-
1
Group Decision Support System for System Design
Nagib Callaos,*** Richard Evans,** William Lesso,* and Belkis
Callaos***
* Professor Emeritus and Former Associate Dean of the College of
Engineering of the University of Texas at Austin ** Department of
Computer Science, George Mason University *** Department of
Processes and Systems, University Simon Bolivar and The
International Institute of Informatics and Systemics (IIIS)
Presented as two Plenary Keynote Addresses at the
5th
World Multiconference on Systemics, Cybernetics and Informatics,
2001
Abstract
Our purpose in this article is to combine three previously
published papers in order to apply a general Methodological frame
to a specific problem. Because system design is a special case of
problem solving, we will try to apply A Generalized Group Decision
Support System (GDSS) for Group Problem Solving: (Callaos, et.al.
1999; Callaos and Lesso, 1999) to the specific area of System
Design. This will be done by means of: 1) describing a GDSS for
Group Decision Making, based on the Mathematical Solution to the
Voter Paradox, 2) analyzing the concept of design, and making
explicit its relations to: 2.1) intention, action, decision
(Callaos and Callaos, 1995), and 2.2) to discovery (Evans, 2000),
and group creativity (synectics). The GDSS for Group Problem
Solving proposed here has two basic sub-systems: one would support
the group creativity process, required for the generation of
alternative designs, and the other would support: a) the group
reasoning process, required to establish the pros and cons of each
design alternative and b) the group decision process, based on the
Ordinal Scales Delphi Method made possible by the Generalized
Absolute Majority Rule, and/or the optimal hamiltonian path, both
included in the Solution of the Voter Paradox. The ideas presented
in the third part of this paper emerged while Richard Evans, one of
the co-authors of this paper, worked with IBM and NASA over the
past four years. The other two parts are suggested as complement to
the third part, and as a way to make operative the basic ideas
through the software developed according the first part of this
paper. Keywords: GDSS, Group Problem Solving, Voter Paradox,
Design, Intension, Action, Options, Invitation, Nomination,
Confirmation, Self-Assessment, Meeting Purity, Customer, Builder,
Associate, Idea-writing, Requirements-as-design-decisions, Ideas,
Great Question, Introductions, Risk-based Design, Uncertainty,
Headwaters, Insights, Issues, Initiatives, Greenhouses, Systems,
Discovery System, Delivery System, Delivered System, 3-D Thinking,
Dimensions, Independence-is-not-Unilateralness
-
2
Introduction The main objectives of this paper are:
1. To relate three previously elaborated papers in order to
achieve an integrated framework aimed at the application of a
Collective Decision Support System (GDSS) to systems design,
especially in the area of complex systems, where design decision
should not be made by a person, due to the multiple effects of such
kind of decisions. So, the content of this paper is based on the
principle that people affected by the effects of design decisions
should participate in such decisions.
2. To generate a working paper, where each one of the co-authors
could elaborate with more details some issues of it, add other
related issues, and where other scholars, researchers, or
consultants could participate via feedback, or by means of
elaborating further related aspects.
3. To entice other researchers, scholars and consultants to
contribute with related papers to the track of this topic included
in the 5th World Multi-Conference on Systemics, Cybernetics and
Informatics (SCI 2001)
The content of the paper will have, basically, three sections,
associated to the three papers to be related. In the first section
we will try to describe a conceptual design and a basic
architecture of a Generalized Group Decision Support System (GDSS)
aimed to sustain processes of Group Problem Solving. This will be
done by means of integrating: 1) problem definition and
typification, 2) macro-phases of synectic methods, 3) GDSS types,
4) Collective Decision Theory with its related mathematical
solution to the Voter (or Condorcet) Paradox and 5) application of
the Operations Research Approach to this specific kind of problems.
