37 Figure 8. Intersection of the theory of action and system simulation. ..............................43 Figure 9. Classification of social systems simulators, indicating the position of this

research in bold. ...............................................................................................46 Figure 10. Thorngate’s one-armed clock. .........................................................................21 Figure 11. Evolution of computer simulations of organizations. .....................................22 Figure 12. Structure of the three-parameter hyperbolic learning curve model. ...............61 Figure 13. Illustration of a negative exponential distribution as a "forgetting" function..62 Figure 14. User view of dedicated Excel spreadsheet. ......................................................64 Figure 15. User view of the simulation. ............................................................................65 Figure 16. User display for example with long window. ..................................................78 Figure 17. Energy for the long window example. .............................................................79 Figure 18. Energy for the short window example. ............................................................79 Figure 19. Base case for affect vs. affect-neutrality. .........................................................81 Figure 20. Pattern of internal energy following external with a strong culture. ...............82 Figure 21. Pattern of internal energy following external with a weak culture. .................83 Figure 22. Pattern of internal energy following external energy with very weak culture.84 Figure 23. Simulation after two energetic events per year, both with affect. Illustrates

queuing effects. ................................................................................................85 Tables

Table 1. Works by Parsons, his supporters, and his critics ................................................2 Table 2. Additional delimitations of the study ..................................................................22 Table 3. The system dynamics modeling process across the classic literature. ..............57 Table 4. Rules of the simulation........................................................................................67 Table 5. Assumptions made in the simulation...................................................................69 Table 6. Map of the theory to the model. ..........................................................................72 Table 7. Correspondence between what was required and what was developed. ............88

v

ABSTRACT Talcott Parsons was perhaps the best known American sociologist of the 20th

Century. He postulated a general theory of the structure and function of social systems, one at all levels of analysis and all levels of abstraction. The center of his theory is action, which he defined in his own terms of situation, conditions, ends, and norms.

For those familiar with Parsons' work, the creation described here simulated using a digital computer a very small subset of Parsons' theory of action, including his frame-work of four functions or functional prerequisites, one of the four pairs of pattern vari-ables, the cybernetic hierarchy, and interchange media. The simulation was meant to be a proof-of-concept, a toy demonstration of the feasibility of modeling such a complete and richly described social action theory.

Most simulations of social systems use a modeling technique called system dynamics, a way of characterizing flows and accumulations over time. Other researchers have tried to simulate the theory of action using system dynamics but have failed. One contribution of this research is the application of a different technique, discrete event simulation, to social systems. There are only two published applications of discrete event simulation applied to social systems. Accordingly, this work offers some insight into how to incorporate time ordering into reasoning about social systems.

Simulations were executed to demonstrate consistency with outcomes predicted by the theory. One finding was that Parsons neglected to take into account the disposition of (motivational) energy transiting through a system or organization when the energy is blocked by having to wait for the processing of predecessor energy. The effect of this oversight can be a very long wait for the availability of a prerequisite function and no guidance on whether, for example, the energy decays, dissipates, waits, interrupts, or is channeled elsewhere.

Page vi

ACKNOWLEDGMENTS "We shape our models and then our models shape us."

- Michael Schrage. (2000). Serious play: How the world's best companies simulate to innovate, Harvard Business School Press, p. 3.

Several fellow students in the Executive Leadership Program (ELP) cohort have sustained my enthusiasm, including, but not limited to, Dr. Betty Beene, Dr. Brenda Conley, Linda Hodo, and Dr. Ted Willey. Dr. Margaret Gorman has been a steady force moving me ahead and she has been a safe harbor for my administrative challenges. I always feel like Margaret treats me as a distinctive student, a gift she has for making each one of us feel special. And her dissertation was breath-taking.

I am grateful to Laura Reid of Simul8 Corporation for attesting to the veracity of the technical operation of the simulation described in this dissertation and along the way helping me to improve its operation and correctness. I am also grateful for those who came before me in the application of computation methods to organizational problems, in particular Profs. Rich Burton at Duke, and John Kunz and Ray Levitt at Stanford, for per-sonally helping me to see that engineering tools such as discrete event simulation could be beneficially applied to social systems. Profs. Kathleen Carley at Carnegie Mellon and Anjali Sastry at MIT were also instrumental in personally inspiring me to try to apply engineering methods to the questions of social systems.

Dr. Chris Johnson gave me the courage to undertake the study of Talcott Parsons. Dr. Johnson is a Parsons scholar and his energy and enthusiasm about Parsons are conta-gious and set the bar high. He is very, very busy and I am especially grateful that he has taken on three roles: expert who attests to the mapping of the theory of action to the simulation model, an outside reader for the dissertation defense, and a friend and col-league.

I consider myself an adequate researcher, but it took me too long to find Prof. Tom Fararo, a professor in the School of Sociology at the University of Pittsburgh and another Parsons scholar. Prof. Fararo, besides being an inspiration in his writing and interpretation of Parsons, has been a ready and energetic reader of my manuscripts. I am grateful for his generous expenditure of time and energy on my behalf.

Professors Walter Brown and Robert Hanneman, members of the dissertation committee, have been generous with their time and energy. They have in their unique ways given me important, stimulating feedback.

Prof. Dave Schwandt has been my close mentor throughout this long journey. He is the person who spoke to me at the very beginning about joining ELP. I was struck im-mediately then by his boundary spanning, openness to people not in his field(s), and how gentle he was with people like me who knew nothing of human resource development. A few years ago he introduced me as, "This is Stan. He is the only person I know who has the whole dissertation in his head!" Prof. Schwandt has been so generous with his time that I feel guilty. It should not have taken all of this prodding to get me to finish, but I am slow and Dave has been a patient, steady, and exacting influence. He, too, is a Parsons scholar and has not been put off by my excursions into what I felt might be important, while he kept me focused and fixed. He has the gift, too, of making each one of his stu-dents feel special and unique, and I am forever grateful for his friendship and guidance.

Page vii

All doctoral work is a family effort. My wife, Jan, from the very first moment we spoke about the Program and the time it would mean away from her, has cheered me on, even during the lonely days and evenings she spent as I studied and wrote. Everyone loves Jan and she, too, has made friends among my ELP colleagues and leaders. She has made this long journey worthwhile and possible. She also proof-read this final version, a sacrifice well beyond the pale. I am responsible for any remaining errors, faults, failures, oversights, defects, and misinterpretations.

I thank all of the people above for their gently persuading me to finish.

1

I. INTRODUCTION "All models are wrong. Some are useful." - George E.P. Box

Creating a computer program to simulate Talcott Parsons’ general theory of action has proved elusive. Two researchers who have had considerable success simulat-ing social systems (e.g., Jacobsen & Bronson, 1985; Jacobsen & Bronson, 1987; Jacobsen, Bronson, & Vekstein, 1990) have written that they were unable to construct a computer program that would simulate Parsons’ theory (Jacobsen & Bronson, 1997). The purpose of this dissertation is to describe a proof of the concept that it is possible to cre-ate such a computer program.

Many simulations of social systems cast those systems in terms of rates of change of key constructs (Hanneman, 1988). They are systems of differential equations, even if somewhat disguised. The current research framed organizations in Parsons’ terms as having a significant time-ordering of events, not in terms of rates of change. Time-ordering, directly according to Parsons’ words, was expressed as "before," "then," "next," "cycles," "phases," etc. Using a time-ordering style of simulation yielded a toy or proof of concept that was animated inside a digital computer and used to ask and answer a new set of questions about the theory, and might serve as a basis for the construction of a high fidelity simulation.

Social scientists have long sought a test bed for their ideas, concepts, and theories. Typically "the scientific method" applies to the "hard" sciences, where researchers con-sult a theory, postulate a question or hypothesis, enter a laboratory where they control environmental factors, and put together an apparatus to generate some phenomena and measure the outcomes (Kuhn, 1970). "Normal" science laboratories do not exist in the social sciences, primarily because environments cannot be controlled – or, more accu-rately, controlling the environment changes it, so normal science cannot be applied. As a result, there is some appeal in attempting to create an environment inside of a digital computer where the structure and function of a social system could be mirrored. Com-puter simulation, if appropriately validated, has offered a means to instantiate and exer-cise richly described social systems (Hanneman, 1988).

In order to instruct a computer to simulate (that is, "act like," in a reified context) a social system – to bring the social system to life – it was required to have a "sufficient" description of the (static) structure of the constructs along with a description of the (dynamic, time-varying) functioning of the static constructs. While no one knows what the definition of "sufficient" is a priori, Parsons’ theory of action was a candidate at least from a volume perspective because Parsons himself wrote thousands of pages about it, supporters have written thousands, and critics another thousand or so (see Table 1 below). It remained to read the contributions to the theory of action and extract its descriptions of salient structure and function to see if they were sufficiently detailed to simulate a social system resembling the one described by Parsons. The main purpose of this dissertation is to document the extraction and resulting simulated social system.

It is useful to note that there are detractors as well as supporters of Parsons' work. Clearly there is not universal agreement on the meaning and importance of his model. A few of the areas of disagreement are described at the end of "Place of Parsons' theory of action in sociology," in the Literature Review chapter, p. 27.

Page 2

Table 1. Works by Parsons, his supporters, and his critics

Works by Parsons Works by supporters Works by critics Parsons, 1951-1978b; Parsons & Bales, 1953; Parsons, Bales, & Shils, 1953a; Parsons, Bales, & Shils, 1953b; Parsons & Platt, 1973; Parsons & Shils, 1951; Parsons & Smelser, 1956

Alexander, 1983; Alexander & Sciortino, 1996; Barber & Inkeles, 1971; Bluth, 1982; Boudon & Bourricaud, 1989; Bourricaud, 1981; Brownstein, 1982; Etzioni, 1975; Hills, 1968; Holmwood, 1996, 1983; Lackey, 1987; Loubser, Baum, Effrat, & Lidz, 1976a, 1976b; Park, 1967; Rocher, 1975; Turner, 1999

Black, 1961; Camic, 1996; Camic, 1998; Dubin, 1960; Habermas, 1981; Kolb, 1962; McGowan, 1998; Savage, 1981; Selznick, 1961

Constructing a model of time-variation, especially of mathematical equations,

inside a computer is not new and is used often in the "hard" sciences. Jacobsen and Bron-son (1985) point out:

Generally, models can be classified as one of three types: iconic, analog, or abstract. An iconic model is one that looks identical to the system it represents. An example is a wood and paint mockup of an automobile shell. From a distance, the mockup appears to be an automobile, but as the mockup has no engine or inte-rior, it is not a complete representation of an automobile. Yet if the purpose of the model is to determine the aerodynamic characteristics of the system it represents by subjecting it to wind tunnel tests, then this iconic model is quite adequate. An analog model is one that does not look like the systems it represents but has corre-sponding behavior. Engineers often build electrical systems as analog models of mechanical systems. By measuring the current in an analogous electrical systems, they can predict the motion in a mechanical spring system. An abstract model is a set of statements about the structure of a real system. If these are formulated as mathematical equations, they may be solved and used to predict the behavior of the real system. It is this last type, abstract mathematical models, which form the basis of continuous simulation. … (pp. 57-58)

By applying this kind of modeling to the study of social systems, a researcher can watch the interaction of social forces reveal themselves with time slowed down or speeded up inside a computer and can obtain a very detailed understanding of the contri-butions of structure and of function to the specific observed behavior of the social sys-tem.

Also, the replication of theory inside a computer is not a new idea, not even for theories of organization, as illustrated in early histories provided by Hanneman (1988) and Garson (1994). The first comprehensive simulation of an organization was probably A Behavioral Theory of the Firm (Cyert & March, 1963). This work was a tour de force integration of microeconomics (that is, the setting by a firm of output level and product

Page 3

price) with organizational goal-setting, choice, and rational decision-making. While the unit of analysis was the description of a single firm among competitors, it could be expanded to the description of aggregates of firms and to non-business organizations, and to the normative analysis of a single firm and of economic policy. Incidentally, while not citing Parsons' constructs, there are many references to them without acknowledg-ment.1

The replication of theory inside a computer is experiencing recent interest in the social sciences as more researchers come to the social sciences from other, hard science areas (e.g., computer science, mathematics, psychology) (Burton & Obel, 1995; Carley & Prietula, 1994; Coleman, 1965; Conte, Hegselmann, & Terno, 1997; Gilbert & Conte, 1995; Gullahorn & Gullahorn, 1963; Hanneman, 1988; Ilgen & Hulin, 2000; Jacobsen & Bronson, 1995; Jacobsen & Bronson, 1987; Jacobsen & Bronson, 1997; Phelan, 1995; Prietula, Carley, & Gasser, 1998; Samuelson, 2000). The normal course of research in computational and mathematical organization theory is to wring structure and function from a theory, operationalize or animate them, and then make changes in the simulated environment or the structure/function and watch the computer’s results for interesting, informative patterns, including validation with respect to the underlying theory. To "ani-mate" in this context means to bring to life graphically on a computer screen.

For example, Sastry for her Massachusetts Institute of Technology doctoral dis-sertation, redacted in Sastry (1997), took a detailed, simulation-oriented look at the struc-ture and operation of how Tushman and Romanelli (1985) explained strategic change. She was able to show inconsistencies in their explanation, a more parsimonious explana-tion, and to more clearly reason about cause and effect. She showed, among other things, that there were loops that reinforced positive or negative feedback, thereby speeding up or retarding the change, respectively. The lines in Figure 1, below, represent flows of information, and the noun phrases (e.g., "Strategic orientation required") represent either inputs or accumulations of values. By simulating the operation of strategic change at the organizational level, Sastry was able to identify which postulated stores (accumulations) would be a priority to measure in the real world because of their dominant effects and which would be a lower priority because they may have only secondary effects.

1 The particular constructs, the four functional prerequisites, are explained later in the text, on p. 11. In Cyert & March

(1963, chap. 6, pp. 114-127) "A summary of basic concepts in the behavioral theory of the firm," there are goals, expectations, and choices. Regarding organizational goals, e.g., "... we have argued that organizational goals change as new participants enter and old participants leave the coalition [making the decisions]." p. 115. This is latent pattern maintenance. Organizational expectations are based on "search," an analog to environmental interface, the adaptation function. p. 116. And as to organizational choice, "the standard decision rules are affected primarily by the past experience of the organization and past record of organizational slack," which are references to pattern maintenance and integration functions, respectively. p. 116.

Page 4

Figure 1. Sastry’s "simplified causal diagram of the punctuated change

theory." (Sastry, 1997)

Sastry accomplished her task by reading the Tushman and Romanelli article (1985) and coding each passage as applying to a definition, a structure (i.e., static relationship), dynamic behavior, or not applicable for her study. She collected the state-ments about structure, for example, and derived constructs consistent with her modeling choice (system dynamics in that case) and training as a system dynamicist. She con-structed a computer replica of the static and dynamic components and animated it by having information from the simulated environment flow along the lines of the diagram. She then graphed the changes in accumulations and how well the strategic orientation tracked the required one. She made changes in the flow rates and accumulation rules as experiments. Her article essentially reports the patterns she observed based on varying what she postulated were independent variables. In conclusion she observed by simula-tion six novel ways that managing strategic change failed (Sastry, 1997).

The approach of this study was to construct a high fidelity replica of the essential aspects of the theory of action, so the replica mirrored the elements of action that Parsons described as indispensable: the situation, conditions, ends, and norms (Parsons, 1968a, p. 44). As well, it modeled time because Parsons’ theory described time-varying behavior: action by its definition and nature is dynamic.

How this study differs from its predecessors Sastry (1997) and Jacobsen and Bronson (1985) both applied system dynamics

(Hanneman, 1988), a symbolic representation of systems of differential equations, to organizational and social studies, respectively. Jacobsen and Bronson (1997), in a paper summarizing their 15 years of social system simulation, note, "A ... theory we tried to model was Parsons' General Theory of Action. We chose it for its renown and because of the controversies around it. We soon found that it could not be modeled at all because of the unclarity and inconsistencies in Parsons' use of concepts." (p. 98 ) Their challenge was to translate elements of Parsons’ theory of action into the standard system dynamics form of information flows among accumulations. They tried having material flow (the

Page 5

kind that is physical and is allocated as part of Goal Attainment) be a model of roles or resources. They tried modeling the way of acting and the method of giving meaning to actions, but were left wondering where the next generation arose. They tried having three of the functions concentric around one of them, but that would contradict Parsons' dia-gram that shows them all interconnected. They tried a causal loop diagram among the four functions, but it traveled in the opposite direction that Parsons envisioned. They tried using the four functions as "valves" or controls on the stock of loyalty, power, order, and goods. They considered pattern variables as "the ranges on which the other concepts could be measured." In the end they abandoned their modeling effort. (Jacobsen, personal communication, October 31, 2000).

Parsons (1953a) writes, referring to his theory of action, "Since we are dealing with processes which occur in a temporal order, therefore we must treat systems and the processes of their units as changing over time." (p. 167) [italics in original.] "The first important implication is that an act is always a process in time. The time category is basic to the scheme." (Parsons, 1968a, p. 45). Jacobsen and Bronson can justify their (failed) approach by these statements (alone) because they sought to replicate the theory in terms of time-varying constructs. The present research took a (slightly) different approach and applied the iconic model, per Jackson’s advice to construct a high-fidelity model (Jackson, 1983, pp. 4-5)2, not the abstract mathematical one of system dynamics. This way there was no need to guess what in Parsons’ theory corresponded to the system dynamics constructs of flows and accumulations (which Jacobsen and Bronson had to). Instead, in this research there was a more literal translation between the elements of Par-sons’ theory and the simulation – though the translation was not total, as many, many bits of the theory were left out. For example, Parsons' posits four pairs of "pattern variables" and this research only models one of them, the one dealing with affect vs. rationality.

In particular, this research will concentrate on the "temporal order" aspect of Par-sons’ descriptions.

The idea of a mathematical model as theory in mathematical form began to take hold [in the 1950s]. Writers of variant interests all agreed that such models per-mitted the logical derivation of falsifiable claims about some class of phenomena. This differentiated mathematical models from curve-fitting and data analytic reduction methods. (Fararo, 1984, p. 152)

Indeed, this dissertation relies on a "cousin" of mathematical models, simulation, and therefore is not of the curve-fitting or data analytic reduction variety. There are no correlations, no Cronbach's alpha, no significance tests. In fact, there are almost no sta-tistics at all because, in great part, it deals with a theory at the analytic level.

Novelty of results Results were sought that are important, significant, but what might be the defini-

tion of those terms? Should "novel," "useful," "interesting" or "surprising" be added? In the spirit of propounding testable hypotheses, Parsons' theory of action does not predict

2 Not all models seek fidelity. One operational measure of fidelity is correspondence: for every important construct in

the world there is a (corresponding) construct in the model. Another term for correspondence might be requisite variety: for every variation in input there is a corresponding control or regulation such that the output matches the variation (Ashby, 1956, chap. 11). Some researchers deliberately translate what they sense into frameworks that are not mirrors of the originals, thereby not seeking correspondence or requisite variety.

Page 6

what would happen in an organization that faced a high frequency of changes in its envi-ronment, too much for it to absorb in the interval in which the changes could be made sense of an acted upon (so called "permanent white water" (Vaill, 1996)). Would the changes accumulate, be discarded, decay, or queue? Perhaps a computer acting like a system of action could shed some light.

What is surprise, what is novelty? Van Fraassen (2002) argues that if science is objective, then when a scientist sets up an experiment he/she anticipates that the values of measured parameters will fall within certain bounds; the experimental setup is established to measure just those parameters and just at those levels. This, after all, is the orthodoxy of the science being invoked. So what can be regarded as novel that would be sensed during such an experiment? In part it might be that the objectively measured results would not match those anticipated by the theory, even though the experimental apparatus were established within the theory in the first place.

And van Fraassen (2002) reminds us that Kuhn (1970) has struggled with the same dilemma and concluded that novelty, when it can be sensed, may yield a change in the orthodoxy – incidentally, still in terms of scientific methods that imply objectivity – if not the facts of the particular theory; there would be no resort to mythology or religion (because that would alter the method). So, novelty in van Fraassen and Kuhn's views is possible and admissible.

Shackle (1969) postulates a calculus of surprise by introducing the notion of potential surprise.

A man cannot, in general, tell what will happen, but his conception of the nature of things, the nature of men and other their institutions and affairs and of the non-human world, enables him to form a judgement as to whether any suggested thing can happen. In telling himself that such and such a s thing 'can' happen, he means that its occurrence would not surprise him; for we are surprised by the occurrence of what we had supposed to be against nature. (p. 67) [italics in original]

Shackle first divides a spectrum into extremes: perfect possibility would not sur-prise a person and impossibility would engage the absolute maximum degree of surprise (p. 68). Between them are degrees of possibility with their corresponding inverse degrees of surprise. While not important for the research here, Shackle posited that the dispersion of degree of possibility and corresponding degree of surprise is not a probability distribu-tion, but rather is non-distributional, that is, is not a function. One can have many events that are not a surprise and their probability would not sum to unity.

To summarize, Shackle relates the degree of belief inversely with the degree of surprise: we are surprised by that which we believe cannot happen.

What might surprise look like in the research to be described here? First, the reader would have to think it was impossible to achieve. To a small subjective degree, this has happened. When this researcher mentioned to other members of his school cohort what he is trying to do many of them expressed doubt that it would be possible. Further, Jacobsen and Bronson tried it and failed, so there is a hint of impossibility. "Some people don't believe that models of human behavior can be developed." (Sterman, 2000)

Second, the method of inquiry, a computer simulation, is far less restrictive of an experimental setup than a traditional laboratory so some behavior might be observed that was not within expectation, not predicted by the orthodoxy, and therefore would be sur-prising within the ambit described by van Fraassen and Kuhn.

Page 7

There are perhaps two more reasons that surprise might be expected, both because Parsons has written so much. First among these is that it is predictable that there might be contradictions or gaps in there somewhere, the precise nature of which might generate surprise.3 And second among these is that so many people have already vetted Parsons' theory that anything new would be unexpected at this late date.

Conceptual framework The fundamental framework that informed this research is that of systems. A sys-

tem is a collection of elements and interactions whose structure and function can be understood by looking at the whole, the summation, the interaction of elements. This description highlights a tension in systems study. Some scholars infer qualities of the whole by studying the parts (methodological reductionism, typical in normal science), others claim that that one can never appreciate the whole by dissecting the parts (holism) (Honderich, 1995, pp. 750, 372); (Sibeon, 1999).

The approach in this study was somewhere in the middle: the whole was studied by understanding its parts, but not separately, rather as they interacted and collaborated in patterns to define the whole. Thus, the emphasis was on how the parts were connected, what flowed along those connections, and how the interplay of connections and flows defined an organic whole.

Parsons (1968b) wrote: Action systems have properties that are emergent only on a certain level of com-plexity in the relations of unit acts to each other. These properties cannot be iden-tified in any single unit act considered apart from its relation to others in the same system. They cannot be derived by a process of direct generalization of the prop-erties of the unit act. (p. 739) [emphasis in original]

The term energy used in this research denotes the material in the environment of the system that is external to it and that can be sensed by the system. That is, energy is the term used to label the elements in the world external to the system under study that can be used to stimulate the system, that can energize and activate the system to respond to the environment. Sometimes Parsons refers to this energy as motivation (Parsons et al., 1953a). Concretely, the energy could be news, ideas, or information, for example. News, of say an invention or a move by a competitor, could stimulate the system (in our case an organization) to evaluate the content and respond to what it sensed in the external world. One important point is that the term energy used here is not the same as that used in physics; in physics energy is conserved, that is neither consumed nor created, but in the use here (social) energy may be infinitely created and transformed and possibly even consumed. Parsons postulated a law of conservation for motivational energy, which in its central part would claim that motivational energy is exchanged for changes in the system (Parsons et al., 1953a, p. 168). Alas, the merits of such a law of conservation of social energy is beyond the ken of this research.

If one considers a unit act to begin with the importation of exogenous energy, then the social system presented by this research will follow the trajectory of that energy as it transits the replica of an organization. In order to imagine what the emergent proper-

3 Brownstein (1982) has found contradictions and gaps by converting a portion of the theory of action into a set of

logic statements and showing the inconsistencies therein.

Page 8

ties could be possible it is necessary to understand at least the rudiments of Parsons' theory of action. Parsons' theory of action

Talcott Parsons, perhaps the best known American sociologist of the 20th century, attempted to construct a theory that explained organizations at all units of analysis. He was attracted to and influenced by the systems view4, among many other influences, and noted that organizations are structured in specific ways, not randomly. He noted that structure and function interplayed, that the functions of an organization were executed by elements of its structure. He noted that the execution also was not random, that it responded to normative pressures that could be applied exogenously and endogenously.5

He sought to understand the patterns of structure and function. His view was not reductionist, he was trying to see organizations at their highest levels of abstraction. In order to construct a high-level framework, Parsons (1960) defined the atomic unit, the unit act, to which everything else would refer:

The unit act involves the relationship of an actor to a situation composed of objects …. The unit act, however, does not occur independently but as one unit in the context of a wider system of actor-situation relationships, … referred to as an action system. … Action is thus viewed as a process occurring between two structural parts of a system – actor and situation. (p. 467) [italics in original]

While the unit act is the atomic level, the social system describes social interac-tion, behavior. Behavior that directly concerned the "cultural level" Parsons called action. Said another way, relying on Weber, which Parsons often did, particularly related to action:

Interpretive understanding of social actions is a prerequisite for the causal analy-sis of social structures and processes. In modern form, we can put it this way: there is an actual world of events and some events are behaviors. Behaviors are treated as actions when they are analyzed relative to cultural frames of reference, according to which the behavior means one or more things to the producer of the behavior and to the situational interpreters of the behavior. (Fararo & Skvoretz, 1984, p. 148) [italics in original]

Action includes four generic types of subsystems (that is, collections at which unit acts occur): organism (atomic level, the individual), social system (generated by the interaction among individual units), cultural systems (patternings of meaning, such as beliefs and ideas), and personality (the learned patterns of social and cultural interaction) (Parsons, 1977b, p. 178). We might re-phrase these units of analysis, in order from small-est to largest, in terms of the sciences that usually describe and study them: biology, the organism's physics and chemistry, its "atomic" structure; psychology, the individual's learned behavior and decisions; sociology, the collective structure and action of individu-als; and anthropology, the national or religious influence.

4 "System seems to me to be an indispensable master concept …." (Parsons, 1977b, p. 101). 5 The non-randomness is the subject of an entire work, Parsons, Bales & Shils (1953b), according to Parsons (1960, p.

195).

Page 9

Parsons described his theory, in the terms most important for this research, using several constructs: pattern variables, functional prerequisites, interchange media, and the cybernetic hierarchy. Parsons theory was much larger than these constituents, but they were the ones replicated in this research. It was an untestable (and therefore not falsifi-able) assumption of this research that the axes mentioned are the core of the theory of action. Or stated more positively, if one can simulate these elements then the theory of action can be simulated. Pattern variables.

Robert Bales, a student of Parsons', was studying small group interaction. Bales' method of primary research was observation: he would watch actual groups deal with real situations. He came to see patterns, broadly tasks and emotions. And he saw in groups that questions about tasks and emotions were asked and answered. He subdivided the patterns into what he called four problem areas: expressive-integrative social-emotive positive and negative reactions, and instrumental-adaptive task area questions and answers. The modern depiction of this is illustrated in Bales (1999, figure 6.1, p. 165).

In a few words, Bales observed small groups and saw patterns in the interactions among the participants. He saw questions and corresponding answers, he saw attention to the work or tasks of the group, he saw positive and negative emotions, he saw reactions to external and internal stimuli, he saw planning of tasks and work processes, and he saw setting of norms and expectations and the response of performance to them, among others.

Parsons adopted Bales' framework and adapted it to describe the patterned struc-ture and function of organizations. He called the original five, later reduced to four, pairs pattern variables (Parsons, 1960):

Each variable defines one property of a particular class of components. In the first instance, they distinguish between two sets of components, orientations and modalities. Orientation concerns that actor's relationship to the objects in his situation and is conceptualized by the two "attitudinal" variables of diffuseness-specificity and affectivity-neutrality. … Modality concerns the meaning of the object for the actor and is conceptualized by the two "object-categorization" vari-ables of quality-performance and universalism-particularlism. It refers to those aspects of the object that have meaning for the actor, given the situation. (p. 468) [italics in original.]

The purpose of the classification was to suggest propositions about action systems in terms of the components and the type of act their combination defines and controls. In this section pattern variables are described, then their patterned movement is described, and finally the patterned movement is structured to yield what becomes in the section after this one the four functional prerequisites.

At base, action is grounded in motivation and emotion or its polar opposite, disci-pline. The emotional pole is called affect and the discipline or deferred gratification pole is called affect-neutral. The affect pole is considered an expressive action and the affect-neutral pole is considered a rational or cognitive pole. Fararo (2001) illustrates the differ-ence: "In the judge-defendant social relation in American society, in the public trial situation, the judge is expected to restrain herself from expressing feelings of liking or not liking the defendant. This constitutes a specific aspect of socially responsible action expected of a judge." (p. 150)

Page 10

Parsons averred – based on Bales' small group interaction observations – that cognitive standards were expressed in a more general form that transcended any particu-lar group, while emotional or "appreciative" standards were applied to more particular collections. The polar opposites then became the universalism-particularism pattern vari-able. Fararo (2001) illustrates the difference: "A judge is expected to apply the same body of law to any defendant before her. Socially responsible action defined by this universal-ism means transcendence of the particular relationship to the specific defendant before her." [italics in original] (p. 151)

The outcome of social action can be characterized either by the result of interac-tion or by the role-status of an actor or actors. The pair of opposites is variously called quality-performance or ascription-achievement. Fararo (2001) illustrates the difference: "To be appointed as a federal judge, a person must satisfy certain performance or achievement criteria pertaining to education and experience. The judge is evaluated by reference to performance, not according to race or gender." (p. 152)

In social action each actor may focus attention on a specific social object or in a "plurality" of social objects. The pair is called the specificity-diffuseness pattern variable. Fararo (2001) illustrates the difference: "A judge is expected to confine her interest in the defendant to trial-related matters." (p. 152) That is, the judge would have a specific focus, not a diffuse interest in the affairs (that is, social actions) of the defendant.

Before relating these pattern variables to each other, it is worth mentioning that either separately or in the combinations to be described next the values of the each pair represents in each social situation what is acceptable, the norm, the expected, the institu-tionalized pattern of appropriate orientation. In this sense, as Fararo reminds us (2001, p.149), the value of pattern variables acts as part of a (yet to be described) control mecha-nism to stabilize social action. When the values are the expected ones, then there is rein-forcement; when the values are not the expected ones then the social system responds to set the value right.

Parsons unlinked and then re-linked the pattern variables, each orientation with each modality, this way: universalism with specificity, particularism with diffuseness, performance with affectivity, and quality with neutrality. While not evident at this point in the exposition, the re-linking corresponded (Parsons said "converged," (1960, p. 468)) to the classification of functional problems or prerequisites that Bales had earlier formu-lated (Bales, 1950).

The researcher asserts that in order to demonstrate the feasibility of simulating Parsons' theory, only one pattern variable needed to be selected. As will be explained on page 46, below, affective and affective-neutrality were selected. The affective orientation is that the actor responds to the situation emotionally, and its opposite, the affective-neu-tral orientation, is that the actor responds rationally, cognitively, not emotionally. (Heise, 1979) states "Events cause people to respond affectively." Clearly, Heise is not going to agree with Parsons on this issue! This would have been important if Heise's work on af-fect, situated action, affective reactions, events, and social processes (op. cit.) were going to inform the work reported here. Instead, Heise notes in his comprehensive and accessi-ble work that his framework is incommensurate with Parsons' (Heise, 1979).

Page 11

An action system is not characterized solely by the actor's orientations and modalities; it is also a structured system with analytically independent6 aspects that the pattern variable combinations cannot take into account. That is, pattern variables are a necessary but not sufficient categorization. In particular, in a structured system both actor and object share norms (Parsons, 1960, p. 468); this is one definition of interpenetration. Four functional prerequisites.

Starting with the pattern variables and after a set of steps that consumed hundreds of pages in Parsons (1960), Figure 2, below, the components of action systems, was ulti-mately offered; it is not necessary to understand everything in the figure. At its heart are four major quadrants at the intersections of external-internal and instrumental-consum-matory. These correspond to the four functional problems that Bales identified and Par-sons refined. Internal and external refer to inside and outside of the organization, endoge-nous and exogenous, respectively. Instrumental applies to means and consummatory applies to ends. The names of the major quadrants, starting in the upper left corner and moving clockwise, are Adaptation, Goal Attainment, Integration, and (Latent) Pattern-Maintenance (AGIL). These four are imperatives, prerequisites for any organization, in fact any organism, to address in order to survive, that is, in systems terminology to maintain its boundary.7 They are also the functions performed with respect to social actions.