The Generalized GDSS to be described in the first section has been
applied to several specific cases as, for example, Strategic
Planning, Reengineering, Management Control, Participatory
Management, etc. In this paper we will try to apply it to
participatory systems design. This is a special kind of group
problem solving. So a Generalized Group Problem solving
methodology, and a Generalized GDSS, should be applicable to the
specific case of Systems Design. In the second section of this
paper we will try to work a systemic notion of âdesignâ, which
would, consequently, be an integrated and an integrative one. To be
integrated, the notion of design should include its contextual
relations, and to be integrative should be a comprehensibly
unifying one. This would give us the conceptual framework for the
principles this papers is based on. Consequently, we will attempt
to identify an initial semantic comprehensive structure of the
notion of âdesignâ, which will lead us to a hypothetical conceptual
infrastructure, which, in turn, will give us the pointers to the
contextual relations and the other notions strongly related to the
notion of design. After examining these contextual notions, we will
check the validity of the hypothesis made about the conceptual
infrastructure proposed for the notion of âdesignâ. Banathy, et.
al. (1979) made a very wide exploration on the different
meanings/definitions given to the notion of âdesignâ and extracted
the commonalties they identified among the high diversity they
found. This is a synthetic-inductive approach to the problem. We
will try a synthetic-inductive/deductive approach i.e. a very brief
semantic synthetic-induction first,
-
3
concluding in a hypothesis formulation, followed by an
analytic/deductive process oriented to the conceptual context.
Furthermore, we will try, in the second section of this paper, to
provide the reasoning about the principle, given above, which will
be one of the basic support of this paper i.e. the people affected
by systems design decisions should participate in the decision
making process of these design decisions.
In the third section we will try to show that design involves
discovery and since discovery is based on options generation, the
foundation of design is option generation and assessment. Option
generation might be supported be one of the GDSSâ sub-systems
presented in the first section (the electronic brainstorming one),
and since the assessment should be a collective one, it might be
supported by another GDSSâ sub-system (the collective decision
support one). The fostering of the emergence of design ideas, and
the nurturing of their growth, is an essential part of system
design, i.e. discovery. An adequate decision related to the
discovered options is the other essential part or any designing
process. The identification of an adequate set of ideas, and their
comprehensive assessment, is effected by inviting, never imposing;
by persuading, never compelling; and by nominating and confirming,
never by directing or "requiring". The so-called requirements are
design decisions, they are never separate stand-alone sentences.
And, as design decisions, they should be participative ones, where
people affected by these decisions are who should participate in
the respective decision making process. The GDSS proposed in the
first section would give the required support for this
participatory process and for the required collective decision. As
we will show in the third section, design as discovery relies in a
major way on the united operation of the three roles of Customer,
Builder, and Associate, as well as their respective relationships.
It is also suggested (in the third section) that there are only
these three roles, whether for person-to-person or
organization-to-organization relationships. In
organization-to-organization relationships, a GDSS like the one
described in the first section would be highly useful. Adequate
design requires a combination of creative intuitions, and
structured thinking along with conversations, with others and/or
with oneself. The GDSS described below have the potential of
effectively supporting this kind of thinking and conversational
processes and, in general, group creative processes as those
described in the third section of this paper. 1. A Generalized GDSS
for Group Problem Solving
1.1 Basic Definitions
The word âproblemâ derives from the Latin term âproblemaâ, and
the Greekâs âprĂłblemaâ, formed on âprobĂĄlleinâ which means âput
forthâ, âto put further on in the direction in questionâ. As long
as the direction (in question) is related to explicit or implicit
objectives, the term âproblemâ would mean, âto put further or in
the direction of achieving explicit or implicit objectives,
purposes or ends. Since ends are achieved through means, there will
be two types of problems and, hence, two non-exclusive types of
problem solving processes:
-
4
P.1. Given a set of ends, inquire about the means to achieve
them. P.2. Given a set of ends and a set of possible means to
achieve them, find the âbestâ set of
means, where âbestâ is related to the given ends (effectiveness)
and to the resources required (efficiency), as well as to possibly
other implicit objectives.