6 "Analytically" is used in the sense of Kant (1896), namely that it is true by definition or logic or deduction, not by

experience (which would be synthetic, inductive, empirical). At one point, Parsons writes (1968a, p. 34), "It is these general attributes of concrete phenomena relevant within the framework of a given descriptive frame of reference … to which the term 'analytical elements' will be applied." [italics in original.]

7 "The difference between system and environment has two especially important implications. One is the existence and importance of boundaries between the two. Thus, the individual living organism is bounded by something like a 'skin' inside of which a different state prevails from that outside it. … The second basic property … is that in some sense they [organisms] are self-regulating. The maintenance of relative stability, including stability of certain processes of change like growth …, in the face of substantially greater environmental variability, means that … there must be 'mechanisms' that adjust the state of the system relative to changes in its environment." (Parsons, 1977b, p. 101)

Page 12

Figure 2. The components of action systems. (Parsons, 1960, p. 470)

The Adaptation (A) function imports and filters energy from the external world, the environment, and, based on the external and internal norms, attaches symbolic meaning to it. The Goal Attainment (G) function sets goals (that is, ends) and allocates resources in the service of those goals, based on the symbolic meaning of achieving them. Integration (I) aligns the structure and function of the organization to the goals in accor-dance with the resources allocated. Latent Pattern-Maintenance (L) establishes and then maintains the internal norms. "Latent" is used to refer to something unseen, the opposite of manifest, and the pattern being maintained is what lay persons call organizational cul-ture. Parsons (1977b) says:

"The most important single condition of the integration of an interaction system is a shared basis of normative order. Because it must operate to control the disrup-tive potentialities (for the system of reference) of the autonomy of units, as well as to guide autonomous action into channels which, through reinforcement, enhance the potential for autonomy of both the system as a whole and its member units, such a basis of order must be normative." (p. 168) [italics in original]

Figure 2 also illustrates a collateral point: Parsons set out to develop a grand uni-fied theory that could be applied up and down the units of analysis, from individual to collective to culture. Accordingly, each of the four functional prerequisites can be sub-

Page 13

divided into four more functional prerequisites, and so on infinitely. The figure shows the first two divisions, one at the systems level and then the next at the level of each func-tional prerequisite. Note that the lower right quadrant, the one corresponding to Integra-tion, contains the four functions in the same order as the square containing it. This illus-trates the importance and centrality of Integration, as indicated by the quotation in the paragraph above. And it also illustrates that the diagrams can be used to designate differ-ent levels of abstraction or granularity.

The conceptualization of the pattern variables potentiated Parsons' understanding of the four functional prerequisites because they all fit together so harmoniously. Cybernetic hierarchy.

Figure 3 is the same as Figure 2 in the sense that it contains the same 16 pattern variable combinations (listed in the upper right hand corner of each box), but the rows and columns are arranged differently; it is not necessary to understand everything in the figure. The rows (i.e., functional prerequisites) are now in the order L-I-G-A, and the columns in the order L-I-A-G. Note along the left edge that there is a direction of control and a direction of limiting conditions. These are the cybernetic hierarchy of control. The organization is controlled, first and foremost, by its internal norms. The norms even con-trol which energy is imported and the sense is made of it; which particular energy is im-ported and what particular sense is made of it depends upon the value ascribed to the norm. Therefore, L is the most controlling and A the least.

Each cell categorizes the necessary but not sufficient conditions for operation of the cell next about it in the column, and in the opposite direction, the categories of each cell control the processes categorized in the one below it. For instance, the definition of an end or goal controls the selection of means for its attainment (Parsons, 1960, p. 477).

Page 14

Figure 3. The action system in relation to its environment.

(Parsons, 1960, p. 476)

Interchange media. Parsons next postulated the means by which the 2 x 2 quadrants intercommuni-

cated. Each quadrant is a function and, to form a system, it communicates to and is com-municated from each other one. As one can see in Figure 4 there are 12 such paths (lines with arrowheads to and from each of the four functional prerequisites); it is not necessary to understand everything in the figure. He called the paths interchange media and along them pass symbols, not (usually) physical objects. That is, each function produces and consumes symbols, and that is how each function intercommunicates. One particularly salient depiction of the interchange media is Figure 5, in which a cycle or phase move-ment is illustrated; note the (clockwise) sequence 1, 2, 3, and 4 among the functional pre-requisites in AGIL order. It is not necessary to understand everything in the figure, only the order in which the phases occur with respect to the situation of the organization.

Page 15

Adaptation

Latent Pattern Maintenance

Integration

Goal Attainment

Figure 4. Interchange media (whose paths are represented by arrows) in

the general theory of action. (Adapted from Parsons & Platt, 1973, p. 435); see also Parsons, 1977c, p. 263).

The intuition is that the Adaptation function scans and senses the external envi-ronment and might find some information there that could be imported as energy and passed along (via the interchange medium) to the Goal Attainment function. The Goal Attainment function then might use that information either to change its goals or to change its resource allocation. These changes, expressed symbolically as new goals or new resource budgets, would travel along an interchange medium to the Integration function. The Integration function would then decide how best to arrange the elements of the organization in order to achieve the goals in light of the resources. One can imagine, for example, goals around improved quality and productivity and these would get trans-lated by the Integration function into organizational entities (e.g., VP of Quality or the Quality Improvement Department), job descriptions, new methods of rating job and unit performance, new methods of incenting the desired behavior, new methods of recruit-ment and advancement, and new training. In turn these new structures and functions would activate the Latent Pattern Maintenance function via an interchange medium and the L function would respond, principally by trying to reset the organization to the status quo in ante. L communicates via interchange media connected to the other three func-tions.

The L function works internally by manipulating a construct called tension, which is the difference between what the organization aspires (expressed by goals and resource allocation, that is, the Goal Attainment function) and what it achieves. When achieve-ment is low with respect to aspiration and the environmental situation, the L function is more controlling, it tries to track more closely the energy being imported so that it can match the organization to the environment. Symmetrically, when there is little difference between aspirations and achievement, that is, when tension is low, then the L function is less controlling, more "quiet."

Page 16

Figure 5. Phases in the relationship of a system to its situation.

(Parsons et al., 1953a, p. 182)

Recapitulation. To recapitulate the structure of the theory of action, the 16 possible combinations

of the four pairs of pattern variables gave rise to the 2 x 2 arrangement at the next higher level, the so-called AGIL framework that captures the functional prerequisites that every organization has to address to establish and sustain itself. The quadrants communicate via interchange media and there is a priority of control in that communication, in accordance with the cybernetic hierarchy.

As stated in Section I, Introduction, no one knows a priori how much or little is needed to simulate a particular social system. The researcher speculated that in order to simulate the theory of action there must be at least representatives of the pattern vari-ables, functional prerequisites, interchange media, and cybernetic hierarchy.

Problem The interface between description of systems and social action was informed by

soft systems methodology (Checkland, 1999; Checkland & Scholes, 1999). Checkland realized that many "hard" engineering projects fail because they do not take into account the "soft" factors that humans introduce, such as the power structure around the project. He offered a step-by-step method for integrating hard systems and soft ones. His was a

Page 17

pioneering effort to integrate social and engineered systems, and was a source of inspira-tion for this research.

However, present computer simulations require specific details about the structure and behavior of variables and the interactions they animate. In addition, much of the work in social systems is limited in detailed description, what Checkland calls "rich description." (Checkland & Scholes, 1999, p. 45). Thus, one of the problems is that either sufficient data for detailed description or a comprehensive and appropriately complex theory (e.g., that with requisite variety (Ashby, 1956)) needs to be found, and then one needs to see if it is sufficient for a computer simulation to be constructed and operated.

More specifically, can Parsons’ large body of descriptive text be understood? Can the salient factors (structure and function) be extracted? Finally, is it possible to instanti-ate, make concrete, those salient factors so that a high fidelity representation of the descriptive theory of action can be constructed?

Even if the questions could be answered, one is left with: Are there any novel insights? Is there anything useful to be learned? Can anything significant be predicted? Can the simulation predict something on which Parsons is silent? Is it possible to obtain enough confidence in such an undertaking that it could function as "the right answer"? Asked a different way, "Is it possible to develop a laboratory replica of the theory of action, and if it is then can anything interesting be inferred from operating it"?

In addition, there is no published attempt that successfully simulates any part of Parsons’ theories. Also, there are few published applications of discrete event simulation to social systems. Therefore, this contribution is an early and humble set of results in the use of that tool to be added to the existing scarcity.

Research question The question guiding this research was "What is the minimum set of structures

and related functions that can simulate Parson’s theory of action to some criteria of validity?" That is, what was the most parsimonious selection of theory of action con-structs that, when animated, achieved a given level of fidelity? Can the theory of action be simulated using only the functional prerequisites, (one pair of) the pattern variables, (four of) the interchange media, and the cybernetic hierarchy of social control.

Significance This study contributes to the three areas traditionally addressed by social science

research: • Theory building – This work may enrich Parsons’ description by making con-

nections that Parsons did not, for example between the frequency of changes in the environment and the rate at which change can be sensed and incorpo-rated by an organization. In addition it may identify gaps in description, at least gaps needing to be filled in order to simulate. In addition, this study will contribute to the evolution of applying Parsons’ theory to additional contexts, following a long tradition (Black, 1961; Etzioni, 1975).

• Methods – This work may add methods of translating theory statements into structure and function constructs. These constructs can in turn act as testable hypotheses amenable to a range of theory validation techniques. It also may help to make the case for additional study of time-varying research.

Page 18

• Theory (of science) – This work may add a brick in the discussion of where simulation fits into the practice of science: it is a tool of theory understanding, a tool for theory building, and/or a tool for theory testing.

The reason "game" appears in the title of the dissertation is that there is something of a game that the user of the simulation described in this research can play by varying the inputs and seeing what an execution will produce.8Simulation

The conceptual framework for constructing a simulation from descriptive text emanates from the flow from theory to action, Figure 6.

Theory

Model

Constructs

Variables

Data

Analysis

Action

Figure 6. Flow from theory to action. (David Schwandt, personal communication)

Theories are our ontologies, they are the bases for our beliefs about what we can know for sure (epistemology) and what constitutes valid activities to seek certainty (methodology). We extract from theories various features and organize them, calling that organization a model, which is the theory with some parts left out (that is, the translation from the theory to the model is incomplete). The features and organization are at a level of abstraction, usually the highest, the one with the largest blocks and thickest lines between them. Sometimes collections of the blocks and connections are named or renamed. The thing renamed is called a construct. For example, we use the term "orienta-tion" to identify the performance and learning perspectives of Parsons theory of action. We invent the term "orientation" to be used in that sense. Constructs in turn are com-posed of variables, factors that take on different values, that is, that eponymously vary. The collection of values is called data, which are analyzed so that inferences about actions can be obtained.

The description of Parsons' theory of action forms the theory referred to in Figure 6. The model in that figure is the same theory but only with certain (not all) elements and connections and is the subject of this research. As stated above, one should at least be able to discern in the model to be presented the pattern variables, four functional prereq-uisites, interchange media, and cybernetic hierarchy because they are the cornerstones of the descriptive theory.

8 Using the terms described in the Conceptual Framework section, the "game" is to see whether latent pattern

maintenance will follow the input energy.

Page 19

The goal of any simulation is to animate the elements, to put the time-varying elements onto a "canvas" or work space where there "movement" through time can somehow be visualized. In this research the canvas is a computer screen with a drawing resembling Figure 5 on it and "behind" the picture, in a way not seen by the user, the computer simulates the path of energy entering the organization and transiting in turn through the AGIL cells. Details are provided in Chapter III.

The simulation represents a set of choices – and inventions.9 Explicating what choices are available and what choices were made and why is the subject of this sub-section. Farraro and Hummon (1994, pp. 29 ff), mirroring Fararo (1989, ch. 2), provide a conceptual framework for presenting the choices and for making clearer which parts of the simulation are provided by Parsons and which are provided by the researcher. There are six "key menus" that have to be selected and explained (these are categories and their scales):

i. State space: categorical or continuous ii. Parameter space: categorical or continuous iii. Time domain: discrete or continuous iv. Timing of events: regular, incessant, or irregular v. Generator: deterministic or stochastic vi. Postulational basis: equations or transition rules The state space is the cross product – the combination – of all valid values of all

variables, including how the "boxes" are connected and what flows among them. In the instant case the space is made up of category values, not continuous ones. For example, the Adaptation function is connected to the Goal Attainment function; both of these functions are categories, as is "connected." Parameter space is the cross product of all fixed properties of the system. In this case, parameters include, but are not limited to:

• The magnitude of energy entering the system – A small integer, ordinal scale. • Energy threshold; value that has to be compared and if true then the energy passes

into the system – Same units as the magnitude of energy. • Sense of the comparison test – Category: >, >=, =, <, <= • Whether the energy will be dealt with affectively or not – Boolean. • Time to be spent in each functional area if affective – A quantity of simulated

time; without loss of generality time is represented as a positive integer. • Time to be spent in each functional area if not affective – A quantity of simulated

time. • With respect to learning and forgetting:

o Value of prior learning – A quantity of simulated time. o Time to reach the current pattern – A quantity of simulated time. o Time since the last change – A quantity of simulated time. o Starting value of Latent Pattern Maintenance energy – Same units as magni-

tude of energy • Length of time to simulate – A quantity of simulated time.

9 This point is important because we define model as a subset, an incomplete isomorphism, of the theory. So the model

cannot contain anything invented. But the simulation might in the service of finding a computer method of replicating the elements, structure, or flow of the model. And that is what is meant by the additional clause "inventions."

Page 20

The time domain is described in discrete units. Parsons did not indicate what reasonable time values were, so the researcher assumed that the basic unit was one day. Accordingly, one day passed for every click of the simulated clock, and all clicks occur in integer multiples: time is discrete, not continuous. The timing of events was irregular and depended upon what has happened before. In fact, the simulation clock did not click, rather it moved ahead to the time of the next event and in general that interval cannot be known a priori. And the state variables will only be defined for the discrete, integer time units.

By the generator of the process, Fararo and Hummon (1994, pp.30-31) signify the means or mechanism by which the system produces changes in its state. The two mecha-nisms are by rolling the dice or determinism. Rolling the dice, or having the transition be probabilistic, can be accomplished in discrete event systems; in fact, any probability dis-tribution can be imitated. Deterministic means that there is certainty (probability = 1) that a state changes from the current one to the next. In the research described here, the tran-sitions were deterministic, there was no randomness to selecting the next state.10

By postulational, the authors mean the mechanism by which transition to the next state is specified. Typically, in discrete event simulation the next state depends directly upon the current state and the transition rules. For the research described here, the pri-mary transition rule was: when it is time for energy to move from one functional area to the next, the system attempts the move; if the next functional area is already occupied then the energy is moved to a corresponding queue instead, otherwise it moves the energy to the (empty) functional area.

Which of the foregoing has been described by Parsons and which was invented/created by the research? The categorical state space has been specified in Par-sons, Bales and Shils (1953a) and so has much of the parameter space, though the actual values of the parameter space was assumed in the research; without loss of generality any user of the simulation can change any of the parameter values, so this invention does no violence to the structure of the theory. That the time domain is discrete is a computational convenience and is not suggested one way or the other by Parsons. The timing of events as irregular is consistent with Parsons, Bales and Shils (1953a) and the other two options (regular and incessant) would not be. The deterministic generation of next states is im-plied in Parsons, Bales and Shils (1953a) because there are no probabilities mentioned or suggested. And the postulational basis is clearly not equations, so transition rules are im-plied.

Therefore, to create a model to represent the dynamics of Parsons' scheme we developed a system that managed energy in discrete units and moved those bundles of energy through the processes in accordance with the AGIL framework and governed by the feedback and control hierarchy. Specifically, we envisioned a concrete organization that processed inputs from its environment, though perhaps Parsons would have argued for the generality of the processes at every level of analysis.

10 Here is an example of a probabilistic transition. Imagine a simulation of a retail store, such as a grocery. A shopper

would select a random number of items to purchase and that number would put the shopper into the cashier line for the appropriate number of items, such as the regular line or 15 or less. It would be impossible to know in advance how many shoppers ended up on the 15 items or less line because the selection is random and generated during the execution of the simulation.

Page 21

Limitations The limitations of a study are those characteristics of design or method that set

parameters on the application or interpretation of the results; that is, the constraints on generalizability and utility of findings that are the result of the devices of design or method that establish internal and external validity. In a quantitative study the most obvious limitation would relate to the ability to draw descriptive or inferential conclu-sions from sample data about a larger group.

There are two viewpoints that both properly identify this first attempt at simulat-ing Parsons’ theory of action: 1. Thorngate (1976), in attempting to describe the range of explanatory power of theory,

drew the one-armed clock in Figure 7. He stated that a particular model or theory can-not simultaneously be general, simple, and accurate. Rather, the researcher must trade among those outcomes. Clearly, the research described here was simple, so it was neither general nor accurate. Accordingly, the results will have to be used with great care (not general) and will not apply to any actual social system (not accurate).

General

SimpleAccurate

Figure 7. Thorngate’s one-armed clock. (Adapted from Thorngate, 1976, p. 406)

2. Thomsen, Levitt and Kunz (1999) suggested that simulations go through stages, Figure 8. The first stage is to build a "toy" to see if the simulation can even be built and whether it will have interesting properties. Again, clearly that was the stage of the simulation presented here. Accordingly, the results are vigorously disclaimed as a modest first attempt, really a toy, that may not be applicable to any set of facts, but rather should be seen as a foundation to be enhanced and expanded. Indeed, some elements of the simulation were given arbitrary values in order to achieve simplicity and the arbitrariness detracts from the significance of the outcome (Fararo & Hummon, 1994).

This theoretical approach to models [theory in mathematical form] included the idea of "successive approximation" articulated in sociological theory first by Comte, then by Pareto and the later stressed by Homans. Models were not expected to be correct in every detail nor to cover the entire potential scope of interest in a class of phenomena. They were to be modified and generalized (in a formal sense) over time. Even though such a model might include entities and processes not presently observable, the logical connections among ideas in a con-ceptual network assured that the theory was testable. (Fararo, 1984, p. 152)

Page 22

Figure 8. Evolution of computer simulations of organizations.

(Thomsen et al., 1999)

As stated in the Introduction, the purpose of this dissertation was to report on a first attempt to simulate Parsons’ theory of action. Accordingly, the study did not seek to operationalize or transform into constructs everything Parsons wrote on the subject, but rather it functioned as a starting point of a single working simulation. Even applying that working simulation will leave much for future research. It was necessary to select the few constructs that formed the kernel of this simulation from all of the possible candidates.

At the outset these additional limitations have been identified in Table 2. Table 2. Additional limitations of the study

Origin of limitation Limiting action Parsons never intended his description of the theory of action to be granular enough to support simulation. He stated that his theory was not at the logico-deductive stage yet (Parsons, 1961c, p. 321). There have been other attempts to identify propo-sitions in Parsons’ work and to test whether they form a set that is logical in the sense that conclusions can be deduced from those propositions (Brownstein, 1982). Those other attempts have found gaps in Parsons’

It is not known why Parsons’ theory cannot be simulated unless and until there has been an attempt. On the other hand, Jacob-son and Bronson (1997) reported failure, and they are experienced, published soci-ologists and modelers. Also, in fairness, Parsons (Parsons, 1968a, pp. 77 ff) contra-dicted his own observation by providing some formalization that could have been interpreted as a beginning of a logico-deductive base.

Page 23

reasoning that Parsons acknowledged and does not apologize for because the theory was never intended to be "logical" yet. Parsons was a prolific writer. He has been roundly criticized for being difficult to understand (Selznick, 1961)11; (Kolb, 1962)12; (Bressler, 1961)13. One reason for the criticism is that Parsons would write a sentence and then in the next use different terms for what appeared to be the same thought. Was Parsons restating the previ-ous sentence but in a different way aimed at increasing our understanding through redundancy or repetition or was he saying something that was (slightly) different than the first sentence?14 Parsons’ writing style confounds understanding, which was a problem because the study sought a deep, detailed understanding so that it illustrated the theory by replicating the understanding inside a computer.

Identify a single work (Parsons et al., 1953a) and acknowledge the implications on generalizability.

One never knows when to stop trying to increase the fidelity of a simulation. This is a problem with all simulations. It is equivalent to the question of validation: when is the computer simulation suffi-ciently like the Real World to be trusted? One can always try one more interesting case, one more tweak that will increase fidelity.

Use the heuristic: can the computer simu-lation serve as a foundation for further research work where only incremental enhancements would be needed, not wholesale simulation changes. The simula-tion constructed here has instances of many of the important structures and functions, so that additional structures and functions can be based on those already represented.

It is tempting to label simulation as reduc-tionist, an especially unfortunate moniker

Keep the unit of analysis at the system, structure, and function levels. Do not per-

11 "It is a case of the Emperor's clothes. Is his [Talcott Parsons'] complexly textured raiment really there? Or is it all (or

largely) an illusion, a conjuration, a bad and costly joke?" p. 932 "The problem of arriving at a reassured assessment of Parsons' thought is greatly complicated by a remarkable obscurity of structure and style. Even those accustomed to abstract philosophical discussion find it a considerable chore to decide what is being said on any page, let alone also to assess its intellectual worth. I suspect that a great many sociologists, otherwise sympathetic to the need for general theory, have simply abandoned the effort." p. 932

12 "The essays ... establish beyond question ... the difficulty of understanding his [Parsons'] work at any but the most generalized level." p. 590 "... [T]here is concern with the obscurity of Parsons' language, the shifting meaning of some of his terms, ..." p. 591

13 ""His detractors have chided Parsons for a linguistic style which reads like pure hardtack." p. 149 14 For example: "This conception of the orbit of the action-process is integral to that of phase movement which will

figure prominently in our subsequent discussion. It is applicable both to the unit and to the system as a whole, the latter distinction being a matter of points of reference, not of the concrete structure of processes." (Parsons et al., 1953a, p. 164) In the second sentence to what does "It" apply? Orbit? Phase movement? Subsequent discussion?

Page 24

when applied to a holistic theory as Parsons purported his to be. The challenge was to maintain the holistic nature of the theory of action and simulate something that is whole. The next level down of this chal-lenge is that Parsons described the func-tional components of the theory of action in terms that looked like tokens that travel along wires (media of interchange) among nodes (functional prerequisites). Therefore, the simulation can have the appearance of rats in a maze because at some level that resembles Parsons’ description.

mit manipulation or reporting at the atomic (what Parsons calls the unit) level. This is consistent with Parsons' view that the unit act cannot be viewed by itself but rather in a much, much larger context.

Simulation often postulates a sequence of states through which the system passes. Simulation, then, presents the states that were encountered, but not all of the possi-ble states.15 That is, simulation cannot give the richness that a grammar or contingent approach could (Fararo, 1984, p. 146).

Noted as the nature of simulation vs. a pro-duction system (i.e., grammar) orientation (see Fararo & Skvoretz (1984) for an example of the production system approach, described above beginning on p. 32).

It may be worth mentioning that a significant limitation is that elements outside of Parsons' theory are outside of the simulation of it.

15 This is the same distinction in biology between ontogeny (an individual instance seen in nature) and phylogeny (all

possible instances for a species), between genotypes (the expression of genes found in an instance) and phenotypes (everything that is possible genetically).

Page 25

II. LITERATURE REVIEW The review of the literature is divided into parts. The first part examines the ques-

tion of what is a theory, what is a model, and how does simulation interact with them. The theory to be simulated is reviewed next, with an eye on a particular articulation from Parsons himself. The salient features are identified, as they formed the basis of the selec-tion of the subset of all constructs that shape the simulation. The review then shifts to the art and science of computer simulation applied to social science theories.

Theory, model, and simulation Simulation-building and theory-building are identical in their notional steps (the

steps are from Hanneman (1988)): 1. Define boundaries of the system. 2. Define the elements of the state space and partition the state space into

subsystems. 3. Describe the connectivity of the state space elements, and the forms of

relations among the states of the system. 4. Define the dynamic aspects of the relations among state space elements.

Simulation is a not the real thing, it is an imitation. In the instant case it is very far removed from anything real because the research here is a simulation of a theory and that theory has never been asserted to be related to reality: Parsons' theory is a frame of refer-ence. "In a certain sense, all theories about social action and interaction are simulations – theories are designed to mimic (albeit in highly selected and abstract ways) characteris-tics of real social action." (Hanneman, 1988) Baudrillard (1995) has coined the term simulacra for a copy without an original. One committee member remarked that this dis-sertation was like The Matrix, a popular movie that incorporates much of Baudrillard (1995), especially the question of simulacra, in it, where it is difficult for the viewer to tease out what is real and what is simulated.16 In fact, simulation is also the name of a social theory that addresses the problem that so much of the culture of developed coun-tries is simulated, not real, put there by advertising and other media (Cubitt, 2001). This dissertation does not use simulation in the sense of a stand-in or substitute, but rather in the sense of an animation or reification, a coming to life of something inanimate (in this case a theory). Again, the simulation is not the real thing, the real thing is Parsons' theory of action.

The theory of action Parsons traced the history of the development of the theory of action in Parsons

(1977a). He came to sociology from economics, starting in about 1930. "It gradually become clear to me that economic theory should be conceived as standing within some sort of theoretical matrix in which sociological theory was also included." p. 24. Parsons studied and tried to find the common threads among Alfred Marshall, a dominant English neoclassical economist (one of his students was John Maynard Keynes); Vilfredo Pareto, an Italian economist and sociologist; Max Weber, a German scholar who had ideas on the nature of modern capitalism and how to organize for economic gain; and Emile Durk-heim, a French scholar who, among many other subjects, wrote about the division of labor. Parsons' effort culminated in 1937 in the two-volume work (Parsons, 1968a;

16 One sight gag is when the protagonist, Neo, gives some contraband diskettes to "clients." He hides those diskettes in

a hollowed out edition of Baudrillard (1995).

Page 26

Parsons, 1968b). "The most immediate interpretative thesis was that the four – and they did not stand alone – had converged on what was essentially a single conceptual scheme. In the intellectual milieu of the time this was by no means simple common sense." (Parsons, 1977a, pp. 25-26)

The conceptualization that Parsons created flowed from his observation that the theories of Marshall, Pareto, Weber, and Durkheim had in common an action system, first suggested by Weber and then elaborated by Parsons; the result of the conceptualiza-tion was Parsons’ theory of action (Parsons, 1968a). That is, what these seemingly dispa-rate writers described in common was a system of actions, human actions that had pat-terns that could be described in accordance with a framework. Parsons has said that scientists of his era were informed by the progress in the conception of systems using mechanics and physico-chemistry (Parsons, 1977a, p. 27). In those disciplines one starts at the atomic level and defines what is meant by a "unit."

Accordingly, Parsons started by defining the "unit act." (Parsons, 1968a, pp. 43 ff):

(1) It implies an agent, an "actor." (2) For purposes of definition the act must have an "end," a future state of affairs toward which the process of action is oriented. (3) It must be initiated in a "situation" of which the trends of development differ in one or more important respects from the state of affairs to which the action is oriented, the end. This situation is in turn analyzable into two elements: those over which the actor has no control … and those over which he has such control. The former may be termed the "conditions" of action, and latter the "means." Finally, (4) there is inherent in the conception of this unit, in its analytical uses, a certain mode of relationship between these elements. That is, in the choice of alternative means to the end, insofar as the situation allows alternatives, there is a "normative orientation" of action. (p. 44)

Tension management and learning One of the functions of Latent Pattern Maintenance is tension management. Ten-

sion is the difference between what the inflexible, external environment demands and what the social system provides inside its boundaries (Parsons et al., 1953a, p. 212). Par-sons did not make completely clear the mechanism that Latent Pattern Maintenance used to manage this dynamic tension, but he did say that Latent Pattern Maintenance learned how to perform the function. In addition, he described classical conditioning at the par-ticular learning style (Parsons et al., 1953a, p. 226).

While there is an extensive literature on how organizations learn, there is very, very little of a quantitative nature; Dar-El (2000) informed the literature review of quantitative organizational learning. And what little existed of a quantitative nature was nearly the only literature on the actual mechanism, on how learning actually took place in an organization.

One line of study particularly stood out as applicable to this research because of its quantitative aspirations: Nembhard and Uzumeri (2000), Nembhard and Osothsilp (2001)17, and Uzmeri and Nembhard (1998). This line relied on Mazur and Hastie (1978), who found that learning was related to the accumulation of knowledge and there- 17 There is a small controversy about the results of this research (Jaber & Sikström, 2004; Nembhard & Osothsilp,

2004).

Page 27

fore an adequate model of learning has to account for accumulation. In essence Nembard and his collaborators studied possible descriptions of the gain in productivity due to learning in a factory and tried to fit it to a mathematical model, evaluating 11 models against actual factory floor data (Nembhard & Uzumeri, 2000). The authors used almost 4,000 data points to test the models, which included all of the popular ones, such as exponential, log-linear, and S-curve. One model fit best under a broad set of criteria, the three-parameter hyperbolic learning curve. It is described – and applied – in the Model section on page 58.

Place of Parsons' theory of action in sociology The purpose of this section is to place Parson's theory in the spectrum of socio-

logical theories of the time and to address a few of the many criticisms, particularly those applicable to the research described here.

• Action systems as unification. Finding a unifying thread in such diverse theories as those of Marshall, Pareto, Weber, and Durkheim was a breakthrough of major proportions. Later Parsons added Marx and Freud, no small accomplishment. The unification put Parsons on the map in sociological theory and he spent the rest of his professional life refining the theory of action.

• Structural-functionalism. Structural-functionalism is a school of thought within sociology that concentrates on describing social systems by describing their (static) structures and (dynamic) functions. Structural-functionalists tend not to address questions of how the structures or functions arise nor whether some are better than others.18 In the framework of Burrell and Morgan (1979), structural functionalists are more interested in the sociology of regulation than of radical change, more interested in objectivity than subjectivity, and "tend to assume that the social world is composed of relatively concrete empirical artefacts and rela-tionships which can be identified, studied and measured through approaches derived from the natural sciences. The use of mechanical and biological analogies as a means of modeling and understanding the social world is particularly favoured." p. 26. Parsons(1977b) wrote:

"I well remember at a meeting of the International Sociological Associa-tion, held in Washington D.C. , in 1961, [Robert] Merton very cogently made the point of objecting to the phrase 'structural-functionalism.' He particularly did not like having it labeled an 'ism' and suggested that the simple descriptive phrase "functional analysis" was more appropriate. I heartily concur in this judgment." "The two concepts 'structure' and 'func-tion' are not parallel. … [T]he concept 'structure' does not stand at the same level as that of function, but at a lower analytical level. It is a cog-nate with the concept of 'process,' not function. Sometimes, the levels are consolidated or fused by reference not to functions but to functioning. From this point of view, the verb form may be considered to be a synonym for process. We do not wish to hypostatize structure. It is any set of rela-tions among parts of a living system which on empirical grounds can be assumed or shown to be stable over a time period…. Thus, … the concept

18 These questions are not usually considered part of positivism, of which structural-functionalism is a school.

Positivists try not to go beyond what can be verified, lest their work be considered metaphysics and religion (Keat & Urry, 1982, p. 5).

Page 28

'function,' unlike that of structure and of process, is not a rubric in terms of which an immediately empirical description of a set of features … can be stated. It is, rather, a concept that stands at a higher level of theoretical generality and is more analytic than either structure or process. Its refer-ence is to the formulation of sets of conditions governing the states of liv-ing systems as 'going concerns' in relation to their environments. These conditions concern the stability and/or instability, the survival and/or probable extinction, and not least, the temporal duration of such systems." [italics in original] (pp. 100-103)

It might be worth mentioning that the whole idea of functional analysis and its related synonyms had come into question in the heyday of Parsons and his collaborators. The central issues were questions of what is a theory, does it have to be empirically veri-fiable or can it be a framework, a naming of the parts. Davis (1959) argues that investi-gating the functions and functioning of a social system is not a special method, does not need a special method, and is not a school of thought.