We will see later that design is strongly related to intention,
purpose and objectives, and designing processes are a special kind
of problem solving processes. Consequently there will be two
non-exclusive types of designing processes, analogous to the types
of problem processes briefly described above. Accordingly, all what
we can conclude, find, or propose in the domain of general problem
solving might also be applied to design, designing processes,
system design and system designing processes. Now let us go back to
the problem-solving domain. A situation where ends or objectives
are not given but should be identified and established is a
meta-problematic situation. (In the design domain we will be
dealing with a meta-design problem) In such a case, we are faced
with a meta-problem, i.e. âthe problem of defining the problemâ,
i.e. our end or purpose is to identify the objectives related to
the situation, and to do so, we have to determine adequate means,
or to select the âbestâ subset from a given set of means. This
recursive characteristic allows us to focus on the problematic
level in order to make a conceptual design of a GDSS oriented to
support problem solving processes, because a similar conceptual
GDSS will also be able to support a meta-problematic situation
solving process. Likewise, this recursive formulation will allow us
to focus in the design process, since meta-design level might be
treated in a similar way. The definition of âproblemâ is by itself
a problem, i.e. a meta-problem. Different authors define âproblemâ
according to implicit or explicit objectives. Consequently, diverse
definitions have emerged, as related to different objectives. Our
definition of âproblemâ does not replace or exclude other
definitions; on the contrary it includes them in a coherent
systemic whole. According the objectives of the definer, a
definition might be more or less adequate. This is coherent with
the pragmatic-teleological truth of the Systems Approach
(Churchman, 1971). Similarly, the definition of âdesignâ is also a
problem because different authors provide different definitions
based on implicit or explicit design objectives. So, a systemic
definition of design could be made using the GDSS we will describe
in this section. Problem solving is the process of finding and/or
selecting the best means to achieve some objectives. Different
problem solving methods has been suggested. These methods are
basically heuristic procedures, because they include phases related
to the stimulation of individual creativity, i.e. creatics and/or
group creativity, i.e. synectics. Both kinds of methods are
important for the notion of design as discovery, as it will be
described below, especially in the third section of this paper.
Creativity heuristic methods have usually two macro-phases: a
non-rational, intuitive, non-structured one, and a rational,
non-intuitive, structured one. Different kinds of Collaborative
Information Systems have been developed to support each one of
these two phases. Electronic
-
5
Brainstorming and Electronic Idea Generation Systems aim at the
support of the first phase (Laplante, 1998; Aiken, et.al., 1996,
Aiken, et.al., 1994; Dennis, et.al., 1998; Easton, et.al., 1990;
Gallupe, et.al., 1992; Herniter and Gargeya, 1995; Nunamaker,
1991). Model-based, Multicriteria and Expert Systems-based
Collaborative Information Systems have been developed to support
the second phase (Belton, 1999; Fjermestad, 1998; Podinouski, 1999;
Siskos and Spyridakos, 1999). According to our objectives in this
paper we can distinguish between GDSSes for non-structured and
structured problems. These two kinds of GDSSes will be used in this
paper for the two macro-phases of synectics, respectively. A more
detailed and comprehensive typification of GDSS could be found in
Mirchandani and Pakath, 1999. 1.2. Collective Decision Theory
One important issue, not frequently treated in GDSS literature
is Collective (or Group) Decision Theory. Group Decision Making
based on ordinal individual preferences, and its respective Voter
(or Condorcet) Paradox, is, to our knowledge, missing in the GDSS
literature. This is a paradox in itself: How could Group or
Collective Decision Theory (CDT) be absent from Group Decision
Support Systems? Consequently, we will try to insert CDT into the
GDSS architecture we are suggesting in this paper. We did this
insertion in special GDSS we developed for different specific
applications, obtaining encouraging results and positive feedback
from our clients, and from the respective GDSS users. Elsewhere we
briefly described the Collective Decision Problem and its related
Voter (or Condorcet) Paradox (Callaos, 1976a; 1976b; 1980; and
Callaos et.al, 1999): Transitive (rational) individual preferences
may generate intransitive (irrational) collective preference, if we
apply the quasi-universally accepted Absolute Majority Rule. We
also presented our Mathematical Solution to the Voter Paradox,
based on what we called The Generalized Absolute Majority Rule
(GAMR). We also contrasted our solution to Arrowâs Impossibility
Theorem. Arrow (1951) made a mathematical demonstration about the
impossibility of solving the Voter Paradox. But, we did solve it
and, consequently, showed Arrowâs axioms inconsistencies and
methodological weaknesses. In the Appendix we present examples of
the Voter Paradox and a visualization of its mathematical solution.
1.2.1 The Voter Paradox
Collective Decision Theory is at the heart of the GDSS proposed
here. Consequently, we will briefly describe the problem of
aggregating individual preferences into a collective preference,
along with the Voter Paradox, and Arrowâs Impossibility Theorem.
Ordinal data aggregation and ordinal social decision-making will be
analyzed and solution to the Voter Paradox will be provided. We
will present our critiques with regards to Arrowâs Axioms logical
inconsistencies and his methodological weakness. This will be done
as a product of our approach to the problem. The proposed solution
has been and is being used in several research projects and in
various real life projects. Software, based on the solution
provided here, is being developed now a Beta test of an initial
version is in progress. A first working prototype is being applied
at present for several group judgment problems for re-engineering
processes (a special kind of
-
6
system design) and group opinion formation, generation and
elicitation in an electoral situation for University authorities.
The problem of aggregating individual preferences into a collective
preference could be stated as follows: Given a set of alternatives
A = {a1, ..., am} and the individual preferences of n persons, find
an accepted rule for determining the collective preference of these
n individuals, i.e. find a function SPF (social preference
function) such that:
SPF: Pn(A) âP(A)
where P(A) = Am and Pn(A) = (P(A))n.
The most universally accepted rule is the Absolute Majority.
This rule is for a set of two alternatives, and its application for
three or more alternatives does not necessarily produce a
collective preference. This characteristic of the Absolute Majority
is better known as the Voter Paradox, which was first formulated by
Condorcet (1743-94). This is why it is also known as Condorcet
Paradox. Since it was formulated, several authors tried to find a
solution. Borda (1781), Laplace (1812), Dodgson (Lewis Carroll)
(1876), Nanson (1907), Galton (1907), Hoag and Hallett (1926), etc.
tried to find an intuitively acceptable rule as substitute for the
Absolute Majority one. Their efforts were oriented toward a
consensual truth1. Arrow (1951) was the first who tried an
axiomatic approach directing his effort toward an analytical truth.
Our research was oriented toward a pragmatic-teleological truth as
a goal and we used the consensual and the analytical truths as
means to achieve such a goal. The efforts of the authors that
looked for a consensual truth were in vain. Arrowâs efforts
conveyed him toward his known Impossibility Theorem, which states
that for a given set of universally accepted conditions, there is
no collective decision rule, i.e. there is no transitive SPF.
1.2.2. Proposed Solution to The Voter Paradox
The problem of aggregating individual preferences into a
collective preference could be restated as follows: Given a set of
alternatives A = (a1, ... , am) and n individuals, each one with a
preference function over the set A, find the âoptimalâ SPF, where
âoptimalâ would mean the collective preference associated the
minimum of total un-satisfactions2, or minimum variance between
un-satisfactions3, or a tradeoff between both minimums. This
problemâs restatement allowed us to find a solution for the Voter
Paradox that contains as a special case the Majority rule for both
cases: 1) for two alternatives, and 2) for cases of more than two
alternatives where no Voter Paradox is presented, i.e. where we can
find transitive collective preference by means of the traditional
Absolute Majority. In this sense we can say the proposed solution
is a generalization of the âMajority Ruleâ. To do so, we went to
the reasons that make the Absolute Majority Rule so intuitively
attractive, just and fair. The answer, in our opinion, is that
the
1 For a detailed description of the meaning of âconsensual
truthâ, âanalytical truthâ and âpragmatic teleological truthâ see
Churchman (1971) and Callaos (1980). 2 Capitalist flavor 3
Socialist flavor
-
7
Absolute Majority Rule is the statistical media and as such is
an âoptimalâ point in the sense of minimizing un-satisfaction (or
maximizing satisfactions). In other words: if we have two
alternatives A = {0,1} and n0 voters preferring 0 to 1, n1 voters
preferring 1 to 0, and we want to decide between 0 and 1 in such a
way as to maximize satisfactions (or minimize un-satisfactions),
where an individual is satisfied when the collective decision
coincides with his or hers, otherwise he is unsatisfied.