• Homeostasis. The structural-functionalists have been criticized for postulating sta-ble structures, for not taking account of social revolution, of a set of norms that aim to upset the status quo. Clearly, Parsons admired and sometimes quoted biologists describing homeostasis, the dynamic balance of elements inside an organization/organism and balance of the organization/organism with respect to changes in its environment (Parsons, 1977a, p. 28). While the structural-functionalists do not, indeed, emphasize upsetting the legitimization mechanism of social entities, they do not obviate it either. Moore (1959) wrote, "I have come to the personal conviction that for most purposes the equilibrium model of social systems must be abandoned, as leading to too much distortion, particularly in treating change as external, accidental, and in any event regrettable." (p. 718) A balanced discussion, relying on cybernetics (à la Ashby (1956)) and the kind of control Parsons characterized, can be found in Cadwallader (1959). Parsons him-self (1977b) wrote:

"[Functional analysis] has nothing essentially to do with judgments about the specific balances between elements of integration in social systems and elements of conflict and/or disorganization. … Biology does not have two basic theoretical schemes, a theory of healthy organisms and one of pathological phenomena in organisms, but health and pathological states are understandable in basically the same general theoretical terms. … A related polemical orientation is the claim frequently put forward that 'functionalists' are incapable of accounting for social change: that is, their type of theory has a built-in 'static' bias. This is also entirely untrue. If we have any claim to competence as social scientists, we must be fully aware that there are problems both of stability and of change, as there are prob-lems of positive integration and malintegration." (pp. 108-109)

• Problem of concreteness. The theory of action is an abstraction, a made-up frame-work, a world view, a way to interpret and make sense of (empirical) phenomena. Its descriptions are not, in Parsons' terms "out there" (1968a, p. 46), but rather are mental constructs. In particular they are analytic (deduced by theory and logic), as

Page 29

opposed to synthetic (induced by observation), referring to a dichotomy described in a work about which Parsons often expressed admiration, Critique of Pure Reason (Kant, 1896). It is important, Parsons reminds us, relying on Alfred North Whitehead (1927), not to (mis)place concreteness on the abstractions (Parsons, 1968a, p. 29); the theory of action cannot itself be observed, but rather what is observed in the world (what is "out there") can be described in terms of the theory of action. Therefore, it would not be appropriate to validate the simulation of the theory by world experience, but rather by careful comparison to the text of the theory.

The bad news As mentioned above, particularly in and near Table 1, p. 2, there are criticisms

and critics of Parsons. While the work described here took Parsons' theory as-is, warts and all, without making a commitment to its veracity or even efficacy, it is useful to expose at least a window on the counter-arguments to the theory of action. Udy (1960) characterized the normalcy of criticizing Parsons, "Certain criticisms of Parsons' work have become virtually traditional…. [T]heory [of action] is by and large equated either with taxonomy or with functionalist arguments as to the requisite character of categories. There is an almost complete absence of propositions containing variables." Likewise, Bressler (1961) observed,

"Parsons has transformed the rhetoric of sociological discourse, and it has long since become de rigeur for every sociologist to strong opinions about his contri-butions to sociological theory. In fact, given the obstacles created by what D. A. Sprott has been pleased to call Parsons' "spritely" prose, it would not be at all sur-prising if Parsons had rather more critics than readers." (p. 149)

Perhaps the most succinct critique was Berger and Zelditch (1968), which took Parsons to task on four grounds in 4+ pages. In the context of a book review, their ques-tions were: (a) had there been an improvement in confirmation status, have there been empirical studies confirming the theory of action; (b) was there increased rigor in the framework or its arguments, had it become more logically structured; (c) had the theory become more precise, more accurate; and (d) had the scope increased in order to make the theory more general?19 In every case the authors believed that Parsons failed. They went on to ask what was the importance of Parsons, why was he (still) read. They con-cluded that there were several reasons: the admiration for the ambitiousness of his enter-prise and as an "inexhaustible source of ideas." p. 450

Locating this work within all of Parsons' A small subset of everything written about the theory of action by Parsons and

others was sought that could form the basis for animating the theory. Accordingly, in harmony with the state of simulation as reviewed in the next large section, rich descrip-tions were sought that illustrated the structure and function of the theory of action. That is, detailed descriptions of static structures and dynamic (time-varying) functions or proc-esses were sought. When found then the same processes that Sastry used were applied, namely parsing them into simulation constructs. In a sense this operation is a culling of a

19 The comparison to Thorngate's one-armed clock is palpable.

Page 30

specific description of the theory of action in order to find the passages best suited to the narrow purpose of simulation.

As stated on p. 48, one writing stood out with respect to this search: Parsons, Bales and Shils (1953a). The volume in which the chapter appears, Parsons, Bales and Shils (1953b), is an historical account of the development of the theory of action, and Parsons, Bales and Shils (1953a) is the last chapter, therefore the most recent in the collection. The chapter traces the path (Parsons et al. (1953a, p. 167) called it an orbit) of energy moving among the four functional prerequisites in accordance with the pattern variables. It is a step-by-step description of how energy enters a social system and trav-erses the four units, possibly transforming the unit or itself or attributes of the social sys-tem as it is passed from unit to unit.

Table 6, below on p. 72, presents in some detail the description of the transit of energy through a social system described in Parsons, Bales & Shils (1953a) and the corresponding elements of the simulation.

Models Young journalist (YJ): Why do you work with models? Why don't you work with

the real world? Albert Einstein (AE): Are you married, young man? YJ: Yes. AE: Do you have a picture of your wife? YJ: (Fetches his wallet and digs around, finally producing a photo and handing it

to AE.) Yes, here. AE: (Looks at the photo for a moment and hands it back.) She is rather small.

- Ronald W. Clark. (1971). Einstein: Life and times. New York: World Publishing. Model ships appear frequently in bottles; model boys in heaven only. Model ships are copies of real ones. Asked to describe a ship, we could point to its model. A model boy, on the other hand, having no earthly counterpart, is everything a boy ought to be. (Brodbeck, 1959) [italics in original]

Model is sometimes used in the sense of a replica, a non-verbal description of the thing being modeled. A replica gives no new knowledge, and in that sense is not scien-tifically interesting. On the other hand, models can help us understand more about the thing modeled.

The term for the similarity between a thing and its model is isomorphism. In order for there to be an isomorphism two conditions have to hold: a one-to-one correspondence between the elements of the model and the elements of the thing modeled, and certain relations must be preserved. If, besides structure and relations, the model works the same way that the thing does then the isomorphism is called complete.

If, for instance, a model of a steam engine is also steam propelled, then the iso-morphism is complete. The similarity or isomorphism of a planetarium with heav-enly bodies is not complete. All the planets and their moons and the sun, together with their spatial relations to each other, are duplicated. But the motions of these bodies across the hemispherical ceiling are not cause by gravitational attraction. (Brodbeck, 1959)

What is the difference between a model and a drawing of a model? In order to pursue this, some distinctions must be made. "The language of science … consists wholly

Page 31

of declarative sentences." (Brodbeck, 1959) The sentences contain two kinds of words in them: names for characteristics or attributes or events, and for relations among them. Characteristics are the name of some state, such as grass that is in the state where its color attribute has the value of green. Relations require at least two individuals or operands, such as before, faster, and between. Some terms, such as north, fast, and first, appear to be about a single individual, but are in fact in relation to some standard, and therefore about at least two elements. While sentences may contain only attributes and relations, the subject matter, the content, differs from one scientific discipline to another.

Sentences may have meaning because of the facts, the content, or they may have meaning no matter what their content. For example, "He is tall and he is blond" is of the form "X is A and X is B." One can speak of the truth value of either sentence, but in our example the truth value of the first will be the only one we would be able to ascertain, based on whether it were true that the subject was both tall and blond. In other words, the truth value of the form cannot be known unless we know the values of X, A, and B. If we can know the truth value of a form, then it is called a logical truth because it is true for all possible values of the variables; it is also called tautological or analytic. An example is X = X, because this is truth for all possible values of X in the sense that we usually give to the equal sign. "Sentences whose truth depends upon their descriptive words as well as on their form are called empirical statements, or also contingent or synthetic." (Brodbeck, 1959) [italics in original]

Perhaps the most common class of logical truths is arithmetic. All statements in arithmetic are true by definition, by form, not because we examine the subject matter of the sentences and from them deduce the truth value.

A concept is a term referring to a descriptive property or relation. A fact is a par-ticular or specific thing, characteristic, event, or kind of event. To state a fact is to state that a concept has an instance. Facts are significant when they are connected with other facts to form generalizations or laws. "A law states that whenever there is an instance of one kind of fact, then there is also an instance of another. Laws, therefore, are empirical generalizations." (Brodbeck, 1959) A theory is a deductively connected set of laws. Some of the laws, called axioms or postulates of the theory, imply others, called theorems. Axi-oms are presupposed, their truth is taken for granted, at least for the purposes of the exer-cise of seeing what else is true of they are.

"Two theories whose laws have the same form are isomorphic or structurally similar to each other. If the laws of one theory have the same form as the laws of another, then one may be said to be a model for the other." (Brodbeck, 1959) [italics in original] How does one know if one theory is the model of another? One puts the second into one-to-one correspondence with the first. If the forms are the same and the relations are pre-served, then they are isomorphic. That is, one translates the form of the second into the first and then ascertains whether the truth of the relations is preserved. If it is, then the translation demonstrates the isomorphism between the theories, and the second can be said to be a model of the first.20

"It is all too easy to overestimate the significance of structural isomorphism. The fact that all or some of the laws in one area [of discourse] have the same form as those of

20 In fairness, the relation between the two theories is completely symmetric, one could be the model of the other and

vice versa.

Page 32

another need not signify anything whatsoever about any connection between the two areas." (Brodbeck, 1959) The example she gives is all things that can be ranked and measured: they are structurally isomorphic with arithmetic addition, yet that is quite pos-sibly the only thing they have in common. Another example is "taller" and "smarter" because they are both transitive (p. 394). Accordingly, we should not be misled by that structural isomorphism is anything but a necessary condition for complete isomorphism. The relations have to hold as well for there to be complete isomorphism.

The flow in the research described here is from (1) Parsons theory to (2) a model constructed by the researcher that is an incomplete isomorphism to (3) a simulation con-structed by the researcher that is both an extension and contraction of the model. That is, the model redacts elements, structures, and relations from the theory, and then the simu-lation further reduces the elements, structures, and relations, and also adds some ele-ments, structures, and relations that are neither in the model nor in the theory.

Formalization In lay terms, formalization is an expression of something so that it can be reason-

ed about. The most common formalization is mathematical, but there are other forms, too. Two others that will be dealt with here are logic, which is more officially called first order predicate calculus, and production systems. The place of formalism is that simula-tion can also be a formalism because it represents an opportunity to reason. Production systems

Approximately how many sentences are there in English (or any natural language; natural languages are the ones we speak and read)? The short answer is: infinite, approximately. How do we learn an infinite thing? How do we teach one? We look for what is finite about it, and in the case of languages, as with many other things, it is the (list of) rules that is finite. The collective rules of the construction of a language is called grammar. The rules can be viewed either as specifying what is legal to read or what is legal to construct, generate. That is, we can hand a sentence to a grammar and ask "Is this sentence in the language, that is, is it properly formed according to the rules?" Or we can "run" the rules of a grammar and generate correct sentences in the language. The rules enable us to say, "That is not a sentence," or more properly, "That is not a sentence that is allowed by the rules of grammar."

Note that the rules at this point evaluate or generate content-free sentences. The rules (of grammar) have nothing to say about the content, only about the form. The form is called syntax. That is, grammar describes syntax, without regard to (truth) value of the words.

The appearance of sentences in a language is guided by grammar and by what symbols and symbol combinations are valid. The symbols (e.g., letters of the alphabet) are the lexical aspect of language. In principle there are two types of symbols: terminal and non-terminal. Terminal are the ones we read, that are being read right now. Non-ter-minal describe constructs in the language, such as sentence, paragraph, verb-phrase, sub-ordinate clause, genitive case, pronoun, etc. In English, as in most natural languages, the non-terminals are also terminals, so it is a bit confusing. But when describing artificial languages there is attention paid to the difference between sentences in the language (terminals) and terms used to describe sentences in the language (non-terminals).

Fararo and his collaborators have developed a grammar of social actions (Axten & Fararo, 1977; Fararo & Skvoretz, 1984; Skvoretz & Fararo, 1980; Skvoretz, Fararo, &

Page 33

Axten, 1980) and symbolic interaction (Skvoretz & Fararo, 1996), drawing on the work of Nowakowska (1973) and Hayes (1981). According to Fararo et al. one of their inspira-tions was Harré (1972), where there are descriptions of rule-condition and role-condition forms. Each of them can be thought of as "if … then" statements: if the condition is true then this rule is executed by the role that matches the role-condition. That is precisely the structure of grammar rules: if the right hand side of the rule matches the state of the parsing, then the state is changed to the value of the left hand side, and the matching pro-ceeds until that are no more matches possible. If the final match is what is called the dis-tinguished symbol21 then the whole sentence or social action is recognized and declared valid, otherwise the sentence/social action is not one that is described by the grammar and is therefore noted as impossible or an error.

Another inspiration was Heise (1979): "[A] simple event is conceived as a syntactically ordered conjunction of cognitive elements (usually culturally defined) des-ignating actor, act, and object." Heise's method of "processing" situations that give rise to actions is to scan actor-object combinations. When a match it found, the associated action is executed. This is precisely the steps that Fararo et al. take in their grammar approach (loc. cit.).

Fararo and his colleagues have created descriptions of valid constructs such that the descriptions can be used like any other grammar, either to assess the validity of an existing "sentence" (that is, social action) or to generate valid social actions. The most important aspect from this dissertation's point of view is that the application of much of Fararo's work was Parsons' theory of action. The grammar described social actions that enacted the theory of action. Logic

Brownstein (1982) has formalized the theory of action, too. He used first order predicate logic, the same axioms and method used in high school geometry and trigo-nometry proofs. In some-odd 300 pages Brownstein in the standard language of logic tries to reconstruct the propositions Parsons intended. His conclusions are a bit discour-aging.

Though his [Parsons'] scheme calls for functional explanations, precise, explicit specifications of functional relations are not particularly salient. Moreover, sub-stantive propositions, definitions, regulative principles, preliminary redescrip-tions, and the like are rarely distinguished, thereby rendering it difficult to assess … its conceptual health. … For a scheme with as many fundamental conceptual disorders as Parsons', conceptual analysis becomes of primary concern …. Grave difficulties attend the conceptualizations of the pattern variables. … Assessment … leads to the conclusion that a proper analysis of action in Parsons' terms demands a revision of the basic analytical framework which Parsons has con-structed. (Brownstein, 1982)

Dubin (1960) used a form of logic, too. He looked at the pattern variables at the personality level of analysis. He succeeded in enumerating all of the possible combina-tions of the pattern variables at that level and offered that the choice among them in a particular instance of action might be based on probability. In other words, Dubin used

21 This non-terminal is usually called "sentence" or in our case "social action."

Page 34

the logic of arithmetic to reason about the number of states that are available for an actor to get into.

Time If one skims the literature on the subject of "social time," one finds complaints everywhere about the neglect and marginality of the time problem in sociology, formulated concisely by Kurt Lüscher in the title of an essay: "Time: A much-neglected dimension in social theory and research." (1974) … However, even more decisive, in my opinion, for the impression of marginality and neglect is the minimal theoretical basis of many of the available studies. … Many authors lose themselves entirely in the momentum of their subject by making philosophical, anthropological and everyday observations without even beginning to achieve conceptual precision and a categorization of time within a sociological theory. (Bergmann, 1992)

Time is the missing variable in modern sociological analysis. … Most sociologists treat time as a contingent feature of their research, rather than a topic in its own right. … Indeed, sociology can be almost said to be "time free." As emphasis has been placed upon developing state perspectives – such as structural-functionalism or system theory – then temporal analysis has been largely ignored. The socio-logical research process has been 'synchronic rather than longitudinal'; that is, it has stressed the enduring features of structure rather than the flux and dynamics of change. The dominant research paradigm has been one favoring 'slice-through-time' investigations, and in particular studies whose conclusions are based on one-shot statistical correlations. In short, time has tended either to be excluded as an explanatory variable, or else introduced only in post hoc justification. (Hassard, 1990) [italics in original]

Greater emphasis has been given to statics than to dynamics in most social sci-ence theorizing. And, while comparative static analysis is a necessary and impor-tant task, too much emphasis can deflect attention from other important theory-building tasks. To the extent that social scientist's theoretical activities seek to build explanations of phenomena, rather than descriptions, they must focus on causal processes that occur over time. (Hanneman, 1988)

The typical quantitative theory in sociology is a function or formula, usually of the form that to compute a value for some dependent variable Y there is an arithmetic combination of independent variables, Xi. The changes in the Xi over time is not usually considered, and therefore there is but a single Y in time.

Abbott (2001), a reprint and possible update of Abbott (1988), makes the point more strongly. He claims, using many detailed examinations of published sociological studies, that the equation described above and ones like it are more than mathematical tools, that they influence sociologists to neglect time, to ignore the time path of actors as they interact with themselves and their environments. He argues that the equations that

Page 35

relate variables linearly have "infected" thinking so that some social science researchers believe that reality is generally linear22. He bases his argument in relevant part on:

1. Fixed entities with attributes. Linear equations, such as those used in regression and structural equation modeling, assume fixed entities that have attributes. The entities are fixed in such equations and the attributes can change value. This clearly assumes that the existence of the entities is stable over the modeling period. Oddly enough, many of the subjects studied in sociology are not fixed, such as occupations, roles, social move-ments, and organizations. Abbott asks us to compare this fixed nature with its most com-mon alternative, central-subject/event model.

A historical narrative is organized around a central subject. This central subject may be a sequence of events, a transformation of an entity or set of entities into a new one, or indeed a simple entity. The central subject includes or endures a number of events, which may be large or small, directly relevant or tangential, specific or vague. (Abbott, 2001) [parenthetical material omitted]

Fixed entities ignore changes that occur due to birth, death, combination, division, and transformation. These changes will need to be simulated in the present research because Parsons describes them in his theory.

2. Monotonic causal flow. The right sides of (linear) equations do not differentiate the value over time or in time that each variable would contribute to the dependent vari-able. They are all equal throughout all of time. That is, each right-hand side variable is equally relevant at all times. There is no contingent time. Perhaps worse, the time horizon for all variables in a single equation is identical. That is, if we are trying to measure the effect of several factors on an outcome, all of the factors would have to be measured over the same time scale and the outcome would have to be expressed in that time scale, too. One can see how this could be a problem in the theory of action on several counts: (a) actions happen on a smaller scale inside the organization than are sensed outside it, and (b) there may be a different scale altogether in each functional prerequisite (that is, there is nothing a priori to suggest that the time scales inside each functional prerequisite are commensurate).

3. Univocal meaning. In linear modeling each variable can have only a positive or negative effect, not one and then another under different situations. But (Abbott, 2001) illustrates many cases where a variable may have at first a positive effect and then later on a negative one.

4. Absence of sequence effects. The order of events does not affect the values of variables in a linear combination, so that the actual time story or path or trajectory or un-folding is completely lost using normal statistical tools. In the present research order matters a great deal, because the timing of an external event has a great impact on the organization's response, in light of its history to date of responses.

While the picture of taking time into account in social setting is a bit dark, there are new methods for dealing with time in structural equation modeling, e.g., (Collins & Sayer, 2001; Hamagami & McArdle, 2001),23 and there has always been auto-regressive

22 The term linear can have many meanings. The shortest one for our purpose is that a change in the value of an

independent variable causes a proportional change in the dependent variable. 23 To indicate the extent that time is rarely accounted for in sociological studies and make the point a bit closer to

home, Ralph O. Mueller is the chair of the George Washington University Graduate School of Education’s

Page 36

integrated moving average time series analysis (ARIMA, also called Box-Jenkins), but it has been applied almost exclusively to economic data until recently (McCleary & Hay, 1980). And ARIMA usually addresses only a single entity and a single variable (Abbott, 2001).

There is some modeling of time in sociology, including longitudinal studies, such as Durkheim’s famous one of suicide (Durkheim, 1951). There is some new interest in time, for example, a 2002 special issue of Academy of Management Journal (Barkema, Baum, & Mannix, 2002).

The treatment of time, above, contains a subtlety: it refers to a combination of clock and social time. Clock time is what one gets from calendars, clocks, and other time pieces. It is the time in physics, the one with which derivatives are taken; it is even rela-tivistic time in the Einsteinian sense. Social time is socially constructed and includes such diverse activities/entities as lunch time, waiting, graduation, career progression, and stages of life. Which type drives Parsons' theory? It must be social time because there are none of the attributes of clock time in Parsons' description, such as uniformity of cadence. How can social time be simulated? Oddly, since it is socially constructed a uni-form cadence can simulate social time as long as the social constructions demarking events are present. After all, calendar time is an adequate backdrop for social time.

In the present case, here are some examples of Parsons' social constructions of time in his theory of action: energy arrives at the system boundary at a particular mo-ment, a functional prerequisite consumes time to perform its function, energy passes (in a time interval) from one functional prerequisite to another in accord with the cybernetic hierarchy, and a message is transmitted across a medium of interchange (in a time inter-val). In fact, Parsons himself recognized the importance of time, "The first important im-plication is that an act is always a process in time. The time category is basic to the scheme." (1968a, p. 45)

The simulation of this social time is simply the ticking of a notional clock whose moments are normatively agreed to mark forward time in an interval small enough to permit the shortest social event to transpire.

Process In the last decade a number of writers have proposed narrative as the foundation for sociological methodology. By this they do not mean narrative in its common senses of words as opposed to number and complexity as opposed to formaliza-tion. Rather, they mean narrative in the more generic sense of process or story. They want to make processes the fundamental building block of sociological analysis. For them social reality happens in sequences of actions located within constraining or enabling structures. It is a matter of particular social actors, in particular social places, at particular social times.

In the context of contemporary empirical practice, such a conception is revolutionary. Our normal methods parse social reality into fixed entities with variable qualities. They attribute causality to the variables—hypostatized social characteristics—rather than to agents; variables do things, not social actors. Sto-

Department of Educational Leadership. He is also the author of an introduction to structure equation modeling (SEM) (Mueller, 1996) that does not mention the problem of time in the general linear model nor the newer approaches to modeling time in SEM.

Page 37

ries disappear. The only narratives present in such methods are just-so stories jus-tifying this or that relation between variables. Contingent narrative is impossible. … While action and process have largely disappeared from empirical sociology, they are by contrast central to much of sociological theory, both classic and recent. (Abbott, 2001), reprinted from (Abbott, 1992)

Too much emphasis in empirical research ... has been placed on the study of indi-viduals rather than social systems, and on single-time points in these systems rather than on their continuing process. [Editors' note] Despite the early preoccu-pation of sociologists with research on social stability and change, much of to-day's research is neither dynamic nor oriented to social systems. [italics in origi-nal] (In a volume honoring Talcott Parsons, Riley & Nelson, 1971, p. 407)

Lave and March (1993) advise modelers to "think process." By this the authors meant that one should seek to describe, explain, predict the unfolding of the interaction of social forces and the emergence of the resulting outcomes. Process has been variously defined as "change that follows a stable pattern long enough for us to recognize continu-ity, transient as the continuity itself may be," "a series of progressive and interdependent steps by which an end is attained," "the interweaving of invariance and variance," "a becoming of continuity," and "a tension between linear succession and sequential recur-rence," as summarized in Abbott (1989).

Process is related to time in a straightforward way: the steps in a process are described from a time perspective. "It is clear that process is inherently temporal." (Rowell, 1989) Time in this sense may be an ordering, such as before, during, or after. Or, "when this happens, then that happens." Or it may be in terms of delays, such as Act B happens about six months after Act A. Or it may be any other indication of time or timing. And it may be necessary to mention that time in the process meaning is social time, not necessarily clock time, that is, how time is sensed, not how it clicks off of an absolute clock.

The intuition is that what happens in a process is that events occur and something inside those events trigger changes in the system state that in turn cause other events to happen. In this way, process, system state, and events are related as follows:

State1 State2

State3

Process

Event1 Event2

Figure 9. Relationship among process, event, and state (notional).

Time is what travels on the lines in the direction of the arrowheads, indicating that State 1 happens before State2, etc. State is the value of all of the variables in the system. In principle, then, a system rests with its variables having some fixed value, then an event happens that changes the values of some variables, putting the system into a different

Page 38

state. The event may consume time, the state change may consume time, and the interval between them may consume time. To the extent that there is a pattern in the transition of states and events, we call it a process and usually give it a name (e.g., adoption of a new idea). This is pure construct; there is no commitment that such juxtaposition of event and state exist independently in nature (this is one meaning that Parsons makes of "analyti-cal."). Sometimes the sequence of state/event pairs (also called feeling-activity states (Bergmann, 1992)) is called history, trace, story, time track, course of events, life cycle, narrative, enchainment, or trajectory. Some sociologists call it cause (particularly those committed to statistical methods, especially structural equation modeling), but we do not.

And process is linked to structure: "theories that focus on 'process' or social dynamics must have (at least implicitly) models of structure embedded in them." (Hanneman, 1988)

The research described here is the process kind. It attempts in very crude and rough terms to explain, among other things, how, for example, Latent Pattern Mainte-nance impacts the Adaptation function with respect to which energy it (Adaptation) allows into the system. There are many steps in the flow between the entry of energy into a system and the response of Latent Pattern Maintenance, and Parsons explains them notionally. The simulation described here attempted to imitate and animate that flow, to cause the stand-in for Latent Pattern Maintenance to react to perturbations of energy entering and flowing through the organization.

The process view imposed a considerable burden on the researcher because much has to be mechanized, in comparison with, say, another handy tool of sociological research, structural equation modeling, in which the researcher collects and feeds num-bers into a "black box" computation engine and interprets the stream of numbers that come out. There is far less of a burden to construct the intricate relations among the moving parts and how exactly each one interacts and impacts the other. There is no "answer" in a simulation: the simulation itself is the answer!

Simulations of social systems Computer simulations of social and organizational systems is not new (Bronson &

Jacobsen, 1986; Bronson, Jacobsen, & Crawford, 1988; Burton & Obel, 1995; Carley & Prietula, 1994; Coleman, 1965; Conte et al., 1997; Coyle, 2000; Cyert & March, 1963; Cyert & March, 1992; Epstein & Axtell, 1996; Gilbert & Conte, 1995; Gullahorn & Gullahorn, 1963; Hamblin, Jacobsen, & Miller, 1973; Hanneman & Patrick, 1997; Hanneman, 1988; Ilgen & Hulin, 2000; Jacobsen & Bronson, 1995; Jacobsen & Bronson, 1985; Jacobsen & Bronson, 1987; Jacobsen & Bronson, 1997; Jacobsen et al., 1990; Lane, 2001; Leik & Meeker, 1995; Lin, 2000; Markley, 1967; Moss, 2000; Phelan, 1995; Prietula et al., 1998; Rasmussen, 1985; Samuelson, 2000; Sastry, 1997; Senge, 1990; Thomsen et al., 1999; Tuma & Hannan, 1984). Even the use of simulation games to illus-trate concepts and let sociology students try their hands at applying what they already have learned by more passive means, such as reading and discussion, are not new (Simulation and Gaming and the Teaching of Sociology, 1997; Coleman, 1965; Cross, 1980; Dukes, 1975; Hanneman & Patrick, 1997; Hanneman, 1988; Markley, 1967; Pfahl, Laitenberger, Dorsch, & Ruhe, 2003). There is an annual conference on computational and mathematical organization theory, including social systems simulation (Computa-tional, Social and Organizational Science), several professional societies (the American

Page 39

Sociological Association section on mathematical sociology,24 European Social Simula-tion Association, North American Association for Computational Social and Organiza-tional Science, International Simulation and Gaming Association), several journals (Journal of Mathematical Sociology, Computational and Mathematical Organization Theory, and the Journal of Artificial Societies and Social Simulation), a university research program (Centre for Research in Social Simulation at the University of Surrey in the UK), and a LISTSERV (SimSoc).

Why use simulation to study a system? Fishman (2001) lists eight reasons: 1. Enables an investigator to organize his/her theoretical beliefs and empirical

observations about a system and to deduce the logical implications of this organization.

2. Leads to improved system understanding. 3. Brings into perspective the need for detail and relevance. 4. Expedites the speed with which an analysis can be accomplished. 5. Provides a framework for testing the desirability of system modifications. 6. Is easier to manipulate than the system. 7. Permits control over more sources of variation than direct study of a system

allows. 8. Is generally less costly than direct study of the system. In addition, Fishman (2001) lists some technical attractions of simulation: 1. Compress time so that years of activity can be simulated in minutes. It can

also expand time so that detailed interaction can be seen and analyzed. 2. Identify and control sources of variation in order to postulate the relationships

among the dependent and independent variables. 3. Correct operation can be at least subjectively assessed during the execution of

the simulation by stopping time and examining the state of the system without impacting the system under observation.

4. The state of the simulation can be stored for later analysis and then replicated with the same initial conditions, enabling a kind of experimentation that is impossible in the real world.

Discrete event simulation Discrete event simulation (DES) was created in the 1960s to address a set of

problems for which there were no closed form equations that could be solved. The prob-lems were an area of operations research called queuing theory, the study of waiting lines. In fact, queuing theory addressed a number of related concerns that were growing in importance, everything from how long to make left-turn traffic lanes to how many of those expensive shopping carts to have in a grocery store. The primary application of queuing theory was a particular type of manufacturing capability called job shop. A job shop is a facility that makes custom parts, not a full-scale production line. Every major manufacturing plant has a job shop and because of the custom nature of its work and that it fills in for unexpected/unplanned incidents on the assembly line floor, it is a challenge to plan its work. Many so-called dispatching schemes were created – including shortest operation time first, longest first, prioritizing those that waited longest – and needed to be tried. But without a formula to solve it was going to be tedious. And part of the pressure

24 Mathematical sociology is a topic much larger than simulation, but simulation is included in its ambit.

Page 40

on the solution was demand for a surge manufacturing capability in a US defense build-up for the Cold War.

And the advent of the Internet, then called ARPANET, presented questions about how big the computer storage on the network had to be in order to hold messages in the event of transient outages, and was it better to have a few long messages or lots of short ones.

The first DES systems were used by manufacturing, transportation, and telecom-munications engineers. Then Leonard Kleinrock, a UCLA engineering professor who pioneered much of the design of ARPANET, analytically solved many of the queuing problems in closed form (Kleinrock, 1975-1976) and some of the pressure to simulate waned.25 Kleinrock’s formulæ assume that inputs arrive at a random rate according to some distribution and are serviced/ processed/transformed at another random rate, possi-bly according to a different distribution; that is, that there is a probabilistic element to the operation of the systems under study. DES views the world as compartments among which items and information flow. The items have to be "born" as they cross the boundary into the system, then are trans-formed or processed or serviced, and then possibly consumed, and finally they exit the boundary of the system and in effect "die." This view is sometimes called process, esp. by social science researchers (for example, (Lave & March, 1993)) who are trying to differentiate themselves from others who take a more static view.

There are two approaches to DES: event-scheduling and process-interaction (Fishman, 2001). At the outset it is important to understand that the results are the same independent of the approach, but during simulation construction there is a trade-off between simulation simplicity and simulation control depending upon which approach is used. "Every discrete-event system has a collection of state variables that change values as time elapses. A change in a state variable is called an event" (Fishman, 2001)[italics in original]. The simulation is thought of, comprised of, a set of events, such as, in this research, energy appears at the boundary of the system, Adaptation filters are altered based on the tension between internal stability and external energy level, if affective energy is present then it takes priority over affective neutral, etc. In this conceptual scheme, "each event contains all decision flows and state variables. Simulation is the execution of a sequence of events ordered chronologically on desired execution times. No time elapses within an event" (Fishman, 2001).

In the process-interaction approach the focus centers on the processing or trans-forming entities, those parts of the simulation that take inputs and transform them. The approach provides a sequence of activities (events) in a time order, other terms for which are flow and process. In other words, the process-interaction approach concentrates on the time history of the transactions and their transformers; it is not a "disconnected" list of events that change the state of the simulation.

The research reported here uses the process-interaction approach as a way to trace the time history of energy as it transits the organization and the time history of the organization as it responds to the energy. This approach was selected because it more

25 This researcher may have written the first discrete event simulation program in a simulation language in Southern

California, in 1965-1966, and was a graduate student at UCLA in the department of and at the time of Kleinrock’s work.

Page 41

closely follows Parsons' style of description in which flows are described in a series of time-related steps.

To reiterate the baking example in the Methods section, p. 47, what would happen if dough could be formed into loaves more quickly than the time it took to cook the loaves in a batch? If the average rate at which loaves were created exceeded the average cooking time then an infinite queue would grow in front of the ovens. If the average rate at which loaves are created is about the same or lower than the rate at which loaves are cooked, then on average any queue that would form would be finite and a queuing theory formula can tell us how long it might be at its maximum. Formulæ could also be em-ployed to evaluate whether there should be multiple slow ovens to compensate for fast loaf making, etc.

In terms of simulating the theory of action, the transactions to be "born" are units of energy, and this researcher thinks of them as news, such as a new idea. The DES com-partments would be Parsons’ functional prerequisites, the processing or transforming would be what happens inside each function (e.g., scanning the environment in Adapta-tion; setting goals and allocating resources in Goal Attainment; recruiting and training new staff, and integrating new processes in Integration; and establishing the filters by which sense is made in the other functions in Latent Pattern Maintenance), and the dying would be what happens to the imported energy after the last function in the flow, Latent Pattern Maintenance, has responded. And the flow would be of (a) energy, and (b) mes-sages along the paths of the interchange media among the functions. If the rate of arrival of energy and other interchange media exceeded the rate by which it could be processed then queues would grow between the producer and consumer. Such attention to rates and queues, while an integral part of DES, is absent from Parsons’ conceptualization and therefore writing. This may be an important indication that using DES is inappropriate for simulating the theory of action, as prominently mentioned in the Limitations section, p. 20.