Mathematically the problem could be described as follows: Maximize
Z = n0(1-x) + n1x (1)
subject to
x = 0,1
So the answer is immediate, i.e.
x=0 if n0 > n1
x=1 if n0 < n1
Therefore, to solve the Paradox, we should generalize the
optimization problem, given above, to more than two alternatives.
There are two kinds of collective ordinal preference, according to
whether the alternatives to be ranked are mutually exclusive (as in
a presidential election or selecting a constitutional clause among
several alternatives) or not (as in election of a board, or a
budget ranking). The mutually exclusive case is the one who
received much attention from different authors. The non-exclusive
case has been less frequently treated. We will focus here on the
mutually exclusive kind. Elsewhere (Callaos, 1971) we treated the
case for non-exclusive alternatives. For the case of mutually
exclusive alternatives we will propose a formulation for what we
might call the âGeneralized Absolute Majority Ruleâ (GAMR).
Immediately we will prove that GAMR (which is an SPF for m
alternatives), implies the Absolute Majority Rule (which is an SPF
for 2 alternatives). In other words: we will show that the Absolute
Majority Rule is a special case of GAMR, i.e. the case where m=2.
Therefore, the same reasons that make the Absolute Majority Rule so
universally acceptable should make the GAMR universally acceptable
as well. In the Appendix we present examples of the Voter Paradox
and a visualization of its mathematical solution. To formulate the
GAMR let us represent the preferences aggregation problem by a
graph. To generalize the Absolute Majority Rule is to find the
hamiltonian path of alternatives aĎ1, aĎ2, ⌠, aĎm (where Ďi â Ďj
âiâ j and Ďi = 1, 2, ⌠, m) in such a way as to maximize: NĎ1,Ď2 +
NĎ1,Ď3 + ⌠+ NĎ1,Ďm + NĎ2,Ď3 + NĎ2,Ď4 + ⌠+ NĎ2,Ďm + NĎm-1,Ďm where
NĎi,Ďj is the number of individuals preferring alternative aĎi to
aĎj. Therefore the mathematical formulation of the problem will be
as follows:
-
8
Max Z = â=
m
i 1â
=
m
k 1
xik (ââ
=
m
rr
11
Nir - ââ
=
m
ss
11
Nis (â
-
9
Z* ⼠ââ
=
1
1
m
iâ
+=
m
ij 1
NĎi, Ďj (8)
Let us show that no alternative aq â aĎ1 could be obtained as a
winning one by applying the AMR. Let us suppose that aq is the AMR
winning alternative, then: Nq,Ďj > NĎj,q â Ďj = 1, ⌠, m; Ďj â q
(9) and let t be the place assigned to aq in the ordering obtained
by GAMR and given in (6), i.e. aĎ1, aĎ2, ⌠, aĎt-1, aĎt, ... , aĎm
(aĎt = aq). We have already shown that for any different ordering
should satisfy (8), then for the ordering: aĎ1, aĎ2, ⌠, aĎt-1,
aĎt+1, ... , aĎm the following un-equality is satisfied:
Z* ⼠â=
m
j 2
NĎ1,Ďj +â=
m
j 3
NĎ2,Ďj + ... + â==
m
tj 1
NĎt,Ďj + NĎt,Ďt-1 +â=
m
tj
NĎt-1,Ďj - NĎt-1,Ďt + â+=
m
tj 2
NĎt+1,Ďj
+ ... + â=
m
mj
NĎm-1,Ďj (10)
From (7) and (10): 0 ⼠NĎt,Ďt-1 - NĎt-1,Ďt and since aq is the
AMR winning alternative, unequality (9) should be satisfied,
then:
NqĎt-1 ⤠NĎt-1q (since aq = aĎt)
NqĎt-1 > NĎt-1q
Then there is no aq â aĎ1 that could be the AMR winning
alternative. Therefore if there is any AMR winning alternative, it
will be the same one as the GAMR winning alternative. Therefore if
there is any AMR winning alternative, it will be the same one as
the GAMR winning alternative. 1.2.3. Arrowâs Impossibility
Theorem
The Voter (or Condorcet) Paradox was still unsolved when Kenneth
Arrow (1951) claimed to have roved mathematically the impossibility
to solve it, Arrow proved that there is no voting rule or method,
i.e. no transitive SPF that could have the followingâapparently
innocousâconditions, which where the initial axioms in his
axiomatic approach:
1. Unlimited Domain: The domain of the SPF should be Pn(A), i.e.
SPF should be defined for all possible sets of individual
preferences.