Without loss of generality, the arrival rate of news and service rates of the func-tions are set to be fixed amounts, so the simulation here is deterministic. The reason is that Parsons gives no insight into the rates, so the assumption of randomness, while more realistic in terms of the real world, would only reflect invention by the researcher in Par-sonian terms.

The researcher could only find two applications of discrete event simulation applied to social systems. Fararo and Hummon (1994) used DES to analyze several aspects of social networks; the senior author is well-known for his contributions and extensions to the theory of action (e.g., (Fararo & Skvoretz, 1984)), so it is worth noting that he (with collaborators) did not employ DES to simulate it. Jin and Levitt (1996) described knowledge-work projects in terms of its how they are organized, the tasks to be performed, and the links between the two. Then tokens, as stand-ins for real work items, are moved through the task network (PERT chart) in a simulation of the work to be accomplished as delays, errors, and noise were inserted into the project so that final per-formance of the project was predicted, taking into account important social (particularly team and organizational) aspects of knowledge work. The actual mechanism of the

Page 42

simulation was discrete-event.26 This project simulation system was described in lay terms in Samuelson (2000).

Accordingly, while simulation itself was no stranger to social systems, discrete event simulation was very rarely used.

26 Disclosure: the system described is an educational version. There were also several commercial versions and the

researcher's employer was a distributor and partner of the Stanford University spin-off created to enhance and market the commercial version of the simulator. The researcher was the in-house expert of his employer.

Page 43

III. METHODS Research overview

Like Sastry’s, this study was a test of our understanding of the interface between (narrative) description and the technical goal of predicting the future of a social system by constructing and "bringing to life" a laboratory replica of such a social system (see Figure 10). In another sense, it was an application of the translation of description into enactable constructs that can mirror the structure and function of a social system. The study, therefore, had two conceptual forks: (a) understand the description of the theory, and then (b) reify that understanding so that a laboratory replica can be created and oper-ated. Expanded into more detail it looked like:

1. Understand the theory of action a. Read what Parsons, his disciples, and his critics wrote. b. Select descriptions. c. Translate into constructs (say, structure and function). d. Validate the simulation when it is completed.

2. Construct a computer simulation program a. Develop constructs from the authoritative text, as needed. b. "Program" the constructs into a computer simulation program. c. Assure the correct technical operation.

System simulation

Parsons’theory of

action

Subject of this dissertation

Figure 10. Intersection of the theory of action and system simulation.

Research methods Place of simulation and theory

Parsons (1977b) wrote: Methodologically, one must distinguish a theoretical system, which is a complex of assumptions, concepts, and propositions having both logical integration and empirical reference, from an empirical system, which is a set of phenomena in the observable world that can be described and analyzed by means of a theoretical system. An empirical system … is never a totally concrete entity but, rather, a selective organization of those properties of the concrete entity defined as relevant to the theoretical system in question. (p. 177)

Parsons above restates the definition of "model," just like the subject of this research. In other words, a model is the theory with some details left out, abstracted

Page 44

away. So, is a model the same as a theory? Is a model the same as a simulation of a theory? These are questions still being debated by the social simulation community.

Perhaps the clearest explanation is from an electronic mail message that was in response to this question, posed on the social systems simulation SIMSOC LISTSERV ([email protected]): Is a simulation a (logical) derivation of a theory? Either (a) the simulation is a logical consequence of the theory and is [therefore] a theorem of this theory and then "If you can derive a contradiction from a given set of axioms, then that [sic] axioms are invalid"; or (b) it is an entity that is something like a theory with its "own" axioms and theorems: theory: axioms(T) -> theorems(T) sim(1): axioms(sim(1)) -> theorems(sim(1)) … sim(n): axioms(sim(n)) -> theorems(sim(n)) So, now what is the relationship between the axioms of T and sim(1)...sim(n)? In the philosophy of science there are several attempts to clarify the relationship between theories and models (here: simulation). One attempt is from Morrison/Morgan [(Morrison & Morgan, 1999)]. Another attempt is from Sneed, Stegmüller and other authors [(Balzer, Sneed, & Moulines, 2000; Stegmüller, 1979)]: the structuralist conception of theories (structuralism) (cf. Klaus G. Troitzsch on simulation and structuralism [(Troitzsch, 1998)])27. According to structuralism a theory is a theory-net consisting of several theory-elements connected to a basic theory-element. There are several links between the theory-elements: specialization, extension etc. From the perspective of structuralism, you can specify the relationship (links) between "theory" and simulation(s). The entity called "discursive sociological theory"28 is the basic theory-element that speci-fies the basic concepts, the basic axioms. A simulation is a theory-element, which speci-fies additional, new concepts, new functions and introduces new axioms. The introduc-tion of new ("gap-filling") axioms is in principle not a problem. It becomes only a prob-lem if you add new axioms that are contradictory to the axioms of the basic theory-ele-ment. To the point "explanation and simulation": According to structuralism a simulation is an extension/specialization etc. of a "discursive sociological theory" and can explain certain aspects of reality. In the case of specialization it refers only to a subset of the applications of the "discursive sociological theory"!

27 As evidence that this is not a settled matter, Troitzsch has called for a workshop on Epistemological Perspectives on

Simulation, in Koblenz, Germany, 1-2 July 2004, http://www.uni-koblenz.de/EPOS/. He states in his call for abstracts "Simulation has been a research instrument for long in various disciplines. In recent years, it is gaining further attention. This may be contributed to the lack of theories that would allow for explaining and predicting the behaviour of complex systems. In addition to that, new modelling paradigms, associated with object-oriented concepts, intelligent agents, or models of (business) processes inspire the use of simulation. … Furthermore, it seems that simulation is regarded by some as an alternative to research methods that do not provide convincing support for certain research topics. At the same time, the epistemological status of simulation remains unclear. This is, for instance, the case for its relationship to core epistemological concepts, like truth and reason. Against this background, it seems worthwhile to reflect upon the preconditions of using simulation successfully as a research tool."

28 Parsons' theory of action is this type, a descriptive sociological theory, as opposed to, say, a mathematical or formal expression of a theory.

Page 45

In a word, then, a simulation of a theory is not the theory itself, but rather an extension and specialization, perhaps with some elements added ("gap-filling" axioms) because in order to conduct the simulation they had to be. That is, one of the challenges in simulating a theory is placing into the simulation that which is missing in the theory, but is necessary for the simulation to proceed. An example in the instant case is queues or buffers. There are no queues or buffers or waiting lines in Parsons' theory of action, but what if there is a quick succession of interchange messages, too quickly for a function to process or absorb. What happens to the interchange message? Is it queued or lost or resent? Whatever the answer, it is a "gap-filling" axiom that is added to the simulation but absent from the theory – in order to get the simulation to run.

This research was about a method: applying the method of simulation to a theory of sociology. This section describes simulation and how it was applied in this instance. There would be a scientific elegance if the steps were performed in the order in the out-line. In fact, the steps were applied in a messy fashion because choices made in any step may not work downstream. And any researcher seeking to simulate a theory is, as she reads, mentally applying the techniques known to the researcher, even if subconsciously. The techniques form a part of the grounding (a bias) that any researcher brings to a simulation problem. The particular assortment of techniques that a researcher knows colors her perception of the problem. Accordingly, in this research a prototype approach was taken:

1. Read a representative sample of the theory of action. 2. Try a few descriptions as the basis of a pilot exploration. 3. Try a simulation technique (say, system dynamics or discrete event). 4. Encode the description using the simulation technique’s representation system

(that is, programming language). 5. Operate the simulation to see if the results correspond to what the theory

describes. If the pilot obtains results of sufficient fidelity (an unevaluated term), then the

researcher will expand on the sample, will expand on the descriptions to be encoded, will stay with the simulation technique but may increase the fidelity of the representation (which in principle can be done nearly infinitely), and will operate a number of cases to increase confidence that the simulated situation corresponds to the description of theory. In addition, the operation of the computer program so constructed was validated.

One of the primary contributions of this research was the application of a par-ticular simulation technique to a sociological theory: discrete event. It was central to the contribution even though its choice will not occur first in the sequence of research events. Choose appropriate simulation technology

There are styles of (social system) simulation. Two at the top of the description tree (Figure 11) are the main ones: continuous and discrete event. The continuous style mirrors systems that are continuous in time, such as most physical phenomena (e.g., dis-tance, velocity, acceleration). The discrete mirrors step-by-step events that occur at a particular time or in a particular order and for a particular duration, such as taking a test, filling a car with gasoline, cooking a meal, etc. One of the differences is the mathematics involved and the underlying mechanical way that time is advanced by the computer simulator.

Page 46

Traditional, single agent Agent based

Discrete event

System dynamics ...

[Time] Continuous Other methods of accounting for time

Social system simulation(Treatment of time)

Figure 11. Classification of social systems simulators, indicating the

position of this research in bold.

An example of continuous simulation is system dynamics, of which Sastry’s work is illustrative (Sastry, 1997). System dynamics formulates its problems as one of flows and accumulations. The standard example is a bathtub: water in the tub accumulates as water flows in from a tap and water in the tub falls as water flows out through a drain. What does this have to do with social systems? Nothing directly, but it can be used to explain any accumulation, such as competence, performance, and the ability to change, to mention a few in Sastry’s case, per Figure 1.

For the simulation described here, the discrete event technique was selected because its style most naturally reflects a description of the sort "This happened, then that happened, after which this happened." Discrete event simulation reflects the importance of time-ordering. This is how, by and large, Parsons described the dynamics of his theory, therefore there is a match, fit, and correspondence between the description and the com-puter simulation technique. Such correspondence is the operational definition of "high fidelity." For example, in describing some of the attributes of the pattern variable affec-tivity/affective-neutrality, Parsons stated that affective actions take less time than affec-tive-neutral ones, that it takes longer to react rationally (affective-neutrality), to study a situation, than it does to respond emotionally (affectively). (Parsons, 1982) Discrete event simulation makes the enactment of significant time-ordering a centerpiece, hypothesized here as a good fit with Parsons’ description of the time-dependent aspects of his theory.

The use of agents is gaining currency in social systems simulation (Epstein & Axtell, 1996; Gilbert & Conte, 1995). Basically, agents are autonomous "computers" that imitate individuals that each execute their own "program" of social actions. They interact with other agents (individuals), using the outputs of those agents to inform their "pro-grams" and possibly changing their internals. They form "communities," "(artificial) societies," and take collective action. There are agents (actors) in Parsons’ theory and he described them and their actions at a lower unit of analysis than that treated here. Simu-lating agents in Parsons’ theory of action might be a future application of the simulation described here. If one viewed agents as co-operating and communicating sequential proc-esses (Hoare, 1985) (in the context of agent-based simulation), then this study gives insight into the program that might be inside each agent, that is, the instant research is a necessary precursor to an agent-based simulation of Parsons’ theory of action.

The use of discrete event simulation was a centerpiece of this research because prior work has almost never used it. Almost all previous social systems simulations have used continuous or other methods of representing time (such as discrete Markov chains (Coleman, 1964), or a cadence at which all agents communicate and change internal state (Lomi & Larsen, 1995)).

The construction of a simulation, once the subject matter is understood, was to translate that subject matter into a language understood by a type of computer program, a simulation engine. The simulation engine acts like the subject matter by, in the case of

Page 47

the discrete event technique, moving "work" through a series of transformations (work stations) that operate on the work in some simulation-useful fashion. The user will spec-ify at what rate or interval "work" arrives, where it will go as it traverses the network of transformations, and how it will move outside the boundary of the simulation (that is, dies).

One can imagine an industrial baking oven in which dough loaves arrive at some rate, move through an oven at some rate, and then move out of the system at some point. Each of the "stations" has attributes, such as oven temperature. And each of the units of "work" has attributes, such as composition (rye, pumpernickel, etc.). And any of the attributes can be a random variable or a stochastic variable (that is, depend upon a ran-dom value).

There are two parts to a simulation system: the engine, which interprets constructs in the simulation language, and the simulation language itself. The engine moves work along in accordance with the specification stated in the language. The language in the case of the discrete event technique is often represented as boxes and lines between them. Work travels along the lines and is transformed inside the boxes. Sometimes the work arrives more quickly than the boxes can service it, so the work has to wait. Work waits in a queue. One can again think of the baking example: what happens if dough loaves arrive more quickly than the ovens can cook what has already arrived? The new dough loaves wait. What if the ovens are always slower than the process that creates the dough loaves? An infinite queue builds if the arrival rate exceeds the service rate.

The internals of the operation of a simulator are beyond the scope of this disserta-tion. Suffice it to say that simulation is a mature discipline and that there are many choices available to the researcher so that she does not have to build a simulation engine or develop a simulation language (Banks & Carson, 1984; Mize & Cox, 1968; Zeigler, Praehofer, & Kim, 2000). The criteria to be applied in the search for such a combination of discrete event simulation engine and language for this research was, in priority order:

1. Building-block approach, with the existence of many pre-built components. This refers to the simulation language. Many computer programming languages are functional, where each line instructs the computer to perform a specific function, such as arithmetic or printing. Building blocks, in contrast, specify the compo-nents (works stations, units that transform the work) and connections in a net-work. The advantage of a building block approach is that less has to be specified and what is specified is particular to discrete event simulation. The pre-built com-ponents simplify the job of specification because at some level of abstraction all discrete event simulations are similar. All create work at some rate, distribute it among work stations, transform the work item, collect like items, and then have them exit the simulation at some rate.

2. The ability to create blocks that are not already in existence (by writing a com-puter program) if the appropriate pre-built component is not available. In the event the selected language could not specify something important in this research, then it would be valuable if the block could be created, even if that meant writing a (presumably) small computer program. This attribute is called extensibility in the computing literature.

3. Ability to create a graphical user interface for the user, in which input values can be requested and outputs can be viewed graphically. The primary users were envi-

Page 48

sioned to be people interested in social systems, not necessarily computer-centric professionals. Therefore, graphic communication was important.

4. Visual programming approach for the developer of the simulation. This technical attribute makes using the simulation language easier because the specification of the components and their interconnections would be graphic, too, like a diagram that becomes animated.

5. Low price. There is no loss of generality in the selection of discrete event simulation for Par-

sons’ theory of action and the computer program could be converted to continuous simu-lation by another researcher (Zeigler et al., 2000). That is, (mathematically) every dis-crete event simulation can be represented in continuous time, but the character of the dis-crete event outlook is destroyed in the process. In other words, the choice of the discrete event technique will not preclude the choice of the continuous time technique, but the transformation of discrete event to continuous time leaves the original, underlying dis-crete event simulation unrecognizable. In yet other words, one might ask, "Does the choice of the discrete event technique mean that there are cases that cannot be repre-sented, that choosing, say, continuous time representation would be strictly more power-ful?" The answer is that the continuous time representation is strictly more powerful (that is, there is a least one model that can be represented in continuous time but not as discrete events), and that any discrete event representation can be translated into the more inclu-sive continuous time representation by a series of (algebra-like) steps. But the translated model will not look like the original discrete event one because the translation process does such violence to it. Translate from description to simulation

Any simulation from descriptive text is built in a step by step process: 1. Identify description that will serve as the authority. 2. Transform the description into constructs:

a. Reflect the structure in the simulation. b. Reflect the functions in the simulation.

3. Build the simulation based on the collection of constructs. 4. Operate the simulation. 5. Validate its fidelity. 6. Run more cases, rework, improve, expand.

To some degree the focus on time-ordering filtered how Parsons’ texts were read and interpreted, how it was parsed into constructs about structure and function. In par-ticular, when reading to find constructs one inevitably was led to questions of timing or precedence. To the extent these questions are not out of place, not forced, the choice of discrete event simulation was (informally) validated. Construct selection criteria

In order to identify description that would serve as an authority, many writings of Parsons were examined. One stood out as especially suitable for research purposes: Par-sons, Bales and Shils (1953a). This 107-page working paper was the anchor29 of the work described here, just as Tushman and Romanelli (1985) was for Sastry. While the work here will not be limited to the anchor, it will likely rely heavily on it.

29 Researchers in artificial intelligence call such a source an "oracle" in order to convey the status as an authority.

Page 49

The following criteria were applied to each candidate construct: 1. Essential – Will it be possible to build a high fidelity simulation without

this construct? If not, then it was included. 2. Parsimony – Will this construct already be present or in a more general

form? If so, then it was excluded. 3. Dependency – Will the construct stand alone or depend on (many) others?

If it will stand alone then it was favored for inclusion. 4. Spectrum of interpretation – Will this construct be relatively free from

interpretative bias (e.g., discrete time) or multiple interpretations? If so, then it was included.

5. Richness – Will the candidate add a lot of meaning or will it be a detail? If it will add a lot of meaning then it was favored.

6. Recency – Favor Parsons' most recent rendition of his theory Validate the simulation

Since the simulation reflects a theory, the validation step attempts to answer the question, "Has the simulation been faithful to Parsons theory of action, according to Par-sons et al. (1953a)?" There were two steps in the validation, both performed by a third-party expert on Parsons' theory of action:

1. Did the structures and functions in the simulation accurately reflect the theory of action?

2. Did the technical operation of the simulation produce results that were predicted by the theory of action?

That is, the validation step was a subjective assessment by experts about the accu-racy of the translation of theory to simulation artifacts. The attestations of the two experts, one for each question, are found in an appendix.

Lave and March (1993, ch. 3, Evaluation of speculations) posit another approach: truth, beauty, and justice.

The construction and contemplation of models are æsthetic experiences. Like other æsthetic experiences they become richer and more enjoyable with an appre-ciation of their nuances. The dicta of methodology are nothing more mysterious than rules of thumb or improving the artistry of speculation (Lave & March, 1993, p. 52).30

By truth the authors mean correctness. The model should accurately (synonym for correctly) reflect the assumptions and derivations of the underlying theory. The emphasis is on testing the derivations, not the assumptions because assumptions are often axioms and therefore true by definition. Truth is sought by several means:

• Testing for circularity. Are the definitions tautological? • Promulgation and evaluation of alternative derivations. Seeking alternatives can

expose errors in the original derivations and at least help sharpen models. • Differentiation among competing derivations. Can an experiment verify or favor

one derivation over another?

30 The comment about the appreciation of nuance can be related to March's long association with models, beginning

with Cyert & March (1963), one of the first simulations of an organization.

Page 50

• Playfulness. By being intellectually playful a researcher can become less invested in and less fervent about his/her model and therefore more open to question its correctness. By beauty the authors mean three criteria:

• Simplicity. A smaller number of assumptions is more attractive than a larger num-ber, all other things being equal. This is Occam's razor and parsimony restated.

• Fertility. The simulation produces a large number of interesting derivations per assumption. That is, the simulation is rich in descriptiveness. This is also a way to say that more general simulations are preferred because they apply to a greater number of situations.

• Surprise. Some of the implications are surprising and not immediately obvious from the assumptions. The critical impetus for system dynamics models, for example, according its creator, Jay Forrester, was to be an instrument of policy study because the outcome of feedback loops is counterintuitive (that is, surpris-ing) (Sterman, 2000). Achterkamp and Imhof (1999) also list surprisability as one of three important features to credibly establish computer simulation in sociology (generality and power to separate theoretical from technical results are the other two). Justice is the third dimension offered. It is a reminder that we should pursue the

explication of social myths and that our own philosophical commitments are not neutral, that we have to expose, examine, and question our own worldviews.

How might we evaluate or judge the truth, beauty, and justice of the model/ mod-eling described in this research? One way would be to ask the dissertation committee members to make subjective evaluations on each of the dimensions and compare them with other models they admire. This way the research described here would be placed in a "quality" context related to other, like models.

It may be worth noting that in the case of pattern variables there were some field experiments that potentially could have been used to validate the model (Cherns, 1980; Park, 1967; Podell, 1966; Podell, 1967; Williams, 1959), but, alas, none of them were able to make inferences at the level of analysis used here, namely organizational. Assure correct technical operation

There are several parts of the technical operation. Conceptually, there is the simulation that is constructed by the researcher, the computer program that interprets and executes it, and the information consumed during the operation of the computer program. There has been general frustration with the quality of computer software since its humble beginnings in the late 1940s and this dissertation's application of computing will not solve those concerns.31

There are two general approaches to increasing the quality of computer software: assuring/proving its correctness and increasing the confidence in the final version by testing. Other hard science disciplines often assure the correctness of their work by proving certain properties of the results. For example, scientists and engineers can use physics to "prove" that a bridge will withstand certain forces. But computer programs are 31 It may be worth noting from an authority perspective that the author of this dissertation is the Associate Editor in

Chief of Quality of the Institute of Electrical and Electronics Engineers (IEEE) Software Magazine. IEEE is the largest professional organization in the world and the Computer Society is the largest subdivision of IEEE. Software is a publication of the IEEE Computer Society.

Page 51

not physical objects and the laws of physics do not apply. On the other hand, computer programs can be mathematical objects and a branch of mathematics could be used to prove properties. The intuition is use to the logic we employed in high school geometry when we proved theorems about triangles, such as congruence. We could say that the lines in a computer program are like a geometry proof, where our task is to show that each line is an entry in a proof that given the inputs and the transformations being applied, the output was what we want. This way computer programmers would be able to prove that their programs worked before they were ever executed, before they were every tried on any computer.

As appealing as this approach was conceptually, it has not been widely applied in practice. For example, even though they are prime candidates for this mathematical proof of correctness, the most popular computer programs for statistical analysis (e.g., SAS, SPSS, BMDP, and Systat) did not use the approach. Rather they, and nearly all other computer programs, are tested as a method of improving confidence in the results pro-duced by the computer program. Testing can only show the existence of errors, never their absence.

Testing is essentially a (serious) mathematics problem. Even a simple computer program has more states in it than there are assumed to be molecules in the universe. Therefore, exhaustive testing – testing of every state that can be obtained in a computer program – is not feasible as a matter of practice. Even with the fastest computers it would take hundreds of years to try all of the states in a simple program.

Accordingly, one application of math to the problem is to find/compute equiva-lence classes, one example of a large subset of the possible states, and have that one example stand-in for all of them. For instance, if we were testing the printing of United States ZIP codes, those five-digit numbers that indicate the general geographic location of mail destinations, then we might select a sample of them instead of all 00000-99999 = one hundred thousand possibilities. In fact, it might be normal to select only three values to test the printing: 00000, 99999, and a random choice in between.

This leads to another testing approach that is ad hoc but often used: test the places in a computer program where errors are known to "hide": boundary conditions and inter-faces. Boundary conditions are the extreme values that a computer program takes in or puts out, such as a large negative number, zero, and a large positive number. Interfaces are places where one computer program uses the services of another, such as, in the cur-rent research, the simulation engine invokes Microsoft Excel for input from a table.

Upon closer examination of the research described here errors could occur in the following places in computer programs:

1. Errors in the specification of the simulation, that is, in what can be specified in the SIMUL8 language. The researcher is the author of the specification. This type of error can be discovered by reading and by being traced back from anomalous results.

2. Errors in the execution of the SIMUL8 language program. The provider of SIMUL8, SIMUL8 Corporation, is the author of the engine. Therefore, errors of this type are the most difficult to discern, hopefully are the most rare, and can be discovered by tracing back from anomalous results. Also, SIMUL8 Corporation regularly updates the engine based on user input from world-wide usage.

Page 52

3. Errors in the values provided in the Excel sheet that is read in during program execution. The researcher is the author of the specification. This type of error can be discovered by reading and by being traced back from anomalous results.

In sum, the quality of the results of the simulation cannot be proved and will always be suspect. Confidence in the results can be improved by increased testing, which is a problem of multiplicity of states, of trying to test as many states as practical given the constraints on the research resources.

Delimitations The delimitations of a study are those characteristics that limit the scope (i.e.,

define the boundaries) of the inquiry as determined by the conscious decisions to exclude and include that were made. Among these were the choice of objectives and questions, variables of interest, alternative theoretical perspectives that could have been adopted, etc. The first limiting step was the choice of problem itself.

This study was highly biased by the search for strong time orderings; that is the basis of discrete event simulation. In order to construct the simulation the researcher scoured Parsons, Bales and Shils (1953a) for even the remotest indications of time order-ing and surely this biased the fidelity of the simulation. Furthermore, it could never be argued that the time ordering in Parsons theory was an essential feature, surely not on the level of the pattern variables, functional prerequisites, interchange media, and cybernetic hierarchy. Accordingly, the claim must be made that the research here was one transla-tion of the theory, not the (definitive) translation. The simulation was not comprehensive; at best it is intended to be a scaffold that other researchers will use to build higher fidel-ity, more comprehensive simulations of the theory of action.

In addition, the theory of action contains many permutations and layers. The study here limited its scope to:

1. Performance; neglects learning – The Parsons theory can be applied to two views of organizations: their performance in pursuit of their "exterior" goals, and their "interior" learning as they perform. In the performance case, energy flows from Adaptation to Goal Attainment; in the learning case energy flows from Integration to Goal Attainment (Parsons, 1960, p. 217). This research addressed only the per-formance view/flow. As an aside, the organization simulated does learn how to perform (well, how to reduce tension, the difference between the energy outside the organization and the energy circulating within), using classical conditioning, which is not what Parsons implied in his learning vs. performance dichotomy.

2. A single unit of analysis: the organization – Parsons illustrated that the unit of analysis of his theory can be any size, from, for example, individual to a nation or national culture. This research selected a single unit in order to demonstrate feasi-bility. Also, because of a single unit of analysis the research did not address inter-penetration, the impact of levels of analysis on each other, as, for example, norms for the personality level can impact the performance of an individual at the col-lective level.

3. Single level of the four functional prerequisites (not the infinite regress) – In addi-tion to multiple units and levels of analysis, Parsons illustrated that each of the four functional prerequisites can, in turn, be subdivided into four units, and each of those four units into four more, ad infinitum. This research addressed a single

Page 53

level in order to demonstrate feasibility of simulating the characteristics of each of the four different prerequisites.

4. One of the pattern variables, not all of them – Parsons constructed four pairs of variables to explain the underlying behavior and motivation of the four functional prerequisites. This research addressed one, affect and affect-neutrality, because it is described in Parsons, Bales and Shils (1953a) as having a time-varying impact. That impact was that events that engage affect (emotional) are dealt with more quickly than those that are affect-neutral (cognitive, rational). Also, on account of not simulating all of the pattern variables the interaction among them, as seen in the proliferation of 16 combinations32 in Figure 2, was not investigated. There-fore, interaction effects were neglected, perhaps to the significant detriment of the simulation's fidelity. Not taking into account the rest of the pattern variables may, in fact, invalidate the model because the feedback and interaction among the pat-tern variables could well change the outcomes entirely, as in any dynamic system where a model simplifies feedback (Hanneman, 1988). On the other hand, since the purpose of this research was to show feasibility, simulating only one pair of pattern variables may serve that goal.

5. Only orientation, not modality – Each pattern variable pair defines one property of a particular class of components. Orientation, one of set components, concerns an actor's relationship to the objects in his/her situation and is conceptualized by the two "attitudinal" pairs of variables of diffuseness-specificity and affectivity-neutrality. Modality, the other set of components, concerns the meaning of the object for the actor and is conceptualized by the two "object-categorization" pairs of variables of quality-performance and universalism-particularism. This research only addressed orientation, and, as mentioned above, only one pair of them, namely affect and affect-neutrality.

6. Sufficient affect-neutrality, not sufficient affect – Fararo observes (2001) that in order to sustain a system there needs to be sufficient attention to the tasks con-fronting it. In order to address those tasks a system must emphasize affect-neu-trality, though, of course, not to the total exclusion of affect. "A necessary condi-tion for social order is that affective neutrality is not too small." (p. 157) This research did not address sufficiency, but did address a related topic, the pursuit of reducing tension.

7. No direct simulation of the cybernetic hierarchy – The hierarchy was implicitly simulated on account of the direction of flow of energy, only Adaptation to Goal Attainment to Integration to Pattern Maintenance. Therefore control is manifest only from Pattern Maintenance to Adaptation. Here, too, the observation in 4, above, about the omission of feedback paths invalidating the fidelity applies. There would likely be an entirely different set of outcomes if all of the paths were modeled. As stated in 4, the purpose of this research was to demonstrate feasibil-ity, so modeling some of the paths may achieve that goal.

8. Only a few of the interchange media – There are twelve interchange media. This research only examined four: Adaptation to Goal Attainment to Integration to

32 Four pairs of pattern variables generate 24 = 16 combinations.

Page 54

Latent Pattern Maintenance, and then Latent Pattern Maintenance back to Adap-tation.

9. Very few of the possible process features in Parsons, Bales and Shils (1953a) were simulated, not all of them. In every case, the choice of boundary was caused by the nature of this research: a

toy simulation to investigate the feasibility of simulating Parsons' theory. That is, by its nature this research was bounded.

Page 55

IV. THE MODEL AND SIMULATION Why simulate?

Regardless of the form of the model or technique used, the result of the elicitation and mapping process is never more than a set of causal attributions, initial hypotheses about the structure of the system, which must then be tested. Simula-tion is the only practical way to test these models. The complexity of our … models vastly exceeds our capacity to understand their implications. (Sterman, 2000)

Chatting about Sociological Laboratory [a book and accompanying computer programs by Bainbridge] over lunch with me one day, George Homans said that he had once hoped to write a third book in the spirit of The Human Group and Social Behavior. To be called A Toy Society, it would start from his set of simple axioms and build a miniature society operating according to principles logically derived from the axioms and through its realism, demonstrating that Homans's approach could indeed explain the chief features of human society. Unfortunately, Homans said, he could not find the means to produce a functioning toy society. Computer simulation, he agreed, could be that means. Unavailable to Homans, modern computer simulation techniques make possible a variety of experiments with toy societies, leading ultimately to a grand test of the logical coherence and sufficiency of any theory of human behavior constructed along the lines proposed by Homans. (Bainbridge, 1992)

There are only a few alternatives to a computer simulation of a social system: normal science, a descriptive model, real world experimentation, or a formal model. Normal science is possible when parameter variables are able to be controlled, and then the whole mechanism of hypothesis testing, experimental design (Campbell & Stanley, 1963) and its corresponding statistics are available.

Parsons provided the second: a descriptive model, using words and drawings to communicate his meaning. How do we come to understand the interactions of the com-ponents he proposed? How can we test our understanding of how the parts fit together and what the outcomes are?

Real world experimentation is a problem in all social sciences because of the lack of control: parameter values cannot easily be fixed and even if they could the very fixing often interferes with the process in situ that was sought to be observed. In addition, there is considerable attenuation (that is, delay between cause and effect) in the real world, particularly with the Latent Pattern Maintenance function. Worse, the very nature of it, its latency, means by definition that it cannot be directly observed.

Formal models are those about which we can reason, usually by manipulating the statements. The most common formal model is mathematical; there are also a number of formal models stated in first order predicate logic or as the type of axiomatic logic one finds in high school geometry or trigonometry theorem proofs. There is a formal model (in the predicate calculus) of Parsons' theory (Brownstein, 1982) and it does illustrate inconsistencies and gaps. However, due to those inconsistencies and gaps, according to its author it cannot be used for inference. There is no complete mathematical model of the theory of action.

Page 56

One objection to formalization in general is that important aspects of a problem might be omitted in order to obtain or preserve tractability. Simulation was specifically developed to overcome the limitation of tractability, but, in fairness, imposes others, such as a significant barrier to entry in terms of computer and operations research knowledge, just as mathematics-based approach does, too (Hanneman, 1988).

Mathematical formulations of complex problems often exceed the capacities of the creators and consumers to understand and explicate them. The introduction of conditional relationships, nonlinear relationships, and complex patterns of coup-ling among even small numbers of variables can rapidly exceed our capacity to solve such systems, or to comprehend the meaning of the solution if one is found. (Hanneman, 1988)

Simulation languages are intermediate between mathematics and description. They are invented languages that are more restrictive than natural languages (such as English) and less restrictive than traditional mathematics. They force upon the program-mer a discipline to describe in some detail the structure and function of the system under study.

Another objection to simulation is that the social world is unsystematic and am-biguous in causality (Tsuchiya, 1966). Two researchers, in particular, have had success using simulation to aid understanding of "soft," that is, qualitative problems, including those of ambiguity, which holds out hope that simulation can be used in poorly-quantified domains (Robinson, 2001; Tsuchiya, 1966). In addition, the whole application of system dynamics to social systems is a response to this objection, as documented in the annual Proceedings of the International Conference of the System Dynamics Society and in the quarterly journal, System Dynamics Review. And (Hanneman, 1988) is a tour de force in the application of system dynamics to social systems. System dynamics, in that role, attempts to aid understanding by showing the consequences of assumptions about struc-ture and function.

In some sense we are "stuck" with simulation as a laboratory workbench for exploring our understanding of social systems and where that understanding might take us. Simulation, while formulated in the orthodox theory it is trying to animate, nonethe-less can lead to radical changes in such fundamentals as the presuppositions, model boundary, time horizon, and dynamic hypotheses (Sterman, 2000).