2. Positive Association of social and Individual Values: If one
alternative raises or remain still in the preference of every
individual, then âceteris paribus,â it must not fall in the
SPF.
3. Independence of Irrelevant Alternatives: If the removal from,
or insertion into, the set of alternatives of a certain alternative
âaâ results in no change in any individual preference
-
10
of the remaining alternatives, then it must cause no change in
the SPF of those alternatives.
4. Citizen Sovereignty: For each pair of alternatives âaâ and
âbâ there is some set of individual preferences for which the
collective preference rank âaâ above âbâ.
5. Non-Dictatorship: The social preference between any two
alternatives must not coincide with any one individual regardless
of the preferences of other individuals.
Arrow proved mathematically that no SPF can fulfill these
conditions (axioms) while keeping its transitiveness. This
diminished, in our opinion, further research with regards to
finding a solution to the Voter Paradox.
1.2.4. Inconsistencies in Arrowâs Axioms and Methodological
Weakness of his Approach
Elsewhere (Callaos, 1976a) we exposed at length our arguments
against Arrowâs approach. Here we will present an adapted version
of the summary we presented in other place (1976b): 1. Several
critiques have been written to Arrowâs approach, most of them from
the philosophical implications of his axioms, and others in
reformulating his axioms in a way that the Voter Paradox could be
solved. But, no criticism has been given yet, as far as we know, on
the logical foundation of his methodology. In our opinion Arrowâs
five axioms (or conditions) are explicitly contradictory among
themselves, so it is not surprising that Arrow would prove their
contradiction after a nice âlogical jugglingâ.
2. In condition 4 (Citizen Sovereignty) Arrow means that the SPF
should not be imposed from outside of the individual preferences
set. On the other hand Arrow imposes other conditionsâcalled
reasonable by himâon the SPF. This is a SPF non-imposed and imposed
from the outside and at the same time. This is a methodological
contradiction.
3. Beside his five conditions, Arrow implicitly imposed another
one, namely âSocial Rationalityâ or transitivity in the SPF.
âRationalityâ and âsocietyâ belong to different logical categories
(unless we adopt and organismic philosophical point of view).
âRationalityâ is an individual characteristic, not necessarily a
social one. Furthermore, even if we can adjudge rationality or
non-rationality to a society, why should we equate it with the
individualistic one? Furthermore, if we were to define âsocial
irrationalityâ as intransitivity in the SPF, it would automatically
depend on the rule, or voting method, used. The same society in the
same voting process could be âirrationalâ according to the Majority
Rule and ârationalâ according to another rule and this is a
contradiction.
4. Condition 3 (Independence of Irrelevant Alternatives) is
untenable under the light of our approach. It is absurd to believe,
and to impose, the same result we get by optimizing for two
variables (two alternatives) than for more of them. Sub-problems
optimization does not assure the global problem optimization.
Optimization of subsystems in an individual and independent way
does not assure the optimization of the whole system.
-
11
5. Condition 3 imposes indirectly the Majority Rule, which is
based on binary choice. So, what Arrow did was to have the Majority
Rule as one of his satisfactory conditions; then he re-proves the
Condorcetâs Paradox, i.e. Majority Rule could lead to intransitive
SPF. Condorcet had already done it in a very simple and still
rigorous way. This would leave Arrowâs work with great
emptiness.