Model construction Models are two things: a choice of goals and a choice of constructs from many.

The goal of this model was to illustrate feasibility: was it feasible to simulate Parsons’ theory of action? The choice of constructs was more multidimensional: unit of analysis, level of abstraction, granularity, time horizon, fidelity. For example, clearly the level of fidelity will drive the choice of granularity (greater fidelity requires greater granularity), and the choice of granularity will drive the choice of the unit of analysis (the greater the granularity the lower the unit of analysis). Since the goal was feasibility, the highest unit of analysis was selected, along with low granularity and low fidelity.

In the sense used here, a model is a theory with bits left out. The construction of a theory – like that of a model – is a messy mental process. Most reports of theories do not, thank goodness, give the details of the creative and cognitive processes that gave rise to them. In its briefest form, the researcher reads something that inspires an outline of a theory and gives rise to a place to "hang" or represent future insights. In other words,

Page 57

there is a preliminary (usually mental, hypothetical) theory that is elaborated as more and more concepts are added. That first version of the theory is sufficient if there is a place to put each elaboration without "too much" work. This is the cognitive mechanism behind grounded theory, a creative process of deriving an explanation from raw data. It is also the cognitive mechanism behind the process point of view: a creative process of deriving the mechanism that yields an outcome.

Numerous authors (e.g., Bainbridge (1992), relying on Rodney Stark's "sociologi-cal process" (2003)) have described their steps for building a theory or model. Here is a sample (Lave & March, 1993):

1. Observe some facts. 2. Look at the facts as though they were the end result of some unknown process

(model). Then speculate about the processes that might have produced such a result.

3. Then deduce other results (implications/consequences/predictions) from the model.

4. Then ask yourself whether these other implications are true and produce new models if necessary.

A recent table, below, shows the variety of steps possible, summarizing those offered by various system dynamics authorities: Table 3. The system dynamics modeling process across the classic literature. (Luna-Reyes & Andersen, 2003)

While all of the proposals, above, appear logical and linear, in fact the process is a

non-linear, iterative one of creative speculation and hypothesis testing. No more will be said of this inchoate process; the point is to appreciate the difference between what is prescribed as a set of steps to establish a theory or model and what really transpires inside the mind and workbench of the theorist or model builder.

The plan of this chapter is, first, a description of the elementary model, followed by a short tutorial on the implementation of learning and tension. After that is a descrip-tion of what the user saw as she operated the simulation, along with the rules and assumptions that were implemented. The chapter concludes with a parsing of selections from Parsons, Bales and Shils (1953a) and their correspondence in the possibly more elaborated model to illustrate the fruits of the process that built a bridge between Parsons' text and the simulation model.

Basic concept The basic concept of the simulation is that of a baking oven fed by a conveyor

belt on which is raw dough. The dough represents the energy outside the system under study (the oven). The oven represents the heat that will be applied in successive internal

Page 58

chambers as the raw dough is transformed into its cooked form. The goal of the oven (system) is to produce bread that is cooked "just right." This particular oven senses how large the dough mass is and adjusts its internal heat based on it. And, in fact, it adjusts based on the pattern of dough masses as it sees them one at a time as they enter the oven door.

In Parsonian terms, the dough represents energy from outside an organization, say news or a new idea. The news will pass through the four functional prerequisites as it is transformed and it transforms the organization. Success is measured by how well the latent pattern maintenance matches the pattern of energy entering the system. The differ-ence between energy presenting itself to the system and the energy inside the system (at the latent pattern maintenance stage) is called tension. Our goal is to minimize tension, so the goal of the interaction of the internal functions and structures is to match or fit the latent pattern maintenance energy to that that entering the organization.

To increase the fidelity a bit, there are not only different bread masses, but differ-ent kinds of bread (rye, poppy, sourdough, etc.). For each the oven has to react differently because in order to cook properly it is not only a matter of temperature but also of time. Some dough has to be cooked more quickly, some more slowly, even at the same tem-perature.

In Parsonian terms, in addition to raw energy in the environment, there are values of pattern variables that are intrinsic to different types of energy. The effect of processing of energy that has one type of pattern variable, affectivity/neutrality, is modification of the time that the system takes to respond to the energy. An affective value moves the energy more quickly; an affective neutral (that is, rational) value moves the energy more slowly. So, external energy is typed – by the value of the affectivity/neutrality pattern variable.

Model of tension and learning To increase the fidelity a bit more, imagine that the oven knows that it works best

with raw dough that exceeds a certain mass, that is, the dough has certain characteristics or a "signature." So it filters out – rejects – dough that is not heavy enough. It takes in only a certain size and above. And -- here we are stretching -- the filter at the opening of the oven is operated from inside the oven: the intuition is that the oven comes to learn the minimum value that it will accept and adjusts the filter as it learns. The filter setting could be different for every lump of raw dough as the oven learns.

In Parsonian terms, the Adaptation function filters the energy that it accepts at the boundary of the system. That filter is set by Latent Pattern Maintenance. If the pattern maintenance function sets the filter too high, then some useful energy in the environment will not be imported, potentially creating tension. If the filter is set too low, then the sys-tem responds to everything and patterns are difficult to develop, expending system energy with no added patterned capability.

Another way to think of the simulation, that is, another analogy, is target tracking. The organization being simulated is trying to track the pattern of energy in the world out-side of it, just as radar tracks a target. If the target turns out to be a "bad guy," then the tracking attention should increase. If the target is noise, not an important thing at all, then it should quickly identify that and not expend extra energy. The difference between what is expended and what should be expended is tension, a quantity to be minimized.

Page 59

Before providing more detail it is important to take note of advice that Parsons sprinkled liberally throughout his writing (for example, (1968a, p. 47)): be careful about the unit and level of analysis. If the system under study is an organization, then it is treated as an indivisible "black box." We should look only at its input and output, not how it processed the input to achieve the observed output. But our situation was a bit dif-ferent – not that we are inattentive to Parsons’ generous advice – because the simulation created here generated the output by processing the input. Therefore, the researcher had to know something of the inner workings lest he could not have transformed the input into the output. Or, put another way, the computer simulation IS the black box.

Accordingly, the flow of energy from outside the system was first filtered in the Adaptation function, as stated above. If the energy passes through the filter, whose value was set by the Latent Pattern Maintenance function, then it will pass to the Goal Attain-ment function after a delay depending upon whether the energy to be responded to is affect or is affect-neutral. If it is to be responded to by affect, then it will traverse now and through the rest of its journey quickly, according to a user-set value. If it is affect-neutral, then it will travel more slowly, consuming time to "think." The energy will then pass from Goal Attainment to the Integration function according to delay and selection rules, and then it will pass from the Integration function to the Latent Pattern Mainte-nance function according to delay rules. Different delay values can be set by the user for each of the functional prerequisites x pattern variable value (affect or affect-neutral).

Systems are goal-oriented, so there must exist some mechanism that matches what outside energy is allowed in compared with the goals of the system. The goal of the toy system described here is to reduce tension – that is, the difference between the energy outside the system and the energy circulating inside the system. The mechanism, then, is to adjust the filter of the in-coming energy so that the energy circulating inside the system matches the outside energy. In principle there are many ways to accomplish the match.

In Latent Pattern Maintenance a complex interaction will occur that will set the filter on the Adaptation function. In essence, the Latent Pattern Maintenance function has as its goal to seek to minimize tension; that is, the system's goal is administered by the Latent Pattern Maintenance function. Symbolically it will do this by learning the pattern of energy arriving from the outside and matching the filter to it in order to let in enough energy to sustain the enterprise. Parsons, Bales and Shils (1953a, Fig. 7, p. 223) call the style of learning classical conditioning. Accordingly, this is the style of learning that was simulated. But it is learning by an organization, not by an individual, despite a question about whether organizations can be classically conditioned. Classical conditioning

As mentioned in the Literature Review, one of the challenges in simulation is transforming qualitative concepts into quantities so that a digital computer can manipu-late them. Unfortunately, there were almost no reports of quantitative measures of organizational learning, despite the abundance of references to organizational learning and learning organizations. One of the only quantitative studies of organizational learning was the formulation of Nembhard and Uzumeri(2000), which was used in this study:

Page 60

⎟⎟⎠

⎞⎜⎜⎝

⎛++

+=

rpxpxky , a three-parameter hyperbolic function, where

p = cumulative prior learning, in clock ticks33. Must be a positive integer. Incre-ments with every clock tick. Default is 500.

r = cumulative time to reach k, in clock ticks. Must be a positive integer. Default is 250.

x = units of time since the last change in k. Must be a positive integer or zero. Increments with every clock tick. Default is zero.

k = asymptotic value for Latent Pattern Maintenance. Initial value is 2.

y = successive values of k as x approaches infinity. This is the value that is pre-sented for Latent Pattern Maintenance in the computation of tension, where it is called the "energy level of L."

Figure 12 illustrates the intuition. k is the value being sought by the system, the value that L is trying to obtain by adjusting the energy filter on A. Given this value, Latent Pattern Maintenance must compute a possibly new value for the energy it lets into the system via the Adaptation function. Basically, the formula smoothes prior values in order to reach its goal of k in a stable and planned way. There are two cases: approaching the target k from above and approaching it from below, both of which are illustrated. Imagine that the L function has determined – in a process opaque to us at the moment – that the target value of some important variable is 2. If the outside environment is pre-senting, say, 4, then L clamps down on the filter that lets values in so that the 4s are not permitted to enter. This is Case 2 in the figure. Given the same target of 2, imagine that the outside energy is less than 2, then L opens up the filter and lets it all inside. This is Case 1.

In each case, the L function responds to the difference between its target value and the value circulating within the system, that is, the value let in. It then exercises its considerable control (it is the highest in the cybernetic hierarchy on the control dimen-sion) to bring the circulating energy closer to the target, either from above or below.

33 Clock tick represents an interval of time. In the simulation the clock "ticks" once every business day, as a default.

Page 61

Figure 12. Structure of the three-parameter hyperbolic learning curve model.

(Nembhard & Uzumeri, 2000)

Setting the target Now we address how the target of the learning model was set. That is, we needed

to determine what the successive values of k, above, should be. There was no reference in any of Parsons' works to guide us, and little elsewhere. One guidepost is that negative exponential distributions have appeared to model forgetting since 1885 (Ebbinghaus, 1987); see Wixted & Ebbesen (1997) for an argument that power functions are a better fit and Nembhard & Osothsilp (2001) for a review of more accurate forgetting models. The negative exponential distribution, Nnew = Nbase • egt, depends upon two parameters, Nbase and g, where t is time, e is the base of the natural logarithm, and Ni is the magnitude of, in our case, the energy. Nbase is the value from which the declining curve begins, the "anchor." g is a negative number that determines the rate of decline and its asymptote. The figure below helps to visualize the general shape of this "forgetting" curve, where Nbase is 4, g is –0.013, and the period is 52 (as in weeks in a year).

Page 62

Negative exponential distribution

0

1

2

3

4

5

0 10 20 30 40 50 6

Time

Energy

0

Figure 13. Illustration of a negative exponential distribution as a

"forgetting" function.

In the model, the setting of the target proceeds as follows: Step Reason 1. Given a range of time over which it is

looking (a window) and a sensitivity factor (threshold to respond to change), is the new value of the external energy higher than any other in the window, is it a new maximum?

2. If so, then adjust to the new maximum by changing the base of the decay and reset the time to 1.

The system does not want to respond to small (presumably random) variations around a maximum, so an absolute value is set by the user above which change can happen. And a window is specified (in an analog to simulation time units) in order to tune the responsiveness to change (shorter window implies more change; longer win-dow implies less change). The formula in the step is:

If external energy > maximum in the window) and that difference exceeds a threshold, then we have a new maxi-mum. So set Nbase to the new maximum and t to 1.

3. Compute the target value as: Nnew = Nbase • egt

The new value would depend upon either (a) the new maximum and t=1, or (b) it will continue to follow the negative exponential decay downward with the old base and the next natural value of t.

Page 63

Operation of the simulation The user of the simulation began by reviewing a dedicated Excel spreadsheet for

two types of values: parameter values and the pattern of energy that the Adaptation func-tion will see. Both can be seen in

Figure 14, a likeness of the user's Excel spreadsheet. The parameter values were entered into column 10. The "periods" referred to are clock ticks, ostensibly one business day by default.

Columns 1 and 2 contained in successive rows the magnitude of external energy and whether the energy will be processed as affect or not. If Affect=1, then the energy will be processed as affect, otherwise (nominally zero) it will be processed by an affect-neutral (rational) mechanism, implying that the processing of that energy will take longer than energy processed by affect. The first row was consumed at time=1, the second at time=2, etc. When a blank row was encountered, then the values began at the top again. In this way, the two columns can be thought of as continuously repeating until a target number of clock ticks has transpired.

Without loss of generality, the range of the magnitude of energy can be thought as having a maximum. It is helpful to have a maximum value for the magnitude so that the value of k (the target value of internal energy) can be easily compared to the external energy as a way to visualize how much learning must take place during the simulation.

Also, once all of the values are set in the spreadsheet they cannot be changed.34 Since the simulation is deterministic (not probabilistic), the outcome of a particular run is completely determined by the values on the spreadsheet and the length of time (that is, number of clock ticks) the simulation is run.

34 It would be a trivial upgrade to the simulation to have the simulation itself change the values or at certain intervals

or on the occurrence of certain events ask the user whether changes were wanted.

Page 64

Ver. 0.8 For The Parsons Game by Stan Rifkin1 2 3 4 5 6 7 8 9

Dwell time ValuesAffect Energy Adaptation

(1=true) (Level) Affect-Neutral (periods) (Should be greater than Affect) 100 2 Affect (periods) 50 2 Goal Attainment0 2 Affect-Neutral (periods) 1200 2 Affect (periods) 200 2 Integration0 2 Affect-Neutral (periods) 1200 2 Affect (periods) 600 2 Latent Pattern Maintenance0 2 Affect-Neutral (periods) 00 2 Affect (periods) 00 2 Percentage of ideas not funded0 2 Percentage (expressed as decimal < 1.0) 0.800 2 Prospensity to change0 2 Energy filter threshold (initial value) 20 2 Sense of test: Pass Energy if Energy [operator] Filter Threshold >=0 2 Learning/Forgetting0 2 Prior learning (p) 5000 2 Time to reach current pattern (r) 2500 2 Starting value of L energy (k) 20 2 Negative exponential decay parameters0 2 Threshold to respond to change (in energy units) 10 2 Decay coefficient (<0 and in time units) -0.0030 2 Response window (in number of Affect-Energy pairs) 52

10

Figure 14. User view of dedicated Excel spreadsheet.

Once the spreadsheet is completed then the simulation program was invoked and a screen like Figure 15 appeared. The user typically performs just two operations on this screen: reset the clock (and all other variables) to zero and then start the simulation. The default duration is 2400 business days, or approximately ten years. There is a row of

buttons along the top of the display: . The left most one is reset; the next one is "step," which advances the clock one tick each time it is touched; and the next one is "run," which starts the clock and the simulation runs automatically until the final value of the clock is reached. If the run button is pushed during the actual simulation then the program pauses; touching it again starts the simulation where it left off. The other icons are not used by the user, only by the researcher to develop the simu-lation in the first place.

Based on the results of the inputs the user can view the convergence of internal and external energy on a graph after the simulation has ended; it is in the file containing the Excel spreadsheet. That is, the user can judge how well Latent Pattern Maintenance performs its function of restoring spikes or challenges to its target value of "culture," which in the simulated case is instantiated by external energy.

Page 65

Figure 15. User view of the simulation.

Figure 15 is the tableau on which the user witnesses the simulation. Energy enters from the left on the device that looks like a conveyor belt, something like the bakery example. The value of energy and its affect/affect-neutral pattern variable comes from successively reading the Excel spreadsheet. This external energy is presented to the test in the spreadsheet: Pass Energy if Energy [operator] Filter Threshold, where operator and Filter Threshold are read from the spreadsheet. If the energy does not pass, then it goes to the element Energy that does not enter. If the energy enters then the Adapta-tion function looks at its affect/affect-neutral pattern variable value. If it is affect then the energy takes the top path, the one marked Affect path, and is processed for a period read from the spreadsheet. During that time, Latent Pattern Maintenance, in an unseen (hence latent) process resets to Adaptation filter to a possibly new value in order to reduce ten-sion. After that the energy goes to Goal Attainment, where it may have to wait in a queue of the GA function is busy. If the energy in Adaptation is affect-neutral, then it takes the lower path, possibly to a queue, where it waits for the affect-neutral processing for the period of time specified on the spreadsheet. At the end of that processing the Latent Pat-

Page 66

tern Maintenance function possibly resets that Adaptation filter in order to reduce ten-sion; that is, on each of the possible exits from Adaptation (affect and affect-neutral), Latent Pattern Maintenance potentially resets the Adaptation filter for the next time it encounters outside energy.

The icon depicts a "work station," where the input is transformed into an output for some duration and decisions are made about where to go next. In our case each

occurrence of a work station can make a different transformation. The and icons are queues where energy waits for a function to become available to process it. The number above the picture is the number of energy "bundles" waiting at the moment. The

icon is where energy exits the simulation. The number above the picture is the num-ber of "dead transactions."

Energy can enter Goal Attainment from two sources, both of which are paths out of Adaptation. The top is the Affect path and it is fed to the Goal Attainment function im-mediately, unless G is already working on energy affectively. If G is already occupied with an affective action, then the in-coming affective energy is queued. If energy enters on the lower path then it is affect-neutral and it enters a queue for rational processing once every resource allocation interval. When the interval occurs, then all of the queued resource requests are read by Prepare budget proposal and a percentage of them are passed on to the Integration function and the rest exit the organization and end up in Ideas not resourced. The percentage of ideas that are not resourced is set on the spreadsheet and remains constant for the duration of the simulation.

Energy enters Integration from two sources, too, both from Goal Attainment. If the energy is to be responded to affectively then it goes directly to the Integration func-tion. If the Integration function was already working affectively-neutral, then that process is suspended and held in Interrupted Integration until the affective processing is completed and then it is restored for the remainder of its time. If the Integration function is working rationally on energy when the next batch of rational energy to be integrated arrives, that arriving batch waits and is processed one at a time on a first come, first served basis. If the energy takes the Affect path from Goal Attainment to Integration and Integration is already working on energy that is to be affectively integrated, then it waits in the queue on the Affect path until the Integration function has completed its processing of the current affective activity.

After the energy is integrated it passes to Latent Pattern Maintenance, where in the current model nothing happens except that the energy is passed out of the organiza-tion, out of its system boundary. L has already had its effect by potentially altering the Adaptation function every time energy leaves Adaptation. The alteration of the filter is truly latent here.

It is very important to note the cumulative delay that has occurred between when the energy first enters the system and when it finally impacts Latent Pattern Maintenance. The delay is the sum of the processing times in Adaptation, Goal Attainment, and Inte-gration, plus waiting times. It is not insubstantial. Delay in time-varying systems can

Page 67

cause many kinds of dysfunctional behavior, including most notably oscillation as the organization tries to respond to reduce tension.

Rules All models are the union of their parameters and rules, as mentioned on page 19.

Here is a list of rules programmed into the simulation. Table 4. Rules of the simulation.

Rule Real world interpretation/application 1. The simulation is deterministic.

There are no random values. All val-ues are entered before the simulation begins.

If the same values are used then the same results will obtain. There is no randomness.

2. Energy comes in bundles, a pair of values: magnitude and type. The val-ues of the magnitude and type (whether it is processed as affect or affect-neutral) are fixed for the simulation period.

Think of energy as "news," the kind that comes from newspapers and other media broadcasts. Then the power of the news mes-sage itself contains whether that information will be handled in a rational or emotional way.

3. There is only one place in the system where the energy can enter from out-side: the Adaptation function. In its role to scan the environment, Adap-tation will permit energy to enter the system if it passes certain tests.

News can only come in through one door in the organization.

4. The tests are: If the magnitude of the external energy is in the appropriate relation to the target value, then the energy is permitted to enter and tran-sit the system. Else it leaves the sys-tem. "Appropriate relation" means that that its value "passes" the rela-tion, where both the value and the relation are found in the spreadsheet. "Passes" means that "External-energy Relation Value-in-spreadsheet" is true.35

If the news is not significant enough then it does not rise to a level of sufficient to get the attention of the organization.

5. If the energy is the type that will be processed as affect, then the proc-essing times are selected from one set of cells on the spreadsheet, else they are selected from the other set.

Information that is sensed as needing to be responded to emotionally is processed in a shorter duration than that requiring rational administration.

35 For example, if the external energy is of magnitude 5, the spreadsheet relation is >, and the spreadsheet value is 2,

then the energy bundle passes because 5>2 is true.

Page 68

6. Adaptation and Goal Attainment can work on only two energy bundles at once, one requiring affect and the other not requiring affect.

The capability of the organization to respond to news is limited to one emotional event and one rational event at the same time.

7. Integration and Latent Pattern Maintenance can only work on one energy bundle at a time.

Only one (major, funded) process can be integrated at a time, be it rational or emo-tional, e.g. Total Quality Management or the loss of the CEO. And LPM can respond to only one event at a time, too, thought its effect can be quite long-lasting, as it controls how much news is let in.

8. If A, G, or I are finished processing an energy bundle but its successor is not ready to accept the bundle, then the bundle is put into a queue between them and the function is given more energy to process if any has arrived at that point in the cycle.

News waits to be responded to, it does not drop or go away.

9. If energy is the type that will be proc-essed by affect, then the (media inter-change) path between A and G and G and I are different than if not proc-essed by affect.

News that will be responded to emotionally takes a "fast path" through the functions.

10. Goal Attainment enqueues the energy passed to it from Adaptation and processes it all at once at a given interval if not affect; if affect, then it is processed as it arrives, consuming the affect-neutral delay according to the spreadsheet.

Goal Attainment simulates the rational budget process (if the news is affect-neutral) and queues resource (that is, funding) requests until a definite period has transpired, such as every six months.

11. Not every queued affect-neutral energy package transits from Goal Attainment to Integration, only a per-centage does. That percentage is set at simulation run time.

Think of these Goal Attainment energy pack-ages as funding requests. The Goal Attain-ment function processes the budget requests all at once every six months, passing on only a portion of them as "approved."

12. Integration addresses the incoming energy for the duration specified in the spreadsheet, if not affect. If affect, then any non-affect current integration efforts are suspended (that is, made to pause and put into a special queue) and the affect energy is given priority of Integration. After all of the interrupting (that is, affect)

Imagine that an integration activity is going on, such as implementing Total Quality Man-agement. Then the organization learns that its CEO is suddenly, unexpectedly deceased. The organization suspends the TQM initiative and focuses on succession and how to respond to the urgent news. After responding to the urgency, the organization goes back to implementing TQM.

Page 69

Integration has been processed, then the non-affect queue is resumed.

13. Latent Pattern Maintenance computes a new value for the Adaptation filter once per energy bundle that arrives at Adaptation. The new value is com-puted according to the learning model.

Every time there is news the culture responds by tuning its filter on what it will sense dur-ing the next news-gathering cycle. The tuning is accomplished as Latent Pattern Mainte-nance attempts to reduce the difference (ten-sion) between the news presented from the outside and its response to it inside.

14. Energy leaving Latent Pattern Maintenance leaves the system.

Once the culture has responded to news then its influence remains, passively.

Assumptions As in all modeling, the underlying theory may not provide enough details for a

machine to operate like the theory. Therefore, the modeler must make assumptions, always in the absence of concrete guidance from the theorist. The following is a list of assumptions made for model of the theory of action: Table 5. Assumptions made in the simulation.

Assumption Real world interpretation/application 1. Transit times through the framework

are significant, that is, they matter. The time it takes for news to pass from one function to the next is significant. This is because if a downstream function takes longer to respond to news than an upstream one, the next news will have to wait on or be lost or preempt the news being responded to by the slower func-tion. And those options are significant to how the organization respond to external events/change.

2. Transit times can be different, depend-ing upon whether the energy to be im-ported will be handled with affect or in an affect-neutral way. Affect-neutral energy can be processed in a longer time period in order to simulate the time to be consumed during rational consideration, and energy that will be addressed by affect can take a shorter period in order to simulate that non-rational actions can take significantly shorter duration than affect-neutral ones.

News that is handled in an emotional way does not take as long as that han-dled in a rational, studied way.

3. The modeled transit times are in units of a business day. There is no guid-

The granularity of "time" described in the simulation is the business day. There

Page 70

ance from Parsons on the numerical or relative values.

is no telling how many business days each function consumes.

4. The paths through Adaptation and Goal Attainment are separate and par-allel for affect and affect-neutral energy. That is, as energy enters the organization it is identified as affect or affect-neutral and then put on its own path.

There is a "fast path" through some functions in order to simulate the effects of immediate response implied by "emotional events" vs. the more studied and time-consuming one of the rational response.

5. The paths are separate because one is accelerated and the other is not. Energy on the accelerated path then may be handled differently than that on the affect-neutral path.

6. The affect-neutral path to Goal Attain-ment queues energy such that the Goal Attainment function empties that queue only every so many days, to simulate the periodic (that is, calendar-driven) resource allocation review process.

7. The paths come together at Integration because an organization can only inte-grate one set of processes at a time. Therefore, the energy that has been identified as affective preempts the energy that has been identified as affectively-neutral.

Integration is so consuming that only one organizational initiative can be ac-complished at a time. And emotional ones have priority.

8. Preempted energy is enqueued. That is, it is put aside and waits for the pre-empting force to finish and then it resumes. No energy is lost, it is stored for later use. Its strength does not diminish during storage.

Rational Integration functions that are interrupted by emotional ones are not lost, but rather are delayed by the time it takes to address the emotional one. Then when the integration of the emotional event has been completed either another emotional event can be addressed if there is one, or the rational event that was interrupted will be restored and continue to process as if nothing had interrupted it.

9. There is a single path through Latent Pattern Maintenance.

Latent Pattern Maintenance actually appears in two places: during Adaptation it changes the filter on incoming news to let more or less in based on its response to tension (the difference between what

Page 71

is outside and what is inside), and after Integration, where it does nothing and lets its lasting influence be the setting the of the Adaptation filter.

10. There are only two threshold relations of interest in the Adaptation function: if the outside energy is greater than the threshold, or if the outside energy is greater than or equal to the thresh-old.

The Adaptation function checks to see whether the strength of the outside news exceeds a certain level or whether it is at or in excess of a certain level. No other relationships are permitted.

11. There are three related quantities that could be conflated: the threshold for energy to enter the Adaptation func-tion, the value Latent Pattern Mainte-nance learns to use as the threshold, and the target Latent Pattern Mainte-nance uses in its learning.

The culture tries to reduce the tension between what is sensed externally and what it can tolerate internally. The cul-ture may or may not respond to a spike in the news, for example, if there is a strong cultural counter-force to ignore that presumably one-time event, so that event is smoothed away by looking at a longer-term trend instead. The culture selects how smooth of a trend it will use as input to what it learns as the trend of the real, external events.

12. The smoothing of the external energy values uses this procedure: if the new energy exceeds the maximum value already seen during the time period being considered (the "look-back" window) by a specified margin (threshold to respond to change), then establish a new maximum and that is the value to be used. Else move along a negative exponential curve from the maximum towards the moving average of the external energy (where the win-dow of the moving average is the same as for locating the local maximum).

If there is a spike (a very disturbing bit of news) then is it greater than the last maximum the organization can remem-ber? If it is, then that becomes the new maximum value. If not, then it is consid-ered a bump along the way to forgetting that maximally disturbing event. By whichever means the maximum is established, if it really is a spike, then gradually forget it in light of new, lower intensity news. This slowly seeks the equilibrium level of the daily news after a very disturbing event by gradually for-getting the effect of the disturbing event until the next very disturbing one.

Mapping the model to theory

This section is an abbreviated version of the parsing that the researcher performed on the 107-page Parsons, Bales and Shils (1953a). For every feature of the model the corresponding phrase in the working paper is identified in the table below.

Page 72

Table 6. Map of the theory to the model.

Phrase in Parsons, Bales and Shils (1953a), with original orthography and punctuation

Interpretation and place in the model

We make four major assumptions in our analysis of boundary-maintaining systems which are composed of a plurality of units, or "particles". We assume first, the princi-ple of inertia, namely that a unit or "parti-cle" always tends to move in the same direction at a constant rate unless deflected or impeded. p. 164

Parsons was referring to Newtonian inertia, something that is observed (or defined) in the physical world. Surely the terms "direction" and "constant rate" have a dif-ferent meaning in an organizational setting. By direction, the model assumes that energy passes first through A, then G, then I, and then exits the system after passing through L. By rate, the model assumes "cognitive rate," the rate at which energy is made sense of in organizations. Since there is no conversion factor to rates in the physical world, the simulation lets the user set the duration (called dwell) for each quadrant; the simulation permits different values for affect and affect-neutral energy, so there are eight possible user-set dura-tions (four functional prerequisites x two values of a single pattern variable).

In no concrete case of system-process can this constancy of direction and rate be maintained for any span of time, since the interdependence of units is the very essence of the conception of system. p. 164

Indeed, as energy transits the system the rate of process can vary (in this research deterministically, not randomly). There are several cases where direction can vary: at the outset energy might not be let into the system due to the setting of the filter in Adaptation; not all energy (news or ideas) will be allocated resources in Goal Attain-ment, so some will travel onward (those approved) and some will exit the system; and if the energy is to be affectively responded to then in Integration it can at least temporarily displace activities that were being addressed affectively-neutral.

The unit in a stable state of the system will tend to follow a sequential pattern of changes of direction as its relations to the other units in the systems and to the exter-nal situation change over time. p. 164

The fixed set of rules and assumptions guarantee that a sequential (in this case deterministic) pattern, both in relation to each of the predecessor functions and the external situation over time.

This sequence may be oscillatory or cycli-cal or it may have some other form, but it

The pattern in this research depends pri-marily on the pattern of the external energy

Page 73

will always involve changes of direction (and of rate). These changes will always follow a pattern, although there may be some random elements intermixed with the pattern. p. 164

over time, and the initial settings of the values that control dwell times, rate of learning and forgetting, and the window over which the system looks back in which to formulate its latent pattern maintenance response.

We assume the principle of action and reaction tend to be equal in "force" and opposite in direction. We interpret this to be another version of, or a premise under-lying, the conception of system-equilib-rium. No more than the statement of the principle of inertia does the statement of this principle imply that actions and re-actions empirically are always equal and opposite; it does imply that where they are not equal and opposite, a problem is pre-sented. p. 164

The implementation of "equal and oppo-site" is the setting of the filter on Adapta-tion by the Latent Pattern Maintenance function. If Adaptation lets in "too much" energy LPM compensates by decreasing the amount to be let in in the future and mutates mutandis for "too little." And if the reaction of LPM is not appropriately "equal and opposite" then tension increases, which in turn increases the pressure on LPM to adjust the incoming energy.

We assume the principle of acceleration which asserts that changes of rates of proc-ess must be accounted for by "forces" operating on (or in) the unit(s) in question. An increase of rate implies an "input" of energy from a source outside the unit in question, and decrease of rate, a loss of energy, and "output" of some sort from the unit. p. 165

The model does not handle this. Rates are not adjusted based on the consumption of energy. There are increases and losses of energy in the sense of sinks and sources in the model, those changes do not affect the rates of anything.

We assume the principle of system-inte-gration. We interpret this to mean that, independently of the operation of the other three principles, there is an imperative placed on systems of action which require that pattern-elements in the organization of their components should be compatible with each other while maintaining the boundaries of the system vis-a-vis its external situation. p. 165

The system and its boundary are given in the simulation and cannot be changed. On the other hand, the components are com-patible with each other in the sense that they non-destructively interchange infor-mation. If there is a question of coexistence of the system in its environment, then it is reflected by increased tension, that is, by an increase in the difference between the pat-tern of energy of LPM and the pattern of energy external to the system.

Central to our scheme is the conception of action as a process occurring in or consti-tuting boundary-maintaining systems con-ceived within a given frame of reference. This frame of reference involves, above all, the four dimensions … and … four pattern variables. p. 165

The model simulates action as process by the four dimensions and one of the four pattern variables.

Page 74

An orientation cannot be both affective and neutral [at the same time]. p. 166

Each bundle of energy crossing the system boundary is tagged as either that which is reacted to by affect or by affect-neutrality. There are no degrees of affect; affect and affect-neutrality are mutually-exclusive and exhaustive.

The dimensions are, we assume, essentially directional coordinates with reference to which the process of action is analyzed. Motivational energy entering the system from an organism cannot simultaneously operate in all possible processes which go to make up the system. It must be specifi-cally located, in the sense that it must be allocated to one or more units of the sys-tem. But at any given time this unit must be located at some definite point in the action space, and must be moving … in a definite manner. p. 166

Energy enters the system and moves along pre-determined paths, operating in one place at a time. Energy in the simulation is always specifically located. And the units are connected by definite interchange paths, along which energy and information move.

The system operates through interaction of its member units. Every change of state of one unit … will affect all of the other units in the system and in turn the effects of these effects on the other units will "feed back" to the original unit. p. 167

This describes interchange media, of which only two types have been implemented in the model: feed forward, the forward transit of energy, and a single feedback path from LPM to Adaptation that sets the filter on A to determine how much energy to let in. So, there is only a single instance of feed-back in the current model; it is not fully-connected.

We derive the conclusion that systems of action must be treated as differentiated systems. It then becomes clear that this dif-ferentiation will work out in two ways. Since we are dealing processes which occur in a temporal order, there we must treat systems and the processes of the units as changing over time. p. 167

Time is a fundamental construct in the model, indeed in discrete-event simulation.

The one way character of the process we have deduced from the nature of motiva-tional energy—the fact that it is "expended" in action. We assume through-out … there is, if not a law of conservation of motivational energy, a law of "equiva-lence" in the sense that this energy does not simply disappear, but, "produces" some kind of consequences, that there is a bal-

The energy that transits the system does produce a kind of reactive consequence: the setting of the filter on the Adaptation func-tion so that the inputs and outputs are, indeed, in balance. It does this in a numeri-cal, quantitative way, but without any measurement in The Real World. That is, the numerical aspect was created in the model for purely illustrative purposes.

Page 75

ancing of input and output. We believe, though at this time we cannot prove, that this is a quantitative balance which will eventually proved to be reducible to terms of numerical equivalence. pp. 168-169 As a process of output contrasted with the inputs of motivational energy, of percep-tion of objects, of facilities and rewards as the two primary categories of possessions, we thus treat the learning process as "oppo-site" in directionality from the motivational input processes. p. 170

Learning is manifest in the simulation as a force against Adaptation taking in all (mo-tivational) energy in the environment. Rather, learning is a counter-force in that it limits, potentially reducing, the amount of energy "ingested."

The distinction between performance and learning aspects of action process forms the basis for a further classification of types of process in systems…. In general the type of analysis … presented … provides a model for the typical performance process where the primary concern is not with changes in the properties of the group and its constitu-ent role-units, but with task performance, that is. in the terminology we are adopting here, with attainment of a system-goal. p. 170

The model focuses on performance, not learning (except classical conditioning), on attaining a system-goal and does not address "changes in the properties of the group and its constituent role-units." In fact, there are no modeled properties or roles.

In the absence of an adequate mathematical model, any feasible form chosen necessar-ily involves elements of arbitrariness which all too readily become distortions. We have chosen one such mode of presentation …, but in order to counteract any tendency to reify such a scheme, we have thought it best to say explicitly that it is arbitrary, that there are many possible types of models appropriate to the fundamental ideas, that we have experimented with several and are looking for others. We feel that in the pre-sent stage of development of this type of theory it is exceedingly important to be highly pragmatic about these matters and to try out a variety of devices. Only in such a way can we be protected against a pre-mature rigidity of formulations…. pp. 171-172

The simulation is a toy, a proof-of-concept, a stepping off point for further inquiry.

Page 76

V. RESULTS The results of the simulation are divided into three broad areas: an example, base

cases, and an extension. The base cases and extension are quite similar in their pattern of presentation: briefly explain the theory and what it would predict, indicate what the inputs to the simulation were, and then illustrate the output of the simulation run com-pared to the theory prediction. They directly address Parsons' theory of action. The example has been offered to provide some concreteness to the application of the theory, a topic largely beyond the scope of this research.

Example Parsons' theory is opaque, so the model of it was correspondingly abstract. An

example taken from real events might aid the comprehension of the theory and therefore the model. At the outset we must be mindful of Parsons' admonitions about misplaced concreteness (see p. 28 above), about the value of analytic thought, so this example must be disclaimed from the outset as present here for illustrative purposes only. Nothing is intended to be proved by it.

Up to now patterns of flows among functions have been described (Parsons' phases), but there have been no acts! The example here is an attempt to show how the flows and functions could describe actual human action.

For the example a situation was sought in which the energy outside of the system was relatively uniform for a long time (stable) and then there was a jolt, an impulse of sudden energy, mirroring some of the events to be presented below in the base cases and extension. Waller (1999) examined the order and timing of events in a commercial airline cockpit simulator during training drills with real airline flight crews while they were addressing "nonroutine" events, the kind that were associated with high outside energy.

On the one hand a real situation was sought, but the example explored here was, too, a simulation of such a real series of events. The problem is a scarcity of reports about real world events in which timing and order are recorded, along with outcomes. There-fore, a simulated though realistic setting is presented.

There were ten flight crews of three persons each, so each crew was a small group, the kind that exhibits collective behavior. The setting was naturalistic, as such training simulators are constructed precisely to mirror real world situations and condi-tions. The nonroutine events were arranged in a sequence of six unexpected items of news during a planned 60-minute flight from Los Angeles to San Francisco.

The unexpected events were: 1. Poor weather forecast; bad weather at San Francisco and its alternates; heavy

takeoff weight. 2. Light to moderate turbulence during the climb and cruise phase. 3. Fast, noisy descent required by air traffic control during approach to San

Francisco. 4. The approach was missed due to hydraulic failure; crew must select an alter-

native destination (Sacramento). 5. During the flight to Sacramento emergency procedures had to be performed,

including trying to manually extend the nose landing gear and force the flap that experienced the hydraulic failure.

Page 77

6. During the landing the crew had to compensate for the non-responsive flap, no steering possible with the nose landing gear, and high landing speed because the non-working flap also acted as an air brake.

Waller hypothesized that success at handling nonroutine events would depend upon information collection and dissemination, task prioritization, and task distribution. She noted that these all dealt with the level of the behavior, but not the timing [emphasis hers]. She noted, for example, "rather than viewing the time of change as a function of internal stages or clocks, the time of change may be seen as more tightly linked to exter-nal events." p. 130, relying on Ancona and Chong (1996, p. 263)

Therefore, she hypothesized and tested for timing by studying whether there was a relationship between the time an external event occurred and when it was reacted to. Waller found that, for example, there was no difference in the level of workload between crews that responded quickly and those that responded less quickly, though the crews that responded quickly to external events all performed much better than the crews that responded less quickly or did not let the external events come to their notice. In other words, there was no difference in the level of behavior, but the difference in timing made the significant difference in crew performance outcome. As Waller pointed out, the higher performing teams did not work harder, did not perform more tasks, but did achieve more, all because of timing, because of noticing and following significant external events. (p. 134)

Two scenarios are described within the theory of action simulation, one with a relatively long window and one with a relatively short one. The window, as one may recall, is how far back the Latent Pattern Maintenance function looks back in order to "remember" what happened historically. Strong cultures look back a long ways and weak ones a short time. Here was the input from the user for the long window, the whole flight.

Page 78

Dwell time ValuesAdaptation

Affect-Neutral (periods) (Should be greater than Affect) 3Affect (periods) 1

Goal AttainmentAffect-Neutral (periods) 10Affect (periods) 1

IntegrationAffect-Neutral (periods) 30Affect (periods) 5

Latent Pattern MaintenanceAffect-Neutral (periods) 0Affect (periods) 0

Percentage of ideas not fundedPercentage (expressed as decimal < 1.0) 0.20

Prospensity to changeEnergy filter threshold (initial value) 1Sense of test: Pass Energy if Energy [operator] Filter Threshold >=

Learning/ForgettingPrior learning (p) 1.00E+08Time to reach current pattern (r) 1.00E+07Starting value of L energy (k) 1

Negative exponential decay parametersThreshold to respond to change (in energy units) 1Decay coefficient (<0 and in time units) -0.001Response window (in number of Affect-Energy pairs) 60

Figure 16. User display for example with long window.

Each period of the simulation is one minute and there are 60 periods, to mirror Waller's experiment. The pattern of external energy, not shown, is eight minutes of rela-tive calm followed by a single one-minute message of high energy (Energy=4) that has to be dealt with affectively. Figure 16 shows that the assumed period of prior learning is approximately ten years (in minutes! "1.00E+08" is 108 minutes) and the period of time to reach the current level of expertise is one year. All external events are permitted to enter (Energy filter=1 and Threshold to respond to change=1). The Adaptation function looks at the external environment once every minute.

Here is the pattern of internal and external energy, assuming such a strong culture that the strength of the culture remains constant during the 60-minute flight.

Page 79

External energy vs. LPM energy

00.5

11.5

22.5

33.5

44.5

0 10 20 30 40 50 60

Simulated time (minutes)

Ener

gy le

vel

InternalExternal

Figure 17. Energy for the long window example.

One can see that after the first (of six) non-routine events, the culture only lets in really high energy events as an indication that it has learned that such non-routine events can occur and that attention has to be paid to them immediately.

Now we make a single change: the window is reduced to a few minutes, as if the crew forgets the (disruptive) impact of each non-routine event. The window is set to five minutes and here are the results:

External energy vs. LPM energy

00.5

11.5

22.5

33.5

44.5

0 10 20 30 40 50 60

Simulated time (minutes)

Ener

gy le

vel

InternalExternal

Figure 18. Energy for the short window example.

Page 80

The energy bounces up and down as the culture more closely follows the pattern of the external energy, with a non-routine event every nine minutes. This, too, is pre-dicted by the theory of action because a weak culture (= short window) will respond much more quickly to changes in outside energy and therefore will maintain less of a pattern, will enforce less of a culture.

In summary, Waller notes that groups can "match the rhythms of [their] task-ori-ented behaviors to exogenous events, rhythms, or deadlines." p. 135. This is precisely what the model described in this dissertation: better performance is achieved by matching the energy in the environment with the energy circulating internally, presumably con-trolled by the most powerful function in the cybernetic hierarchy, Latent Pattern Mainte-nance.

In the Waller example cast in Parsonian terms, the crew executes the Adaptation function itself, sometimes by asking for news (such as weather conditions) or by noticing indicators (such as the nose gear not engaging and the noisy approach descent). Based on its sense-making during Adaptation the crew determines whether each event is routine or not. When it was not routine, then the best crews responded to it affectively, accelerating the transit of the event through the crew's equivalent of Goal Attainment (redirect atten-tion towards the new event, immediately "approving" it for (attention) funding), and Inte-gration (executing the standard procedure for that unexpected event, but interrupting or suspending standard processing).

Base cases Affect vs. affect-neutrality

Affect and affect-neutrality are traditionally ascribed to each of the four functions: affect to G and I, and affect-neutrality to A and L (see Figure 5). That is, A and L are to be more cognitive, rational, thought-out, and G and I address gratification and emotional aspects, they are not rational. There is also the view that energy dealt with affectively transits the functions more quickly than energy that is dealt with in a studied, reflective, rational way. One case, then, examines the extent to which energy that is dealt with affect passes through the organization more quickly than energy dealt with affectively-neutral.

One run of the simulation had the following values:

Page 81

Ver. 0.9 For The Parsons Game by Stan Rifkin1 2 3 4 5 6 7 8 9

Dwell time ValuesAffect Energy Adaptation

(1=true) (Level) Affect-Neutral (periods) (Should be greater than Affect) 100 2 Affect (periods) 50 2 Goal Attainment0 2 Affect-Neutral (periods) 1200 2 Affect (periods) 200 2 Integration0 2 Affect-Neutral (periods) 1200 2 Affect (periods) 600 2 Latent Pattern Maintenance0 2 Affect-Neutral (periods) 00 2 Affect (periods) 00 2 Percentage of ideas not funded0 2 Percentage (expressed as decimal < 1.0) 0.800 2 Prospensity to change0 2 Energy filter threshold (initial value) 20 2 Sense of test: Pass Energy if Energy [operator] Filter Threshold >=0 2 Learning/Forgetting0 2 Prior learning (p) 5000 2 Time to reach current pattern (r) 2500 2 Starting value of L energy (k) 20 2 Negative exponential decay parameters0 2 Threshold to respond to change (in energy units) 10 2 Decay coefficient (<0 and in time units) -0.0010 2 Response window (in number of Affect-Energy pairs) 45

10

Figure 19. Base case for affect vs. affect-neutrality.

The simulation ran for 2400 simulated business days (about ten years) and an energy stream of all 2's that were to be dealt with affectively-neutral, except that every 240 business days (approx. one business year) there was an event of energy 4 and it was to be dealt with affectively (which one would assume, as it represents a large departure from "normal"). Accordingly, there were ten such events of magnitude 4, including one on the last business day simulated. The results: there were 480 events (one every business week, which was the frequency of scanning the environment by the Adaptation function), 408 events did not pass through the Adaptation filter and therefore were not processed further, 49 were not resourced, four were in processing in the four functions, and 22 completed the journey through all four functions. Of those 22 were ALL nine of the mag-nitude 4 events that were to be dealt with affectively, leaving the remaining 14 to be the "normal" affectively neutral events. Clearly, the affectively-neutral events speeded through the organization. Strong vs. weak culture

Organizations with strong cultures have, in essence, a strong Latent Pattern Maintenance function, one that restores any disturbances that might enter. In fact, one of the ways LPM can limit disturbances is to not let them in in the first place, by limiting the information/energy that the Adaptation function lets pass. By restricting the filter on the Adaptation function, LPM limits the excursions of energy inside the organization. In a weak culture, the organization more closely follows the external energy, in a sense tracking it with all of its ups and downs.

Figure 19, above, represents the tableau of a strong culture. First, its memory win-dow is long, 45 events; assuming one event per five days, that's almost one business year. In other words, the reaction time of this organization can be delayed by a year, after

Page 82

which it again reacts to the outside energy. Here is the pattern of external energy and the pattern of following it internally:

External energy vs. LPM energy

00.5

11.5

22.5

33.5

44.5

0 480 960 1440 1920 2400

Simulated time (business days)

Ener

gy le

vel

InternalExternal

Figure 20. Pattern of internal energy following external with a strong

culture.

As can be seen, there are annual jumps in energy to a value of 4, with long peri-ods of 2 between them. This strong culture "forgets" the high values over time until another one hits and then its internal energy jumps up again to follow it. The figure illus-trates the deterministic, repeated pattern of how internal energy followed external pertur-bations. Consistent with the prescription of classical conditioning, there was no long term learning, the pattern of internal energy is completely determined by the pattern of exter-nal energy.

By changing just two parameters, the window to 25, about half of the previous value, and the Threshold to respond to change from 1 to 0.5, the organization mirrors a weaker culture. Here are the results, with the same stream of energy as in the figure above:

Page 83

External energy vs. LPM energy

00.5

11.5

22.5

33.5

44.5

0 480 960 1440 1920 2400

Simulated time (business days)

Ener

gy le

vel

InternalExternal

Figure 21. Pattern of internal energy following external with a weak

culture.

In this case the organization gradually forgets the high values and then when the window has passed it says to itself, in effect, "Let's stop responding to old news and get synchronized with what is happening now, let's loosen the reins a bit and let some new energy in." But, again, consistent with the prescription of classical conditioning, there was still no long-term learning, the pattern of internal energy is completely determined by the pattern of external energy. The only change was the period of looking back.

And here are the results in an organization with really weak culture: the window was set to about one month. This would be the case in an organization where something like the terrorist attacks of 9/11 happened every year and within a month the organization was incorporating the weekly news, as if nothing had happened. It would be as if there were no heritage, no legacy. No pattern maintenance.

Page 84

External energy vs. LPM energy

00.5

11.5

22.5

33.5

44.5

0 480 960 1440 1920 2400

Simulated time (business days)

Ener

gy le

vel

InternalExternal

Figure 22. Pattern of internal energy following external energy with

very weak culture.

Extension The outcomes in the previous cases were easily predicted by the theory of action.

Here is a case in which there is no theory to guide predictions. In some sense, the simula-tion is the prediction.

In this case, Figure 19 is used. The pattern of inputs was varied slightly: instead of there being 49 weeks of a constant value of Energy=2 and no affective processing fol-lowed by a single week of Energy=4 with affect, there are 25 weeks of Energy=2 with no affect followed by one week of Energy=3 with affect, followed by 25 weeks of Energy=2 and no affect, followed one week of Energy=4 with affect. In all there are 52 weeks, during which there is an energetic event in the middle and one at the very end, both the only two events to be dealt with affectively. Everything else is the same dull news, to be dealt with affectively-neutral.

Here was the simulation at the end of 2400 business days:

Page 85

Figure 23. Simulation after two energetic events per year, both with affect. Illustrates queuing effects.

While it is difficult to read, the small numbers inside the main quadrants repre-sented the number of energy bundles presently being processed. As can be seen, there were 28 pending "requests" for Integration, along with 16 Integration processes that were interrupted while higher priority ones were being processed (presumably those that had to be dealt with affectively). Why are they all waiting? It is because the current Integration activity is processing energy affectively. So, with the current values, it will be about two times the 60 business days each that each affective process will wait to complete Integra-tion (that is, about 120 business days, six months) before the first affect-neutral Integra-tion process could even begin. A total of 44 affective-neutral events were funded in Goal Attainment and all of them are queued: 28 in the Integration queue and 16 of them in Interrupted Integration. The queues build because, according to Figure 19, each value for the dwell time increased as the energy made its way around the AGIL circuit. This was logical because Adaptation took less time than Goal Attainment, which in turn took less time than Integration.

To reiterate, all 17 items in Spent Energy, those that have completely transited the organization, are limited to those that were dealt with affectively. That is, in the ten years simulated NO affect-neutral events were processed all the way through! That is due to the

Page 86

frequency of the events that had to be dealt with affectively (two per year) and the long time it takes to integrate the responses to them (60 business days, three calendar months).

Therefore, one of the extensions to the theory of action is the impact of time spent in each functional unit as a function of the rate at which inputs and messages arrive. If the time spent is on average greater than the inter-arrival rate of the inputs + messages, then queues will build. This is a fundamental principle of queuing theory (Kleinrock, 1975-1976). Parsons did not write about what happens when some energy has to wait three years to be integrated.

In sum, the results of operating the simulation both were accurately predicted by the theory and without effort demonstrated potential extensions to the theory. The results were collectively an encouraging step towards a workbench for scientists to use to char-acterize and experiment with their understanding of social systems.

Page 87

VI. CONCLUSIONS AND RECOMMENDATIONS FOR FURTHER STUDY Lave and March (1993) stress the indispensable importance for modelers to

recognize when they are wrong. "A final protection from the danger of believing too fer-vently in a theory [of your own] is to be intellectually playful." (p. 61) [emphasis in original] And in another place: "Play to your analytical strength. Do not be afraid of twisting a phenomena around a bit to make it fit into an analytical scheme that can drive some implications for you. Do not hesitate to look for phenomena that can be examined usefully with the models and techniques you have." (p. 73) One of the key points of the research reported in this dissertation is that attention to timing of social systems events can have an impact on theory. But if timing were not a central focus of the theory of action in the first place, then is the time-related research of no account?

Review of purpose and research question The purpose of this research was to see if Parsons’ large body of descriptive text

could be understood well enough to fashion an animation of it. A partial response was that that was beyond the scope of a dissertation-type research. Rather, a single, indicative work of Parsons(1953a) that described the dynamics of the theory of action was selected as a stand-in for the totality of Parsons' works.

Can the salient factors (structure and function) be extracted? Saliency is clearly in the eye of the beholder. A more concrete response is that a simulation was constructed and in the dissertation committee (and the researcher, of course!) it performed according to the theory.

Finally, was it possible to instantiate, make concrete, those salient factors so that a high fidelity representation of the descriptive theory of action can be constructed? Again, fidelity is in the eye of the beholder, and the dissertation committee agreed that the fidel-ity was sufficient to demonstrate that it was possible to simulate the selected aspects of Parsons et al. (1953a).

Finally, the question guiding this research was "What is the minimum set of struc-tures and related functions that can simulate Parson’s theory of action to some criteria of validity?" That is, what was the most parsimonious selection of theory of action con-structs that, when animated, achieved a given level of fidelity? Can the theory of action be simulated using only the functional prerequisites, (one pair of) the pattern variables, (four of) the interchange media, and the cybernetic hierarchy of social control?

The research did not really address minimality as much as it appeals to the reader's sense of parsimony and asks rhetorically, "Could the theory be simulated with fewer structures and corresponding functions?" This question is the opposite of the usual one, which in the instant case might be "How many structures and functions can be included (jammed?) into a simulation in order to make it high fidelity?" The stated level of fidelity of the simulation described here is one that can serve as a building block for the next researcher to construct a more accurate or more general model of the theory of action; this research was a demonstration of possibility, of proof of concept. Accordingly, the best judge about whether the simulation shall achieve its purpose will be to interview the next person in line to use it!

Review of findings Accordingly, the primary finding is that a toy, proof-of-concept simulation of Par-

sons' theory of action can be constructed and operated. It was constructed mentally and in computer programming terms by scaffolding feature-after-feature, much the way the next

Page 88

researcher might use the model, which adds weight to the practicality of the future pros-pect of a more extensive (deeper and broader) exploration of the theory of action. That is, the current research was conducted by beginning with a modest baseline of operational capabilities and successively adding to it in order to increase the coverage of, and there-fore fidelity to, the theory.

The fidelity was illustrated by several "base cases," whose outcomes were pre-dicted by the theory. One case was for a strong culture, one with a strong Latent Pattern Maintenance function that could remember for a long time. Theory would predict that such an LPM function could reset new information entering the organization by quickly decreasing the amount of new information permitted in until the organization was "over" (in the sense of forgot) the out-of-the-ordinary impulses. Another case of weak culture illustrated what the theory of action predicts: the organization under study closely follows the pattern of external information (as though it had no memory) and was therefore "whip sawn" by the shape of external events; there was virtually no counter-force to the im-pulses entering the Adaptation function from a possibly-turbulent environment.

On account of these base cases, one can have a degree of confidence that the model enacts the theory. In addition, an Appendix contains the attestation of an expert on the simulation language selected that the modeling and the model achieved what was sought.

Discussion As presented above in the Literature review, p. 25, there was a notional set of

steps to be taken to build the simulation: Table 7. Correspondence between what was required and what was developed.

What was required (Hanneman, 1988) What was developed Define boundaries of the system. The system boundary was defined in terms

of "internal" and "external," meaning inside and outside of the system or organization under study. The important element that came from the outside and was evaluated and contingently transformed inside was energy, notionally in the form of news, information.

Define the elements of the state space and partition the state space into subsystems.

The partitioning was given by Parsons' four functional prerequisites, so there are four elements of the state space. All four were simulated. In addition, Parsons' defined four pairs of pattern variables that will define the dynamic aspects, below. A single pair, affect and affect-neutrality, was simulated.

Describe the connectivity of the state space elements, and the forms of relations among the states of the system.

Parsons defined a fully-connected space using interchange media as the paths among all possible combinations of ele-ments. That is, there was a 12-part bi-

Page 89

directional connection in Parsons' model, but only four of them were simulated. And only a single connection between inside and outside was defined, namely the one-way connection from the outside to the Adaptation function.

Define the dynamic aspects of the relations among state space elements.

Parsons stated that energy passed along interconnections from each state space element to each other one. The simulation defined only two paths, really: a clockwise one from A -> G -> I -> L and a counter-clockwise one from L -> A. Another dynamic aspect suggested by Parsons but not described in much detail was the classi-cal conditioning that L learns as it reduced tension. One more dynamic aspect was that energy that was to be treated affectively had a faster path through the state space than that treated affectively-neutral. In addition, there were other dynamic aspects that were in the simulation but not in Par-sons: energy queued when the next func-tion was not able to absorb it, a target value of external energy was selected in order to achieve gradual learning, under affect-neu-tral conditions G empties its queue all at once, and it could take longer and longer for each function in the cycle of A -> G -> I.

Latent Pattern Maintenance Latent Pattern Maintenance has a profound and profuse effect on the other three

functions, according to the theory. Qualitatively, LPM reset every function whenever a disturbance entered from the outside. LPM performs the reset in order to preserve what it has learned is the value to keep the difference between the energy outside and the energy inside (the difference is defined as tension) within a threshold. According to Parsons the style of this learning is classical conditioning (Parsons et al., 1953a, p. 226). The chal-lenge in the research here was to translate the notional, qualitative learning into one that could be simulated, that is, was quantitative. The simulation enacted the computation of a reset value according to the formula in Nembhard and Uzumeri (2000). It is a subject of future work to identify other, potentially better, formulæ for learning. Time

As mentioned in the Literature Review on p. 34, time is almost never taken into account in social systems studies. But time was the organizing principle for the research presented here, so it differs markedly from other sociological tracts. The way time was presented in the theory of action was in terms of "before," "during," "after," "longer," "shorter," and "then." There are sequences of patterned action described in (Parsons et al.,

Page 90

1953a), in particular phases and cycles. Sequence by its definition suggests ordering and therefore time and timing. The translation of Parsons' time to the model developed here was aided in a straightforward way using the trick of discrete event simulation, a method of simulation that explicitly specified order and time.

The challenge with respect to time was obtaining times (intervals) for the duration of each of the four functional prerequisites. It was not sufficient to leave that to the user. The arbitrary unit of a single business day was selected as the atomic unit of time, with-out loss of generality. Then the user of the simulation specified durations in units of busi-ness days in the hopes that that somehow was consistent with what Parsons envisioned but never wrote. Again, the unit of time could be changed throughout the simulation to another other one, as long as it was the same atomic unit everywhere. That is, there is no subjective or social time in the simulation; all time is in terms of a clock tick or cadence of equal duration.

The impact of using uniform time instead of subjective time is not clear. If there were some way to model subjective time then the mismatch among durations of the first three functional prerequisites would still exist, queues would still build, and learning would still be time-based. That is, in the main the results would be the same whether time was modeled as uniform or subjective. Process

The description of social systems as process was introduced in the Literature Review on p. 36. There it was argued that the process focus imposed a heavy burden be-cause it required a rich description of the mechanism and steps by which states change inside an organization in response to external stimuli, as opposed to what one usually finds in School of Education dissertations, which are statistical analyses of scores; there is no dynamics, no detailed mechanism of how a score gets its value.

The heavy burden is manifest in writing a simulation because the computer has to be told everything! Not only was the structure and function to be made manifest for the computer, but also the many details about which Parsons gave no guidance: were there queues between functions, how exactly did Latent Pattern Maintenance learn (we know that it was classical conditioning, but what was the model and what were the values of its parameters?), how exactly did Latent Pattern Maintenance affect Adaptation (that is, how did LPM affect the energy that Adaptation sensed or not?), how did LPM measure or sense tension, and then what exactly did it do to Adaptation in order to present a counter-force to energy that disturbed the previous state, etc.

As the simulation results unfolded, another process question arose: did Parsons foresee that organizations had a capacity to respond that is finite over an arbitrarily small period? Did he foresee that the functions might get to a state where they could no longer absorb or respond to any more energy? And then would he have predicted what would happen? While these questions might properly belong in the section below on recom-mendations for further study, in fact rather they suggest the fruits of the process view, as taken during this research. Discrete event simulation as a technique

Most modelers of social systems use the techniques of system dynamics for good reason: there is a community of practice centered around MIT and other distinguished universities (e.g., System Dynamics Society), an excellent text with many examples (Hanneman, 1988), and there is a growing corpus of applications (quarterly System

Page 91

Dynamics Review and the annual Proceedings of the International Conference of the System Dynamics Society). However, system dynamics did not appear to be up to the task of modeling the theory of action, particularly using the direct words of Parsons as the oracle.

Probably the breakthrough in this research came when Parsons' description of the phase movement was seen as a partial time-ordering and then discrete event simulation techniques were applied to see if there was a fit. The single largest contribution of this research may be the application of discrete event simulation to a social system, as this was only the third recorded instance of such an application.

Those who most often use discrete event simulation are trying to understand how waiting lines form, so it was no surprise that the waiting lines in the theory of action were exposed. This, too, may be a contribution of this research, as the topic appeared to be neglected by Parsons, his supporters, and his critics. Other difficulties

The effort to model the theory of action was difficult for several additional reasons: (a) so much is written by and about Parsons; (b) what is written is difficult to understand; and (c) the paucity (well, complete absence) of empirical, time-varying results that could be used to verify the simulation.

Each of these difficulties was addressed, though not all to the same level of rigor. In order to not claim any relation to the totality of Parsons' work, a single chapter was selected as indicative, and then only a very small portion of it was selected to be simu-lated. There is likely no antidote to the difficulty of understanding what has been written by and about Parsons. And one can only try to triangulate among Parsons and experts, and then to have it reviewed by experts in order to increase confidence in the fidelity of the understanding.

Above all, this research should be seen for what it was: a small, toy experiment – without verification – to see if something bigger is possible. Only by seeing that bigger thing, produced by a future researcher perhaps built on the foundation presented here, can the import of the current research be assessed.

Implications For theory of social systems

One implication for social systems research is the consequence of framing inter-actions in terms of time sequences but not attending to the impact of those time sequences. For example, the instant research illustrated the impact of not attending to the relationship between arrival rates and service rates. That is, in an external or even internal environment of turmoil and "white water," (Vaill, 1996) a scan of that environment can identify many items that need the attention of the organization (high arrival rate). Will there be enough time (high enough service rate) to attend to them all? What happens to the ones not attended to? These are questions about which Parsons offered no guidance.

In addition to the typical problem of queues building when the average arrival rate exceeds the average service rate, there is also the question of priority queues and high priority processing. While usually the domain of industrial engineering and operations research, those topics entered this research during the simulation of an organization responding affectively to stimuli. In the affective case the service rate is faster, higher (Parsons et al., 1953a, p. 201), at least one reason for which is that affect is by definition emotional and its opposite, affect-neutrality is rational, reasoned, cognitive, and it takes

Page 92

longer to be rational than not rational. The faster service rate of energy being addressed affectively can compensate for a higher arrival rate of external energy. The dilemma is that the quality of decisions arrived at affectively is lower than those decided affectively-neutral (that is, rationally). As Fararo (2001, p. 157) avers, the problem for further research is finding the "sweet spot," the stable region, between the two. Fararo said it is the kind of theorem one would like to see for the operation of pattern maintenance (loc. cit.). For research in simulation

Clearly discrete event simulation is underused in social systems research; only two previous examples were found. Perhaps the most compelling reason for that is the paucity of social systems research that incorporates time; time is the main independent variable in discrete event simulation. In that sense Parsons was many decades ahead of social systems research. And it might be premature to suppose that sociology has caught up with his practice of seeing social processes as events in a time sequence, the kind of string of actions that are ideally-suited to be simulated in a discrete event framework.

Another force that might augur for additional application of the discrete-event approach is the increased cross-over between sociology and other computer- and mathe-matics-related disciplines. One finds the CMOT (computer and mathematical organiza-tion theory) community increasingly using engineering-oriented tools to address socio-logical problems. For example, Burton and Obel (1995), management scientists, have found that the design of an organization (structure) can be optimum, a term never used by sociologist or organizational designers. Burton and Obel cast the problem as one in linear programming, where an objective function was trying to be maximized (such as decision speed or decision quality) or minimized (such as communication expense or overhead, or rework), subject to constraints. This framing as a linear programming problem is an example of the intersection of two disciplines and that new techniques grew from it.

The application here is that perhaps the confluence of people outside of sociology being interested in the theory of action with the increased capabilities and expressiveness of simulation languages and tools might result in a renaissance in investigations of the meaning of the theory of action using the mechanism of a workbench that a researcher could manipulate to explore understanding. For practice

The scientist finds his reward in what Henri Poincaré calls the joy of comprehen-sion, and not in the possibilities of application to which any discovery may lead. - Albert Einstein36

There is very little here for the practitioner. After all, this is a simulation of a theory, in effect an abstraction of an already abstract theory. If there is one finding for a practitioner it is the power of affect and the need to balance it with affect-neutrality for the long-term health of the organization. This was not lost on Fararo (2001), who notes "A 'functional necessity' or 'functional imperative' for an ongoing social system is that the element of affective neutrality be built into it (i.e., action in some situations should take the form of disciplined attention to instrumental and moral considerations in priority over immediate gratification)." p. 137 36 In Alice Calaprice (ed.). (1996). The quotable Einstein. Princeton, NJ: Princeton University Press, p. 173, in turn

quoting from "Prologue" in Max Planck. (1932). Where is science going? New York: Norton, p. 211.

Page 93

One might be able to speculate under what conditions that the balance between affect-neutrality and purely affective response might be struck, namely that tension is not being tracked, that Latent Pattern Maintenance is not matching the internal energy of the organization with the energy of the environment and that the pattern of matching is diverging over time, getting worse, more tense. In that case, one cause might be that greater affect-neutrality (that is, cognition) is required in order to make higher quality decisions and those decisions in turn would better match the internal energy to the external.

Recommendations for further study Surely the most commonly-heard expression by a reader upon reaching the end of

this dissertation will be "Why did he stop here? This is just the beginning. There is so much further to go." The dilemma when attempting something never done before is to determine when a beginning has been accomplished, when is it time to declare the end of one phase so that another may begin (possibly conducted by another researcher).

This dissertation presented a simulation of a part of Talcott Parsons’ theory of action. Like all simulations its fidelity can be improved in a number of dimensions, in this case: taking more theory into account, being more accurate, being more general, being more user-friendly, accounting for more pattern variables and their interaction, having finer granularity, having courser granularity, having adjustable granularity, and dealing with more interchange media. In fact, one way to generate a list of considerations for further study would be to systematically address this study's Limitations section, p. 52.

The decision about where/when to stop was based on a single judgment: could another researcher pick up where this work left off and continue along a path of refine-ment or generality? While the author could have gone further (in fact, without limit), it is the researcher's judgment that the state of the simulation is complete enough so that an-other person can carry on. Accordingly, while this work could have continued, it is also true that other researchers can join in the fun now.

[The model here] exhibits a logic of "theoretical models in progress." This usually means starting with initial simplifications and then adding complications in suc-cessive revisions. In [computer] programming terms, any one theoretical model becomes successively embodied in a series of program, the later programs cor-recting and extending the earlier. … At any one point in this series of develop-ments, a simulation model is both a theoretical model and a program. There is really never a last program in the series, only a place of rest or termination through exhaustion of the creative possibilities or diversion into work on other such projects. (Fararo & Hummon, 1994)

Increase simulation fidelity As mentioned in Delimitations and Limitations, above, fidelity can be increased

infinitely. In particular, there are four obvious areas of concentration: 1. The number of interchange media could be increased from four to the full comple-

ment of 12. See Figure 4. 2. The number of interpenetrations could be increased from zero to at least one, as illus-

trated in Figure 2 and Figure 3. That is, in addition to the four functional prerequisites alone, each of them has inside four of its own and this could be simulated, too.

3. The cybernetic hierarchy could be simulated explicitly.

Page 94

4. The remaining three pattern variables could be simulated, along with their (combi-natorial, multiple order) effects on each other. Dubin (1960) gives insight into the size of the combinatorics and suggests a probability associated with each pattern variable occurrence, something quite beyond Parsons' conceptualization. While there, though, it would be possible to augment the measure of the magnitude of external energy (0 to 4 in the current simulation) with some measure of the certainty (or ambiguity) of the information communicated by the energy, and thereby give the probabilistic approach greater richness in explaining potentially non-deterministic situations. Again, we note that this was quite beyond what Parsons explained.

In addition, the unit of analysis could be changed either up (to culture) or down (to personality) or one embedded in the other, which is a variation of item 2, above.

Zelditch (1955) adds considerable details to the explanations of an orbit and phase movement that originate in Parsons et al. (1953a), so incorporating Zelditch's work might be an important exercise to determine whether the simulation presented here would be extensible along the lines that Parsons and his colleagues might have taken it. Apply to more reported situations where there is a response to external energy

Only one application was made in the Results chapter to an instance reported in scholarly journals. Therefore, it would be instructive to move from the demonstration of a toy to a tool that explained reported structural and functions responses to external stimuli. Candidates might include (Audia, Locke, & Smith, 2000; Barr, 1998; Chattopadhyay, Glick, & Huber, 2001; Haveman, Russo, & Meyer, 2001; Hoffman, 1999; Holmwood, 1983; Marcus & Nichols, 1999).

In addition, a future approach might also focus on affect vs. affect-neutrality in decision making, relying on such sources as " Toxic decision processes: A study of emo-tion and organizational decision making" (Maitlis, 2004), The neurotic organization: Diagnosing and changing counterproductive styles of management (Kets de Vries, 1984), Unstable at the top: Inside the troubled organization (Kets de Vries, 1987), and The Icarus paradox: How exceptional companies bring about their own downfall; new lessons in the dynamics of corporate success, decline, and renewal (Miller, 1990). One aspect of the focus on affect vs. affect-neutrality that is missing in the current research is that of decision quality or organizational fitness: is there a better or worse Latent Pattern Main-tenance function with respect to a realistic goal to be optimized. The current research makes the single and naïve goal of matching internal energy to the pattern of external. Clearly there is much room here for improvement. Apply to agent-based systems

In the Research Methods section of the Methods chapter on the topic of selecting the appropriate simulation technology, the observation was made that agent-based simu-lation systems were gaining currency. In addition, there it was stated,

Simulating agents in Parsons’ theory of action might be a future application of the simulation described here. If one viewed agents as co-operating and communi-cating sequential processes (Hoare, 1985) (in the context of agent-based simula-tion), then this study gives insight into the program that might be inside each agent, that is, the instant research is a necessary precursor to an agent-based simulation of Parsons’ theory of action.

Now, therefore, to apply the instant research to agent-based simulation, one must construct a hierarchy or network into which agents fit. Much of this has been worked out

Page 95

for the theory of action in a different context (Fararo & Skvoretz, 1984), namely, a hierar-chy of interconnected automata that operate at different levels of interpenetrating abstraction, different units of analysis. One of the advantages of the approach described by Fararo and Skvoretz (1984) is that it preserves the non-determinism of agent-based systems and, again, it is entirely grounded in the theory of action. Address the dual of performance: organizational learning

The theory of action contains a duality of performance and learning. The simula-tion reported here deals only with performance and neglects learning. Therefore the simulation could be expanded to take into account learning. It is not clear how organiza-tions learn and particularly how Parsons thought they did. Therefore, further research could experiment with how each function makes sense of the energy presented to it and how it changes its internal processing correspondingly. Increase technical robustness

The human-computer interface could be improved. The current version is, to be charitable, unusable by anyone but the researcher. There is a significant literature written about how to construct effective interfaces between computers and humans. The ecologi-cal interface seems particularly applicable (Bennett & Flach, 1992; Chistoffersen, Hunter, & Vicente, 1998; Goldstein, 1969; Hoffman & Ocasio, 2001; Howie & Vicente, 1998a; Howie & Vicente, 1998b; Howie, Sy, Ford, & Vicente, 2000; Janzen & Vicente, 1998; Mitchell & Miller, 1986; Pawlak & Vicente, 1996; Rasmussen & Batstone, n.d.; Rasmussen, Duncan, & Leplat, 1987; Shneiderman, 1983; Vicente & Rasmussen, 1990; Vicente & Rasmussen, 1992; Weir, 1991; Woods, 1984; Woods, 1991).

In addition, the simulation could be rewritten with provable correctness in mind, so that testing and evaluation by a third party would be less important because the com-puter program could be proved correct, given its specification. There are several (award winning) methods for constructing and proving correct computer programs where the programs contain timing (Hoare, 1985; Manna & Pnueli, 1991; Manna & Pnueli, 1995).

In sum Alas, the real estimate of whether the model reported here will be sufficient for

further enrichment – which was the purpose of this research – can only be made by the next researcher in turn, who will evaluate this scaffold for its ability to continue the con-struction of a high fidelity replica of Parsons' theory of action.

Page 96

EPILOGUE As can be said of so many other doctoral candidates, this was not the dissertation I

set out to write. My first idea was to create a method of translating causal loop diagrams (CLDs) into system dynamics models (an example of which is shown in Figure 1, p. 4). Causal loop diagrams are informal drawings that show what are presumably causes and effects among circular influences. They were made popular in Senge (1990); Senge was a student of Jay Forrester, the "father" of system dynamics (Forrester, 1968). No one has ever been able to translate from CLDs to system dynamics models because there is (so much) information missing. I found some patterns that – when a few additional questions were asked and answered – would provide a first draft system dynamics model from CLDs. I was going to use some then-new results from qualitative physics, a branch of mathematics that does not rely on exact quantities, in order to reason about the relation-ships among variables.

In addition, I thought that causal loop diagrams might help with an endemic problem in system dynamics modeling: the misperception of feedback (Diehl & Sterman, 1995; Kleinmuntz, 1993; Paich & Sterman, 1993; Sterman, 1989a; Sterman, 1989b). It seems that our human cognition is not very good at seeing non-linear or cyclic or attenu-ated cause and effect connections. And this has been demonstrated even among people who construct such connections every day.

During tea at a George Washington University function I was chatting with Karl Weick about my work because I knew that he was interested (I had written a school paper in which I pointed out that I thought he was mistaken in Weick (1979, p. 69 ff) about cer-tain system dynamics applications). He asked me whether I thought I was solving a problem of ambiguity or of uncertainty. These are his shorthand terms for the two types of equivocality. Weick has written that the purpose of organizations is to reduce equivo-cality. Uncertainty is the want of information. Ambiguity is the want of sense(-making), there may be enough information and it may be contradictory.

I was stunned because I did not know the answer to that simple question. I pon-dered it a long time and spoke with system dynamics experts, including the author of Figure 1. I came away with no answer, so I abandoned that work, in which I had invested a significant portion of my research energy.

My next attempt was to see if I could apply some of the concepts of complex adaptive systems (CAS) – also called chaos or complexity theory – to some real organ-izational events. I had an idea that some of the arguments in the field – particularly about whether change is (a) rapid and cataclysmic (averred by those supporting punctuated equilibrium (Romanelli & Tushman, 1994) and quantum change (Miller, Friesen, & Mintzberg, 1984)), or (b) incremental (Donaldson, 1996) – are simply on a continuum of rate of change and that those changes could be more parsimoniously (and less polemi-cally) explained by a fact of (non-linear) differential equations, the staple of CAS.

The problem was that I could not figure out what to measure. I still cannot, and nor it seems can anyone who studies organizations from the CAS perspective. CAS appears to be a metaphor, not really yet a computational tool.37

37 "Despite the promise indicated by various authors within the field, complexity science has thus far failed to deliver

tangible tools that might be utilized in the examination of complex systems." (Richardson, Cilliers, & Lissack, 2001)

Page 97

One of the turning points in my search to apply what I knew as an engineer and physical scientist to social systems came when David Schwandt, the chair of the disserta-tion committee and the Director of the Executive Leadership Program, invited me to read Daft and Weick (1984), which relied on Boulding (1956). Basically, that work argues that social systems interpret the forces that impinge upon them, they do not "robotically" absorb and then reflect the energy that is aimed at them, as billiard balls would. That is, a social system could absorb energy, reflect energy, multiply energy, delay energy, con-sume energy, or do with energy whatever it wanted to, completely and totally unlike physical systems; physical systems conserve energy. That is, in physical systems there is a fixed amount of energy and for an object to gain more means a loss of some somewhere else, and vice versa. In social systems there is no conservation, no limit to the energy in the system. Nearly all of physics is based upon an equation, an equality, that connects energy to its other embodiments. What would the energy in a social system be equal to? What equality would be preserved/conserved across social acts? I could not and cannot answer those questions, so I dropped my search for physics-like thinking, especially complex adaptive systems (also called complexity theory), applied to social systems.

The current research flowed from my interest in Parsons theory of action because I apply it every day in the delivery of advisory services. I use the theory of action to evaluate the situation, diagnose the current state, and look for leverage for change. I wondered whether I could animate the theory, as so many have for other social systems before me.

I started to go to college by attending night school. It was the time of the military draft and students were deferred if they made normal progress. In trying to make normal progress I was forced to take courses that I could get into, whether I had the inclination or prerequisites or not. During an early semester I took a computer course and did badly. The next semester the only course for which I really had taken the prerequisite was the follow-on computer course. In that more advanced course (it was the most advanced offered at the university at the time) the instructor asked me to learn about a new thing, discrete event simulation (DES). I learned the principles (current in 1967) and wrote a computer program in the General Purpose System Simulation (GPSS) language, which was brand new at the time, that simulated a grocery store, in particular something that was first being tried in that era: designated lines for a small number of items. I was curi-ous about whether those lines worked or not.

The effect on waiting times notwithstanding, my GPSS program impressed quite a few people, so it ended up impressing me! And by that experience simulation became something of a lens through which I viewed a part of the world, particularly the world of management decision-making, which was to become my undergraduate focus in business administration.

In 1975 for my employer at the time I was trying to predict the growth of adoption of a new product. I already knew about the usual S-shaped growth pattern that one gets in a restricted medium, like a Petri dish, and it had been applied to the adoption of technol-ogy despite the obvious violation of assumptions. I was looking for something, well, more human.

Limits to growth (Meadows, Meadows, Randers, & Behrens, 1972) had been recently published and made fascinating reading. It was a simulation of how the world would grow in the next century. It was my first exposure to system dynamics and it was

Page 98

an impressive one. I wrote a simulation of product adoption based on what I learned from Limits to growth. And I tried to stay current with what was called the world model by the system dynamics community.

In the end I did not let my initial exposure to computing in that first university course influence my final direction, mostly because I did so much better in the second course, and that by learning and applying discrete event simulation. DES and system dynamics have different heritages and often appear as intellectual schools that fight over the same turf, much like any school of thought before it becomes "normal" (Kuhn, 1970). Having some facility in both and no commitment to either, this would not be the last time I would be spanning boundaries.

By the time I had earned my undergraduate degree in business I was very inter-ested in computers, so tried to pursue a graduate program in that field, ostensibly inside a graduate school of business. The business school I selected turned out to be having a bat-tle about the place of computing inside it and I could see myself as becoming a pawn in the conflict, so I sought another place at the same university where I could learn com-puting in a different setting. In the end I entered the school of engineering and applied science, for which I virtually completely lacked the prerequisites and did not understand most of the course titles! I had a lot of catching up to do.

I completed my masters work with a thesis that was widely regarded and earned me a visiting scientist position for a year at a distinguished physics institute. I was work-ing on my PhD dissertation there, a simulation system that would permit an arbitrary level of detail. One of the challenges in creating any simulation is that there are some things you care about and some you do not. In each simulation system what a researcher might select as the choices to care about and not to are already made. I wanted to permit the modeler an arbitrary level of concern. As part of my literature search I read about 300 engineering dissertations, nearly all of which had been one year of work and did not build on previous work, so none of them could attack the arbitrariness of the level of detail. I ran out of time, too, and never completed the research. And there has never been a simu-lation system that lets the researcher select an arbitrary level of detail/concern/ abstrac-tion.

Much later in my career I became a consultant to the parts of organizations in which software is developed. Gradually the level of my clients inside those organizations rose and the nature of their questions changed from technical to organizational: "You are advocating that we work in teams. How long does it take a team to do its work?" "What’s the best way for me to organize the 7500 people who work for me?"

This set of questions, and ones like it about how innovation is adopted, took me away from my technical background and placed me on weak ground. So I pursued the learning offered by the Executive Leadership Program’s doctoral degree. Again, I had none of the prerequisites and had to study very hard just to catch up and then stay in place.

I use every day what I learned and I believe that, despite my fits and starts on a dissertation topic, I am a poster-child for the Program, a Program that encourages bound-ary spanning by the example of its leader, Prof. Dave Schwandt, who is a recovering physicist.

Page 99

REFERENCES

(1997). Simulation and Gaming and the Teaching of Sociology. ASA Resources Materials for Teaching. American Sociological Association. 19 pages.

Abbott, A. (1988). Transcending general linear reality. Sociological Theory, 6(2), 169-186.

Abbott, A. (1992). From causes to events. Sociological Methods and Research, 20, 428-455.

Abbott, A. (2001). Time matters: On theory and method. Chicago: University of Chicago Press.

Achterkamp, M., & Imhof, P. (1999). The importance of being systematically surprise-able: Comparative social simulation as experimental technique. Journal of Mathe-matical Sociology, 23( 4), 327-347.

Alexander, J. C. (1983). The later period (1): The interchange model and Parsons' final approach to multidimensional theory. In J. C. Alexander, The modern reconstruction of classical thought: Talcott Parsons (Vol. Four, pp. 73-118). Berkeley, CA: University of California Press.

Alexander, J. C., & Sciortino, G. (1996). On choosing one's intellectual predecessors: The reductionism of Camic's treatment of Parsons and the Institutionalists. Sociological Theory, 14(2), 154-171.

Ancona, D. G., & Chong, C. (1996). Entrainment: Pace, cycle, and rhythm in organizational behavior. In L. L. Cummings, & B. M. Staw, (Eds.), Research in organizational behavior (Vol. 18,Chap. 251-284, ). Greenwich, CT: JAI Press.

Ashby, W. R. (1956). An introduction to cybernetics. New York, NY: John Wiley.

Audia, P. G., Locke, E. A., & Smith, K. G. (2000). The paradox of success: An archival and a laboratory study of strategic persistence following radical environmental change. Academy of Management Journal , 43(5), 837-853.

Axten, N., & Fararo, T. J. (1977). The information processing representation of institutionalized social action. Sociological Review Monograph, 24, 35-77.

Bainbridge, W. S. (1992). Social research methods and statistics: A computer-assisted introduction. Belmont, CA: Wadsworth.

Bales, R. F. (1950). Interaction process analysis. Cambridge: Addison-Wesley.

Bales, R. F. (1999). Social interaction systems: Theory and measurement. New Brunswick, NJ: Transaction Publishers.

Page 100

Balzer, W., Sneed, J. D., & Moulines, C. U. (2000). Structuralist knowledge representation: Paradigmatic examples. Amsterdam: Rodopi.

Banks, J., & Carson, J. S.II. (1984). Discrete-event system simulation. Englewood Cliffs, NJ: Prentice-Hall.

Barber, B., & Inkeles, A. (Eds.). (1971). Stability and change: A volume in honor of Talcott Parsons. Boston: Little, Brown and Co.

Barkema, H. G., Baum, J. A. C., & Mannix, E. A. (Eds.). (2002). A new time. [Special research forum]. Academy of Management Journal, 45(5).

Barr, P. S. (1998). Adapting to unfamiliar environmental events: A look at the evolution of interpretation and its role in strategic choice. Organization Science, 9(6), 644-669.

Baudrillard, J. (1995). Simulacra and simulation. Ann Arbor, MI: University of Michigan Press.

Bennett, K. B., & Flach, J. M. (1992). Graphical displays: Implications for divided attention, focused attention, and problem solving. Human Factors, 34(5), 513-534.

Berger, J., & Zelditch, M., Jr. (1968). Sociological theory and modern society. [Book review]. American Sociological Review, 33(3), 446-450.

Bergmann, W. (1992). The problem of time in sociology: An overview of the literature on the state of theory and research on 'Sociology of Time,' 1900-82. Time & Society, 1(1), 81-134.

Black, M. (Ed.). (1961). The social theories of Talcott Parsons. Englewood Cliffs, NJ: Prentice-Hall.

Bluth, B. J. (1982). Parsons' general theory of action: A summary of the basic theory. Granada Hills, CA: NBS.

Boudon, R., & Bourricaud, F. (1989). A critical dictionary of sociology. London: Routledge.

Boulding, K. E. (1956). General systems theory: The skeleton of a science. Management Science, 2, 197-207.

Bourricaud, F. (1981). The sociology of Talcott Parsons. Chicago, IL: University of Chicago Press.

Bressler, M. (1961). Supplement: Some selected aspects of American sociology, September 1959 to December 1960. Annals of the American Academy of Political and Social Science, 337, 146-159.

Brodbeck, M. (1959). Models, meaning, and theories. In L. Gross, (Ed.), Symposium on

Page 101

sociological theory (pp. 373-403). New York: Harper & Row.

Bronson, R., & Jacobsen, C. (1986). Simulation and social theory. Simulation, 47(2 ), 58-62.

Bronson, R., Jacobsen, C., & Crawford, J. (1988). Estimating functional relationships in a macrosociological model. Mathematical Computer Modelling, 11, 386-390.

Brownstein, L. (1982). Talcott Parsons' general theory of action: An investigation of fundamental principles. Cambridge, MA: Schenkman Publishing Co.

Burrell, G., & Morgan, G. (1979). Sociological paradigms and organisational analysis. Portsmouth, NH: Heinemann.

Burton, R. M., & Obel, B. (Eds.). (1995). Design models for hierarchical organizations: Computation, information, and decentralization. Boston, MA: Kluwer Academic Publishers.

Cadwallader, M. L. (1959). The cybernetic analysis of change in complex social organizations. American Journal of Sociology, 65(2), 154-157.

Camic, C. (1996). Alexanders' antisociology. Sociological Theory, 14(2), 172-186.

Camic, C. (1998). Reconstructing the theory of action. Sociological Theory, 16(3), 283-291.

Campbell, D. T., & Stanley, J. C. (1963). Experimental and quasi-experimental designs for research. Boston: Houghton Mifflin.

Carley, K. M., & Prietula, M. J. (Eds.). (1994). Computational organization theory. Hillsdale, NJ: Lawrence Erlbaum Associates.

Chattopadhyay, P., Glick, W. H., & Huber, G. R. (2001). Organizational actions in response to threats and opportunities. Academy of Management Journal, 44(5), 937-955.

Checkland, P. (1999). Systems thinking, systems practice (30-year retrospective ed.). West Sussex, England: John Wiley & Sons.

Checkland, P., & Scholes, J. (1999). Soft systems methodology in action (30-year retrospective ed.). West Sussex, England: John Wiley & Sons.

Cherns, A. (1980). Work and values: Shifting patterns in industrial society. International Social Science Journal, 32(3 ), 427-441.

Chistoffersen, K., Hunter, C. N., & Vicente, K. J. (1998). A longitudinal study of the effects of ecological interface design on deep knowledge. International Journal of Human-Computer Studies, 48(6), 729-762.

Page 102

Coleman, J. S. (1964). Introduction to mathematical sociology. New York, NY: Free Press of Glencoe.

Coleman, J. S. (1965). The use of electronic computers in the study of social organizations. Archives Européennes De Sociologie, VI(I), 89 ff.

Collins, L. M., & Sayer, A. G. (Eds.). (2001). New methods for the analysis of change. Washington DC: American Psychological Association.

Conte, R., Hegselmann, R., & Terno, P. (Eds.). ( 1997). Simulating social phenomena. Heidelberg, Germany: Springer.

Conway, R. W., & McClain, J. O. (2003). The conduct of an effective simulation study. INFORMS Transactions on Education, 3(3).

Coyle, G. (2000). Qualitative and quantitative modelling in system dynamics: Some research questions. System Dynamics Review, 16(3), 225-244.

Cross, W. M. (1980). The use of situation-generated simulation games in the teaching of sociology. For presentation at Annual Meeting of the Illinois Sociological Society .

Cubitt, S. (2001). Simulation and social theory. London: Sage.

Cyert, R. M., & March, J. G. (1963). A behavioral theory of the firm. Englewood Cliffs, NJ: Prentice-Hall.

Cyert, R. M., & March, J. G. (1992). A behavioral theory of the firm (2nd ed.). Cambridge, MA: Blackwell.

Daft, R. L., & Weick, K. E. (1984). Toward a model of organizations as interpretation systems. Academy of Management Review, 9(2), 284-295.

Dar-El, E. M. (2000). Human learning: from learning curves to learning organizations. Boston, MA: Kluwer Academic Publishers.

Davis, K. (1959). The myth of functional analysis as a special method in sociology and anthropology. American Sociological Review, 24( 6), 757-772.

Diehl, E., & Sterman, J. D. (1995). Effects of feedback complexity on dynamic decision making. Organizational Behavior & Human Decision Processes, 62(2), 198-215.

Donaldson, L. (1996). For positivist organization theory: Proving the hard core. London: Sage.

Dubin, R. (1960). Parsons' actor: Continuities in social theory. American Sociological Review, 25(4), 457-466.

Dukes, R. L. (1975). An evaluation of six prominent simulation games for teaching undergraduate sociology. Presented at 50th Annual Meeting of the Southwestern

Page 103

Sociological Association .

Durkheim, E. (1951). Suicide, a study in sociology. Glencoe, IL: Free Press.

Ebbinghaus, H. (1987). Memory: A contribution to experimental psychology. New York: Dover Publications.

Epstein, J. M., & Axtell, R. (1996). Growing artificial societies: Social science from the bottom up. Washington, DC: Brookings Institution Press.

Etzioni, A. (1975). A comparative analysis of complex organizations. New York, NY: Free Press.

Fararo, T. J. (1984). Neoclassical theorizing and formalization in sociology. In T. J. Fararo, (Ed.), Mathematical ideas and sociological theory (pp. 143-175). New York: Gordon and Breach.

Fararo, T. J. (1989). The meaning of general theoretical sociology: Tradition and formalization. Cambridge, England: Cambridge University Press.

Fararo, T. J. (2001). Social action systems: Foundation and synthesis in sociological theory. Westport, CT: Praeger.

Fararo, T. J., & Hummon, N. P. (1994). Discrete event simulation and theoretical models in sociology. In B. Markovsky, K. Heimer, & J. O'Brien, (Eds.), Advances in group processes (Vol. 11,pp. 25-66). Greenwich, CT: JAI Press.

Fararo, T. J., & Skvoretz, J. (1984). Institutions as production systems. Journal of Mathematical Sociology, 10, 117-182.

Fishman, G. S. (2001). Discrete-event simulation: Modeling, programming, and analysis. New York: Springer-Verlag.

Forrester, J. (1968). Principles of systems. Cambridge, MA: MIT Press.

Garson, G. D. (1994). Social science computer simulation: Its history, design, and future. Social Science Computer Review, 12(1), 55-82.

Gilbert, N., & Conte, R. (Eds.). (1995). Artificial societies: The computer simulation of social life. London: UCL Press.

Goldstein, M. (1969). Display aspects of algebra. Psychological Reports, 24, 937-938.

Gullahorn, J. T., & Gullahorn, J. E. (1963). A computer model of elementary social behavior. Computers in Behavioral Science.

Habermas, J. (1981). Talcott Parsons: Problems of theory construction. Sociological Inquiry, 51(3/4), 173-196.

Page 104

Hamagami, F., & McArdle, J. J. (2001). Advanced studies of individual differences: Linear dynamic models for longitudinal data analysis. In G. A. Marcoulides, & R. E. Schumacker, (Eds.), New developments and techniques in structural equation modeling (pp. 203-246). Malwah, NJ: Lawrence Erlbaum Associates.

Hamblin, R. L., Jacobsen, R. B., & Miller, J. L. L. (1973). A mathematical theory of social changes. New York, NY: Wiley-Interscience.

Hanneman, R., & Patrick, S. (1997). On the uses of computer-assisted simulation modeling in the social sciences. Sociological Research Online, 2(2).

Hanneman, R. A. (1988). Computer-assisted theory building: Modeling dynamic social systems. Newbury Park, CA: Sage.

Harré, R., & Secord, P. F. (1972). The explanation of social behaviour. Oxford, England: Basil Blackwell.

Hassard, J. (1990). Introduction: The sociological study of time. In J. Hassard, (Ed.), The sociology of time (pp. 1-18). New York: St. Martin's Press.

Haveman, H. A., Russo, M. V., & Meyer, A. D. (2001). Organizational environments in flux: The impact of regulatory punctuations on organizational domains, CEO succession, and performance. Organization Science, 12(3), 253-273.

Hayes, A. C. (1981). Structure and creativity: The use of transformational-generative models in action theory. Sociological Inquiry, 51( 3-4), 219-239.

Heise, D. R. (1979). Understanding events: Affect and the construction of social actions. Cambridge, England: Cambridge University Press.

Hills, R. J. (1968). Towards a science of organization. Eugene, OR: Center for the Advanced Study of Educational Administration.

Hoare, C. A. R. (1985). Communicating sequential processes. Englewood Cliffs, NY: Prentice-Hall International.

Hoffman, A. J. (1999). Institutional evolution and change: Environmentalism and the U.S. chemical industry. Academy of Management Journal, 42(4), 351-371.

Hoffman, A. J., & Ocasio, W. (2001). Not all events are attended equally: Toward a middle-range theory of industry attention to external events. Organization Science, 12(4), 414-434.

Holmwood, J. (1996). Founding sociology? Talcott Parsons and the idea of general theory. New York: Longman.

Holmwood, J. M. (1983). Action, system and norm in the action frame of reference: Talcott Parsons and his critics. Sociological Review, New Series, 31, 310-336.

Page 105

Honderich, T. (Ed.). (1995). Oxford companion to philosophy. New York, NY: Oxford University Press.

Howie, D. E., & Vicente, K. J. (1998a). Making the most of ecological interface design: The role of self-explanation. International Journal of Human-Computer Studies, 49(5), 651-674.

Howie, D. E., & Vicente, K. J. (1998b). Measures of operator performance in complex, dynamic microworlds: Advancing the state of the art. Ergonomics, 41(4), 485-500.

Howie, E., Sy, S., Ford, L., & Vicente, K. J. (2000). Human-computer interface design can reduce misperceptions of feedback. System Dynamics Review, 16(3), 151-171.

Ilgen, D. R., & Hulin, C. L. (Eds.). (2000). Computational modeling of behavior in organizations: The third scientific discipline. Washington, DC: American Psychological Association.

Jaber, M. Y., & Sikström, S. (2004). A note on "An empirical comparison of forgetting models". IEEE Transactions on Engineering Management, 51(2), 233-234.

Jackson, M. A. (1983). System development. Englewood Cliffs, NJ: Prentice-Hall.

Jacobsen, C., & Bronson, R. (1995). Computer simulation and empirical testing of sociological theory. Sociological Methods & Research, 23(4), 479-506.

Jacobsen, C., & Bronson, R. (1985). Simulating violators. Operations Research Society of America [now Institute for Operations Research and Management Science].

Jacobsen, C., & Bronson, R. (1987). Defining sociological concepts as variables for system dynamics modeling. System Dynamics Review, 3(1), 1-7.

Jacobsen, C., & Bronson, R. (1997). Computer simulated empirical tests of social theory: Lessons from 15 years' experience. In R. Conte, R. Hegselmann, & P. Terno (Eds.), Simulating social phenomena (pp. 97-102). Heidelberg, Germany: Springer.

Jacobsen, C., Bronson, R., & Vekstein, D. (1990). A strategy for testing the empirical adequacy of macro-sociological theories. Journal of Mathematical Sociology, 15 (2), 137-148.

Janzen, M. E., & Vicente, K. J. (1998). Attention allocation within the abstraction hierarchy. International Journal of Human-Computer Studies, 48(4), 521-545.

Jin, Y., & Levitt, R. (1996). The virtual design team: A computational model of project organizations. Computational & Mathematical Organization Theory, 2(3), 171-196.

Kant, I. (1896). Critique of pure reason. New York, NY: Macmillan.

Keat, R., & Urry, J. (1982). Social theory as science. London: Routledge & Kegan Paul.

Page 106

Kets de Vries, M. F. R. (1984). The neurotic organization: Diagnosing and changing counterproductive styles of management. San Francisco, CA: Jossey-Bass.

Kets de Vries, M. F. R. (1987). Unstable at the top: Inside the troubled organization. New York, NY: New American Library.

Kleinmuntz, D. (1993). Information processing and misperceptions of the implications of feedback in dynamic decision making. System Dynamics Review, 9(3), 223-237.

Kleinrock, L. (1975-1976). Queueing systems (Vols. 1-2). New York: Wiley.

Kolb, W. L. (1962). The social theories of Talcott Parsons: A critical examination. American Journal of Sociology, 67(5), 590-591.

Kuhn, T. S. (1970). The structure of scientific revolutions. Chicago, IL: University of Chicago Press.

Lackey, P. N. (1987). Invitation to Talcott Parsons' theory. Houston: Cap and Gown Press.

Lane, D. (2001). Rerum cognoscere causas: Part II -- Opportunities generated by the agency/structure debate and suggestions for clarifying the social theoretic position of system dynamics. System Dynamics Review, 17(4), 293-309.

Lave, C. A., & March, J. G. (1993). An introduction to models in the social sciences. Lanham, MD: University Press of America.

Leik, R. K., & Meeker, B. F. (1995). Computer simulation for exploring theories: Models of interpersonal cooperation and competition. Sociological Perspectives, 38(4), 463-482.

Lin, Z. (2000). Organizational performance under critical situations -- exploring the role of computer modeling in crisis case analysis. Computational & Mathematical Organization Theory, 6(3), 277-310.

Lomi, A., & Larsen, E. R. (1995). The emergence of organizational structure. In R. M. Burton, & B. Obel, (Eds.), Design models for hierarchical organizations: Computation, information, and decentralization ( pp. 209-231). Boston: Kluwer.

Loubser, J. J., Baum, R. C., Effrat, A., & Lidz, V. M. (Eds.). (1976a). Explorations in general theory in social science: Essays in honor of Talcott Parsons Vol. I. New York: Free Press.

Loubser, J. J., Baum, R. C., Effrat, A., & Lidz, V. M. (Eds.). (1976b). Explorations in general theory in social science: Essays in honor of Talcott Parsons Vol. II. New York: Free Press.

Luna-Reyes, L. F., & Andersen, D. L. (2003). Collecting and analyzing qualitative data

Page 107

for system dynamics: Methods and models. System Dynamics Review, 19(4), 271-296.

Lüscher, K. (1974). Time: A much-neglected dimension in social theory and research. Sociological Analysis and Theory, 4, 101-117.

Maitlis, S. O. H. (2004). Toxic decision processes: A study of emotion and organizational decision making. Organization Science, 15(4), 375-393.

Manna, Z., & Pnueli, A. (1991). The temporal logic of reactive and concurrent systems. New York: Springer.

Manna, Z., & Pnueli, A. (1995). Temporal verification of reactive systems: Safety. New York: Springer.

Marcus, A. A., & Nichols, M. L. (1999). On the edge: Heeding the warnings of unusual events. Organization Science, 10(4), 482-499.

Markley, O. W. (1967). A simulation of the SIVA model of organizational behavior. American Journal of Sociology, 73(3), 339-347.

Mazur, J. E., & Hastie, R. (1978). Learning as accumulation: Reexamination of the learning curve. Psychological Bulletin, 85(6), 1256-1274.

McCleary, R., & Hay, R. A., Jr. (1980). Applied time series analysis for the social sciences. Beverly Hills, CA: Sage.

McGowan, J. (1998). Towards a pragmatic theory of action. Sociological Theory, 16 (3), 292-297.

Meadows, D. H., Meadows, D. L., Randers, J., & Behrens, W. W. (1972). The limits to growth. New York, NY: Universe Books.

Miller, D. (1990). The Icarus paradox: How exceptional companies bring about their own downfall; new lessons in the dynamics of corporate success, decline, and renewal. New York, NY: Harper Business.

Miller, D., Friesen, P. H., & Mintzberg, H. (1984). Organizations: A quantum view. Englewood Cliffs, NJ: Prentice-Hall.

Mitchell, C. M., & Miller, R. A. (1986). A discrete control model of operator function: A methodology for information display design. IEEE Transactions on Systems, Man, and Cybernetics, SMC-16(3), 343-357.

Mize, J. H., & Cox, J. G. (1968). Essentials of simulation. Englewood Cliffs, NJ: Prentice-Hall.

Moore, W. E. (1959). The whole state of sociology. [Book reviews of Sociology today:

Page 108

Problems and prospects and Symposium on sociological theory]. American Sociological Review, 24(5), 715-718.

Morrison, M., & Morgan, M. S. (1999). Models as mediating instruments. In M. S. Morgan, & M. Morrison, (Eds.), Models as mediators: Perspectives on natural and social science (pp. 10-37). Cambridge, England: Cambridge University Press.

Moss, S. (2000). Canonical tasks, environments and models for social simulation. Computational & Mathematical Organization Theory, 6(3), 249-275.

Mueller, R. O. (1996). Basic principles of structural equation modeling: An introduction to LISREL and EQS . New York: Springer.

Nembhard, D. A., & Osothsilp, N. (2001). An empirical comparison of forgetting models. IEEE Transactions on Engineering Management, 48(3), 283-291.

Nembhard, D. A., & Osothsilp, N. (2004). Authors' reply to "A note on 'An empirical comparison of forgetting models'". IEEE Transactions on Engineering Management, 51(2), 235.

Nembhard, D. A., & Uzumeri, M. V. (2000). An individual-based description of learning within an organization. IEEE Transactions on Engineering Management, 47(3 ), 370-378.

Nowakowska, M. (1973). A formal theory of actions. Behavioral Science, 18, 393-416.

Paich, M., & Sterman, J. D. (1993). Boom, bust, and failures to learn in experimental markets. Management Science, 39(12), 1439-1458.

Park, P. (1967). Measurement of the pattern variables. Sociometry, 30(2), 187-198.

Parsons, T. (1951). The social system. Chicago, IL: Free Press.

Parsons, T. (1954). Essays in sociological theory (Revised ed.). New York: NY: Free Press.

Parsons, T. (1960). Pattern variables revisited: A response to Robert Dubin. American Sociological Review, 25(4), 467-483.

Parsons, T. (1961a). The general interpretation of action: Editorial forward. In T. Parsons, E. Shils, K. D. Naegele, & J. R. Pitts, (Eds.), Theories of society: Foundations of modern sociological theory (pp. 85-97). New York: Free Press.

Parsons, T. (1961b). An outline of the social system. In T. Parsons, E. Shils, K. D. Naegele, & J. R. Pitts, (Eds.), Theories of society: Foundations of modern sociological theory (pp. 30-79). New York: Free Press.

Parsons, T. (1961c). The point of view of the author. In M. Black (Ed.), The social

Page 109

theories of Talcott Parsons (pp. 311-363). Englewood Cliffs: NJ: Prentice-Hall.

Parsons, T. (1968a). The structure of social action: A study in social theory with special reference to a group of recent European Writers (With a new introduction ed.). Vol. I. New York: NY: Free Press.

Parsons, T. (1968b). The structure of social action: A study in social theory with special reference to a group of recent European Writers (With a new introduction ed.). Vol. II. New York: NY: Free Press.

Parsons, T. (1970). The impact of technology on culture and emerging new modes of behaviour. International Social Science Journal, XXII(4), 607-627.

Parsons, T. (1974). The institutional function in organization theory. Organization and Administrative Sciences, 5(1), 3-12.

Parsons, T. (1977a). On building social systems theory: A personal history. In T. Parsons, Social systems and the evolution of action theory (pp. 22-76). New York, NY: Free Press.

Parsons, T. (1977b). Social systems and the evolution of action theory. New York: NY: Free Press.

Parsons, T. (1977c). Some problems of general theory in sociology. In T. Parsons, Social systems and the evolution of action theory (pp. 229-269). New York, NY: Free Press.

Parsons, T. (1978a). Epilogue. In The doctor-patient relationship in the changing health scene (pp. 445-455). Washington, D.C.: U.S. Department of Health, Education, and Welfare.

Parsons, T. (1978b). Comment on R. Stephen Warner's "Toward a redefinition of action theory: Paying the cognitive element its due" . American Journal of Sociology, 83(6), 1350-1358.

Parsons, T. (1982). The pattern variables. In L. Mayhew, (Ed.), Talcott Parsons on institutions and social evolution: Selected writings (pp. 106-114). Chicago: University of Chicago Press.

Parsons, T., & Bales, R. F. (1953). The dimensions of action-space. In T. Parsons, R. Bales, & E. A. Shils, Working papers in the theory of action (Chap. 3, pp. 63-109). Glencoe, IL: Free Press.

Parsons, T., Bales, R. F., & Shils, E. A. (1953a). Phase movement in relation to motivation, symbol formation, and role structure. In T. Parsons, R. Bales, & E. A. Shils, Working papers in the theory of action (Chap. 5, pp. 163-269). Glencoe, IL: Free Press.

Parsons, T., Bales, R. F., & Shils, E. A. (Eds.). (1953b). Working papers in the theory of

Page 110

action. Glencoe, IL: Free Press.

Parsons, T., & Platt, G. M. (1973). The American university. Cambridge, MA: Harvard University Press.

Parsons, T., & Shils, E. A. (1951). Toward a general theory of action. Cambridge, MA: Harvard University Press.

Parsons, T., & Smelser, N. J. (1956). Economy and society: A study in the integration of economic and social theory. Glencoe, IL: Free Press.

Pawlak, W. S., & Vicente, K. J. (1996). Inducing effective operator control through ecological interface design. International Journal of Human-Computer Studies, 44(5), 653-688.

Pfahl, D., Laitenberger, O., Dorsch, J., & Ruhe, G. (2003). An externally replicated experiment for evaluating the learning effectiveness of using simulations in software project management education. Empirical Software Engineering, 8(4), 367-395.

Phelan, S. E. (1995). Using Simulation for Theory Generation in Strategic Management. 2nd Australasian Conf. in Strategic Management. La Trobe University, Melbourne, Australia: 6 pages.

Podell, L. (1966). Sex and role conflict. Journal of Marriage and the Family, 28(2), 163-1165.

Podell, L. (1967). Occupational and familial role-expectations. Journal of Marriage and the Family, 29(3), 492-493.

Prietula, M. J., Carley, K. M., & Gasser, L. (1998). A computational approach to organizations and organizing. In M. J. Prietula, K. M. Carley, & L. Gasser, (Eds.), Simulating organizations: Computational models of institutions and groups (p. xiii-xix). Menlo Park, CA: AAAI Press/MIT Press.

Rasmussen, J. (1985). The role of hierarchical knowledge representation in decision making and system management. IEEE Transactions on Systems, Man, and Cybernetics, SMC-15(2), 234-243.

Rasmussen, J., & Batstone, R. (n.d.). Why do complex organizational systems fail? World Bank.

Rasmussen, J., Duncan, K., & Leplat, J. (Eds.). (1987). New technology and human error. Chichester, England: John Wiley & Sons.

Richardson, K. A., Cilliers, P., & Lissack, M. (2001). Complexity science: A "gray" science for the "stuff in between". Emergence: A Journal of Complexity Issues in Organizations and Management, 3(2), 6-18.

Page 111

Riley, M. W., & Nelson, E. E. (1971). Research on stability and change in social systems. In B. Barber, & A. Inkeles, (Eds), Stability and social change (pp. 407-449). Boston, MA: Little, Brown and Co.

Robinson, S. (2001). Soft with a hard centre: Discrete-event simulation in facilitation. Journal of the Operational Research Society, 52, 905-915.

Rocher, G. (1975). Talcott Parsons and American sociology. New York: Barnes & Noble.

Romanelli, E., & Tushman, M. L. (1994). Organizational transformation as punctuated equilibrium: An empirical test. Academy of Management Journal, 37(5), 1141-1166.

Rowell, L. (1989). Foreword. In Time and process: Interdisciplinary issues (The Study of Time VII) (p. vii-ix). Madison, CT: International Universities Press.

Samuelson, D. A. (2000). Designing organizations: CMOT launches success on a solid scientific foundation (What is CMOT and why is it taking off?). ORMS Today, 27(6), 24-27.

Sastry, M. A. (1997). Problems and paradoxes in a model of punctuated organizational change. Administrative Science Quarterly, 42(2), 237-275.

Savage, S. P. (1981). The theories of Talcott Parsons: The social relations of action. New York, NY: St. Martin's Press.

Selznick, P. (1961). The social theories of Talcott Parsons. American Sociological Review, 26(6), 932-935.

Senge, P. M. (1990). The fifth discipline: The art & practice of the learning organization. New York, NY: Doubleday.

Shackle, G. L. S. (1969). Decision order and time in human affairs (2nd ed.). Cambridge, England: Cambridge University Press.

Shneiderman, B. (1983). Direct manipulation: A step beyond programming languages. IEEE Computer, 57-69.

Sibeon, R. (1999). Anti-reductionist sociology. Sociology: Journal of the British Sociological Association, 33(2), 317.

Skvoretz, J., & Fararo, T. J. (1980). Languages and grammars of action and interaction: A contribution to the formal theory of action. Behavioral Science, 25(1), 9-22.

Skvoretz, J., & Fararo, T. J. (1996). Generating symbolic interaction: Production system models. Sociological Methods & Research, 25(1), 60-102.

Skvoretz, J., Fararo, T. J., & Axten, N. (1980). Role-programme models and the analysis of institutional structure. Sociology, 14(1), 49-67.

Page 112

Stegmüller, W. (1979). The structuralist view of theories. Berlin: Springer-Verlag.

Sterman, J. D. (1989a). Misperceptions of feedback in dynamic decision making. Organizational Behavior and Human Decision Processes, 43( 3), 301-335.

Sterman, J. D. (1989b). Modeling managerial behavior: misperceptions of feedback in dynamic decision making experiment. Management Science, 35(3), 321-339.

Sterman, J. D. (2000). Business dynamics: Systems thinking and modeling for a complex world. Boston: Irwin McGraw-Hill.

Thomsen, J., Levitt, R. E., & Kunz, J. (1999). A trajectory for validating computational emulation models of organizations. Computational & Mathematical Organization Theory, 5(4), 385-401.

Thorngate, W. (1976). "In general" vs. "It depends": some comments on the Gergen-Schlenker debate. Personality & Social Psychology Bulletin, 2, 404-410.

Troitzsch, K. G. (1998). Multilevel process modeling in the social sciences: Mathematical analysis and computer simulation. In W. B. G. Liebrand, A. Nowak, & R. Hegselmann, (Eds.), Computer modeling of social processes (pp. 20-36). London: Sage.

Tsuchiya, S. (1966). A new role for computerized simulation in social science: Summary thoughts on a case study. Simulation & Gaming, 27( 1), 103-109.

Tuma, N. B., & Hannan, M. T. (1984). Social dynamics: Models and methods. Orlando, FL: Academic Press.

Turner, B. S. (Ed.). (1999). The Talcott Parsons reader. Malden, MA: Blackwell.

Tushman, M. L., & Romanelli, E. (1985). Organizational evolution: A metamorphosis model of convergence and reorientation. Research in Organizational Behavior, 7, 171-222.

Udy, S. H., Jr. (1960). Structure and process in modern societies. [Book review]. American Journal of Sociology, 66(1), 96.

Uzmeri, M., & Nembhard, D. (1998). A population of learners: A new way to measure organizational learning. Journal of Operations Management, 16(5), 515-528.

Vaill, P. B. (1996). Learning as a way of being: Strategies for survival in a world of permanent white water. San Francisco: Jossey-Bass.

van Fraassen, B. C. (2002). The empirical stance. New Haven, CT: Yale University Press .

Vicente, K. J., & Rasmussen, J. (1990). The ecology of human-machine systems II:

Page 113

Mediating 'direct perception' in complex work domains. Ecological Psychology, 2(3), 207-249.

Vicente, K. J., & Rasmussen, J. (1992). Ecological interface design: Theoretical foundations. IEEE Transactions on Systems, Man, and Cybernetics, 22(4), 589-606.

Waller, M. J. (1999). The timing of adaptive group responses to nonroutine events. Academy of Management Journal, 42(2), 127-137.

Weick, K. (1979). The social psychology of organizing (2nd ed.). New York: McGraw-Hill.

Weir, G. R. S. (1991). Living with complex interactive systems. In G. R. S. Weir, & J. L. Alty, (Eds.), Human-computer interaction and complex systems (Chap. 1, pp. 1-21). London: Academic Press.

Whitehead, A. N. (1927). Science and the modern world: Lowell lectures, 1925 . New York, NY: Macmillan.

Williams, R. M., Jr. (1959). Friendship and social values in a suburban community: An exploratory study. Pacific Sociological Review, 2(1), 3-10.

Wixted, J. T., & Ebbesen, E. B. (1997). Genuine power curves in forgetting: A quantitative analysis of individual subject forgetting functions. Memory & Cognition, 25(1), 731-739.

Woods, D. D. (1984). Visual momentum: A concept to improve the cognitive coupling of person and computer. International Journal of Man-Machine Studies, 21, 229-244.

Woods, D. D. (1991). The cognitive engineering of problem representations. In G. R. S. Weir, & J. L. Alty, (Eds.), Human-computer interaction and complex systems (Chap. Ch. 7, pp. 169-188). London: Academic Press.

Zeigler, B. P., Praehofer, H., & Kim, T. G. (2000). Theory of modeling and simulation: Integrating discrete event and continuous complex dynamic systems (2nd ed.). San Diego, CA: Academic Press.

Zelditch, M., Jr. (1955). A note on the analysis of equilibrium systems. In T. Parsons, & R. F. Bales, Family, socialization and interaction process (pp. 401-408). New York: Free Press.

Page 114

APPENDIX – ATTESTATION OF AN EXPERT

Page 115

APPENDIX – SIMULATION PROGRAM LISTING

This section contains the complete simulation model in the language of SIMUL8, a product from Simul8 Corporation, 26th Floor, 225 Franklin Street, Boston, MA 02110; telephone: 800 547 6024.. More information is available at http://www.simul8.com The model shown here was written and executed in Release 10 Standard.

The most important part of the listing is the last one, Learning Model Common. It enacted the learning portion of Latent Pattern Maintenance and adjusted the Adaptation filter in order to reduce tension. It was invoked on each exit from the Adaptation function.

There is a narrated illustration of the model in action at http://www.Master-Systems.com/Parsons.ivnu

SIMUL8 Documentation for: Game 1.0.S8 at time 10/9/2004 10:15:10 PM Version: 10.0.0.678 ----------------------------------------------------------------------- Parsons' Game A "game" where an organization tries to match its internal energy to the external, whose future pattern is unknown. It performs the match my imitating Parsons' theory of action, and in particular the Latent Pattern Maintenance function that adjusts internal energy to patterns in the past external energy. Created by: Stan Rifkin Last opened by: Stan Rifkin *********************************************************************** General Simulation Information ------------------------------ Warm Up Time: 0 Results Collection Time: 2400 (Days) Start of day: 540 Length of day: 480 , Days per week: 5 Current Random Stream Set: 1 Data display when simulation stopped: Work Item Count *********************************************************************** Distributions SetEnergy External Column of Data Cell R6C2 GameInput.xls SetAffect External Column of Data Cell R6C1 GameInput.XLS Energy Label Based :Energy ADwell Label Based :ADwell GDwell

Page 116

Label Based :GDwell IDwell Label Based :IDwell LDwell Label Based :LDwell SelectIDwell SetDUE Labels Energy (Number) Affect (Number) Label (Number) ADwell (Number) GDwell (Number) IDwell (Number) LDwell (Number) DUE (Number) AlwaysOne (Number) WAIT TIME (Number) WORK TIME (Number) PRIORITY (Number) Images Default Image Entry Width: 32 Height: 32 Transparent Color: 16777215 Default Image Storage Bin Width: 32 Height: 32 Transparent Color: 16777215 Default Image Work Center Width: 32 Height: 32 Transparent Color: 16777215 Default Image Exit Width: 32 Height: 32 Transparent Color: 16777215 Default Image Resource Width: 32 Height: 32 Transparent Color: 16777215 Default Image Conveyor Width: 32 Height: 32 Transparent Color: 16777215 Default Image Tank Width: 32 Height: 32 Transparent Color: 16777215 Default Image Rotz

Page 117

Width: 32 Height: 32 Transparent Color: 16777215 Default Image Process Width: 32 Height: 32 Transparent Color: 16777215 Default Image Loader Width: 32 Height: 32 Transparent Color: 16777215 Default Image Vehicle Width: 32 Height: 32 Transparent Color: 16777215 Default Image Component Width: 32 Height: 32 Transparent Color: 16777215 Default Image 3D Light Width: 32 Height: 32 Transparent Color: 16777215 Default Image 3D Object Width: 32 Height: 32 Transparent Color: 16777215 Bolt_m Width: 16 Height: 13 Transparent Color: 16777215 Image 2 Width: 32 Height: 32 Transparent Color: 16777215 Image 3 Width: 31 Height: 32 Transparent Color: 16777215 SIMUL8 Windows and sub-windows ------------------------------ Open Icon Location X:640 Y:497 W:32 H:32 Window Location X:4 Y:124 W:1255 H:823 Color 16777215 Work Item Types --------------- Main Work Item Type Image: Bolt_m Length 1 Attached Labels: Energy Affect Label ADwell GDwell IDwell LDwell DUE AlwaysOne WAIT TIME WORK TIME

Page 118

PRIORITY *********************************************************************** Simulation Objects ------------------ Work Entry External World -------------- Display Parameters 4 X:205 Y:298 W:32 H:32 Xinc -10 Yinc 0 Image 0 Default Image Conveyor Show Image Do not collect results Work Item Type: Main Work Item Type Inter-arrival time Distribution Detail: Fixed 5 0 0 0 Route Out Objects AFilter On Label Action Visual Logic: VL SECTION: Set dwell time properties SET PRIORITY = Affect 'If Affect = 1 then Affect is required. IF Affect = 1 SET ADwell = Table[10,6] SET GDwell = Table[10,9] SET IDwell = Table[10,12] SET LDwell = Table[10,15] ELSE SET ADwell = Table[10,5] SET GDwell = Table[10,8] SET IDwell = Table[10,11] SET LDwell = Table[10,14] SET DUE = IDwell Label Actions PRIORITY None AlwaysOne Set Distribution Detail: Fixed 1 0 0 0 Affect Set Distribution Detail: Uses: SetAffect External (Excel) Energy Set Distribution Detail: Uses: SetEnergy External (Excel) GDwell None

Page 119

ADwell None LDwell None IDwell None DUE None Work Center AFilter ------- Display Parameters 4 X:345 Y:225 W:32 H:32 Xinc -10 Yinc 0 Show Title Show Count Show Image Replicate 1 Do not collect results Priority 50 Routing In Priority Route In Objects External World Require resources before collecting any work items Routing Out Label On label: Label Preference only Route Out Objects Energy that does not enter Waiting for Adaptation Affect processing Release resources as soon as task complete Operation Time Distribution Detail: Fixed 0 0 0 0 On Label Action Visual Logic: VL SECTION: AFilter Action Logic CALL Learning Model Common 'GT is the greater than relation (>) and GTE is greater than or equal to (>=). 'Temp contains the new threshold, computed from the learning function. IF Relation = GT IF Energy > Temp IF Affect = 0 SET Label = 2 'Path 2 = (Normal, affect neutral) Adaptation ELSE SET Label = 3 'Path 3 = Adaptation with Affect ELSE SET Label = 1 'Path 1 = exit the organization ELSE IF Relation = GTE

Page 120

IF Energy >= Temp IF Affect = 0 SET Label = 2 'Path 2 = (affect neutral) Adaptation ELSE SET Label = 3 'Path 3 = Adaptation with Affect ELSE SET Label = 1 'Path 1 = exit the organization ELSE SET Label = 1 Label Actions Label None Work Center Adaptation ---------- Display Parameters 4 X:444 Y:292 W:32 H:32 Xinc -10 Yinc 0 Show Title Show Count Show Image Replicate 1 Do not collect results Priority 50 Routing In Priority Route In Objects Waiting for Adaptation Require resources before collecting any work items Routing Out Circulate Preference only Route Out Objects Queue for Goal Attainment Release resources as soon as task complete Operation Time Distribution Detail: Uses: ADwell Label Based Work Center Integration ----------- Display Parameters 4 X:660 Y:509 W:32 H:32 Xinc -10 Yinc 0 Show Title Show Count Show Image Replicate 1 Priority label: Affect Routing In Priority

Page 121

Route In Objects Queue from GA affect neutral Goal Attainment Affect Interrupted Integration Require resources before collecting any work items Routing Out Circulate Preference only Route Out Objects Queue for Latent Pattern Maintenance Release resources as soon as task complete Operation Time Distribution Detail: Fixed 120 0 0 0 Interruptable Work Center Latent Pattern Maintenance -------------------------- Display Parameters 4 X:376 Y:506 W:32 H:32 Xinc -10 Yinc 0 Show Title Show Count Show Image Replicate 1 Do not collect results Priority 50 Routing In Priority Route In Objects Queue for Latent Pattern Maintenance Require resources before collecting any work items Routing Out Circulate Preference only Route Out Objects Spent Energy Release resources as soon as task complete Operation Time Distribution Detail: Uses: LDwell Label Based Label Actions Affect None Work Exit Point Energy that does not enter -------------------------- Display Parameters 4 X:238 Y:368 W:32 H:32 Xinc -10 Yinc 0 Show Title Show Count Show Image Input Objects

Page 122

AFilter Storage Bin Waiting for Adaptation ---------------------- Passed the Adaptation filter and now waits for a blocked Adaptation activity, which is evidently busy making sense of "old" news. Display Parameters 0 X:392 Y:292 W:32 H:32 Xinc -10 Yinc 0 Show Count Show Image Do not collect results Capacity: -1 Input Objects AFilter Output Objects Adaptation Storage Bin Queue for Goal Attainment ------------------------- Display Parameters 5 X:563 Y:292 W:32 H:32 Xinc -10 Yinc 0 Show Count Show Image Do not collect results Capacity: -1 Input Objects Adaptation Output Objects Prepare budget proposal Storage Bin Queue from GA affect neutral ---------------------------- Display Parameters 0 X:658 Y:438 W:32 H:32 Xinc -10 Yinc 0 Show Count Show Image Capacity: -1 Input Objects Prepare budget proposal Output Objects Integration Storage Bin Queue for Latent Pattern Maintenance ------------------------------------ Display Parameters 0 X:473 Y:508 W:32 H:32 Xinc -10 Yinc 0 Show Count Show Image Do not collect results

Page 123

Capacity: -1 Input Objects Integration Output Objects Latent Pattern Maintenance Work Exit Point Spent Energy ------------ Display Parameters 4 X:374 Y:696 W:32 H:32 Xinc -10 Yinc 0 Show Title Show Count Show Image Input Objects Latent Pattern Maintenance Work Center Goal Attainment Affect ---------------------- Display Parameters 4 X:733 Y:225 W:32 H:32 Xinc -10 Yinc 0 Show Title Show Count Show Image Replicate 1 Do not collect results Priority 50 Routing In Priority Route In Objects Queue for Goal Attainment Affect Require resources before collecting any work items Routing Out Circulate Preference only Route Out Objects Integration Release resources as soon as task complete Batching Product type of fixed value: 1 Operation Time Distribution Detail: Uses: GDwell Label Based Storage Bin Interrupted Integration ----------------------- Display Parameters 0 X:836 Y:520 W:32 H:32 Xinc -10 Yinc 0 Show Title Show Count Show Image

Page 124

Capacity: -1 Output Objects Integration Storage Bin Queue for Goal Attainment Affect -------------------------------- Display Parameters 0 X:610 Y:225 W:32 H:32 Xinc -10 Yinc 0 Show Count Show Image Do not collect results Capacity: -1 Input Objects Affect processing Output Objects Goal Attainment Affect Work Center Affect processing ----------------- Display Parameters 4 X:496 Y:225 W:32 H:32 Xinc -10 Yinc 0 Show Title Show Count Show Image Replicate 1 Do not collect results Priority 50 Routing In Passive Route In Objects AFilter Require resources before collecting any work items Routing Out Circulate Preference only Route Out Objects Queue for Goal Attainment Affect Release resources as soon as task complete Operation Time Distribution Detail: Uses: ADwell Label Based Label Actions Label None Work Exit Point Ideas not resourced ------------------- Display Parameters 4 X:827 Y:271 W:32 H:32 Xinc -10 Yinc 0 Show Title

Page 125

Show Count Show Image Do not collect results Input Objects Prepare budget proposal Work Center Prepare budget proposal ----------------------- Display Parameters 4 X:623 Y:292 W:32 H:32 Xinc -10 Yinc 0 Show Title Show Count Show Image Replicate 1 Do not collect results Priority 50 Routing In Priority Route In Objects Queue for Goal Attainment Require resources before collecting any work items Label Batching Label Min batch quantity 1 Max batch quantity 10000 Routing Out Percent Route Out Objects Ideas not resourced (0%) Queue from GA affect neutral (100%) Release resources as soon as task complete Operation Time Distribution Detail: Uses: GDwell Label Based Information Store ----------------- Simulation Time --------------- SIMUL8 Data Current Value 0 Warm Up Period -------------- SIMUL8 Data Current Value 0 Results Collection Period ------------------------- SIMUL8 Data

Page 126

Current Value 2400 Current Work Item ----------------- SIMUL8 Data Current Value 0 Overhead Cost ------------- SIMUL8 Data Current Value 0 Overhead Revenue ---------------- SIMUL8 Data Current Value 0 Graph Sync Interval ------------------- SIMUL8 Data Current Value 1 Temp ---- Number Current Value 0 Reset Value 0 Table ----- Spreadsheet EnergyThreshold --------------- Number Current Value 0 Reset Value 2 Relation -------- Text Current Value >= Reset Value NOCHANGE GT -- Text Current Value > Reset Value NOCHANGE GTE --- Text Current Value >= Reset Value NOCHANGE Detail Timing Log

Page 127

----------------- Spreadsheet y - Number Current Value 0 Reset Value 0 x - Number Current Value 0 Reset Value 0 p - Number Current Value 0 Reset Value 0 k - Number Current Value 0 Reset Value 0 r - Number Current Value 0 Reset Value 0 Saved_clock ----------- Time Current Value 0 Reset Value 0 RowCtr ------ Number Current Value 29 Reset Value -2147483648 ValCol ------ Number Current Value 10 Reset Value 10 BotRow ------ Number Current Value 29 Reset Value 29

Page 128

TimeOfMax --------- Time Current Value 0 Reset Value 0 MaxEnergy --------- Number Current Value 0 Reset Value 0 ThresholdToRespond ------------------ Number Current Value 0 Reset Value 0 PercentNotFunded ---------------- Number Current Value 0 Reset Value -2147483648 DecayCoefficient ---------------- Number Current Value 0 Reset Value -0.001 ResponseWindow -------------- Number Current Value 0 Reset Value 52 Summarize Input Data -------------------- Spreadsheet MovingAverage ------------- Number Current Value 0 Reset Value 0 NbrInputs --------- Number Current Value 0 Reset Value 0 RunningTotal ------------ Number Current Value 0 Reset Value 0

Page 129

TempRow ------- Number Current Value 0 Reset Value 0 I - Number Current Value 0 Reset Value 0 Divisor ------- Number Current Value 0 Reset Value 0 Reset Visual Logic: VL SECTION: Reset Logic 'Obeyed just after all simulation objects are initialized at time zero Clear Sheet Table[1,1] 'Read external Excel sheet into internal one in order to speed up execution. Get from EXCEL Table[1,1] , "[GameInput.XLS]ParsonsGameInput" , 1 , 1 , ValCol , BotRow-1 'Set constants from the table just read in. SET PercentNotFunded = Table[ValCol,17] SET p = Table[ValCol,22] SET r = Table[ValCol,23] SET k = Table[ValCol,24] SET Table[3,BotRow] = "Affect" SET Table[4,BotRow] = "External" SET Table[5,BotRow] = "Clock" SET Table[6,BotRow] = "Internal" SET RowCtr = BotRow Clear all Display Plus "" SET EnergyThreshold = Table[ValCol,19] SET Temp = EnergyThreshold Get from EXCEL Relation , "[GameInput.XLS]ParsonsGameInput" , ValCol , 20 , 1 , 1 'GT is the greater than relation (>) and GTE is greater than or equal to (>=). SET ThresholdToRespond = Table[ValCol,26] SET DecayCoefficient = Table[ValCol,27] SET ResponseWindow = Table[ValCol,28] Set Route Out Discipline Prepare budget proposal , Percent Set Route Out Percent Prepare budget proposal , PercentNotFunded*100 , Ideas not resourced Set Route Out Percent Prepare budget proposal , [1-PercentNotFunded]*100 , Queue from GA affect neutral Start Run Visual Logic: VL SECTION: Start Run Logic

Page 130

'Obeyed every time the user clicks the RUN button (at any simulation time) End Run Visual Logic: VL SECTION: End Run Logic 'Obeyed when the simulation reaches end of "Results Collection Period" SET RowCtr = RowCtr+1 SET Table[3,RowCtr] = "" SET Table[4,RowCtr] = "" SET Table[5,RowCtr] = "" SET Table[6,RowCtr] = "" Set in EXCEL Table[3,BotRow+1] , "[GameInput.XLS]ParsonsGameInput" , 3 , BotRow+1 , 4 , [RowCtr-BotRow]+2 Other Visual Logic: VL SECTION: Learning Model Common 'Common processing for the learning model. 'Construct the data that are in the "Window". SET NbrInputs = NbrInputs+1 IF NbrInputs > ResponseWindow 'Find the Maximum in the new range. SET MaxEnergy = 0 SET TempRow = [BotRow+NbrInputs]-ResponseWindow LOOP TempRow >>> I >>> [[BotRow+NbrInputs]-1] IF Table[4,I] > MaxEnergy SET MaxEnergy = Table[4,I] 'Adjust the MovingAverage by dropping off the least recent entry in the Window. SET RunningTotal = RunningTotal-Table[4,TempRow] SET Divisor = ResponseWindow ELSE SET Divisor = NbrInputs IF Energy > MaxEnergy IF [Energy-MaxEnergy] > ThresholdToRespond 'The new Energy has to be greater than Max by a threshold in order to make a change. 'Here if there is a new maximum in the external energy (within the Window). SET MaxEnergy = Energy SET k = MaxEnergy*EXP[DecayCoefficient] SET TimeOfMax = Simulation Time SET RunningTotal = RunningTotal+Energy ELSE 'Standard case: no new maximum. SET k = MaxEnergy*EXP[[Simulation Time-TimeOfMax]*DecayCoefficient] SET RunningTotal = RunningTotal+Energy SET MovingAverage = RunningTotal/Divisor IF k < MovingAverage SET k = MovingAverage 'The code below represents the learning by LPM in order to adjust the energy filter to a target value in the Adaptation function. 'p represents cum prior learning interval, so have to add current elapsed time to running total SET p = p+Simulation Time

Page 131

'x is the time since the last change SET x = Simulation Time-Saved_clock 'Now compute the new value to be used as an energy filter in Adaptation. SET Temp = k*[[x+p]/[[x+p]+r]] 'Save the current time so that it can be used next time for the computation of x SET Saved_clock = Simulation Time 'Prepare elements needed to be saved in spreadsheet so that we can follow the graph of the external energy vs. the internal energy generated by LPM. SET RowCtr = RowCtr+1 SET Table[3,RowCtr] = Affect SET Table[4,RowCtr] = Energy SET Table[5,RowCtr] = Simulation Time SET Table[6,RowCtr] = Temp