1.3. Advantages of Including the GAMR in GDSS
Existing GDSS might get great benefits inserting adequately our
Solution of the Voter Paradox (Callaos, et. al., 1999) as a
subsystem in their architecture. There are fundamentally two basic
reasons to think so: 1. GDSS could help in getting a
group/collective decision based in ordinal scales, avoiding
both: 1.1. The Voter Paradox, i.e. the possible generation of
intransitive group preferences, or
âirrationalâ group decisions. Simulations programs showed that
the Voter Paradox would emerge in more than 30% for three
alternatives, and more than 90% for six alternatives (or more), if
we apply the Majority Rule, but the Voter Paradox does not exist
with our solution.
1.2. The uncertainty of working with cardinal scales, because
possible inter-subjectivity
inconsistencies. Different subjects, from the group, could use
different cardinal scales, and there is no way to know it, or to
prevent its possible occurrence. Ordinal scales do not have this
kind of uncertainty.
2. The Delphi Method could be applied through our solution to
the Voter Paradox preserving
and maintaining intact the benefits of the method, which
basically are two: 1) Consensus building (variance reduction among
individual judgments), and 2) moving group decision closer to
facts, veracity, or agreement with reality. As it is known the
Delphi Method is based on group members who would: 1) provide their
opinions or judgments in cardinal scales; 2) get feedback related
to opinions/judgments average and variance; and 3) have the
opportunity to change their opinions/judgments according to the
feedback received. It is also known that the Delphi Method (in
cardinal scales) build consensus (decreasing the variance) and move
the average to the real value with each feedback loops, until
getting stabilized after 3 to 5 of these loops.
Ordinal Delphi could not be conducted because the Voter Paradox,
and because other group decision rules, different to the Absolute
Majority Rule (relative majority rule, for example) do not preserve
the Delphi characteristic benefits. In fact some voting rules
(different to the Absolute Majority) proved to be detrimental,
because they may move the group preference and, hence, their
decision away from the real value or from the truth. Our Solution
to the Voter Paradox, the Generalized Absolute Majority Rule (GAMR)
preserves the Delphi benefits.
-
12
Consequently, our GAMR might complement GDSS, by means of
removing the uncertainty of group decision making in cardinal
scales, making feasible the use of ordinal scales and Ordinal
Delphi, and preserving the benefits of Cardinal Delphi in Ordinal
Delphi (Figure 1).
Figure 1 1.4. Basic Architecture of a Generalized GDSS for Group
Problem Solving
The two sections we presented above converge in this one. As we
said above, there are two basic kinds of problems (means
identification and/or selection for given objectives) and two
macro-phases in synectics (unstructured and structured). Both
typifications relate to each other: 1. Means identification (for a
given set of objectives) requires synecticsâ unstructured phase. 2.
Means selection (for a given set of objectives and means) requires
synecticsâ structured phase. On the other hand, problem solving
requires creativity and/or creative methods. Hence, synectics could
support group problem solving, and GAMR-based GDSS could, in turn,
support synectics processes (figure 2).
Individual
opinions, judgements
and/or decisions
Group Opinions, judgements or decisions: 1. Without
cardinal scales uncertainties
2. With Delphi
methods benefits ⢠Consensus
building ⢠Group
decisions closer to facts and/or truth, and/or real values
Ideas
from individuals
Decision Alternatives
Group Preference or Decision
Initial
Group
Preference
Final Group Preference
GAMR-Based GDSS
EXISTING
GDSS
OUR VOTER
PARADOX
SOLUTION:
Generalized Absolute
Majority Rule:
GAMR
ORDINAL SCALES
DELPHI METHOD
-
13
In addition, we said that, according to our aims in this paper,
we might differentiate between GDSSs designed for unstructured
problems and GDSSs conceived for structured problems. This division
relates: 1) to our typification of problems; and 2) to the two
phases of synectics. Then, figure 2 might be detailed as it is
shown in figure 3.
Figure 2
Figure 3
GROUP PROBLEM SOLVING
SYNECTICS METHODS
GAMR-Based GDSS
MEANS
IDENTIFICATION
MEANS
SELECTION
SYNECTIC
INTUITIVE PHASE
SYNECTIC
RATIONAL PHASE
ELECTRONIC
BRAINSTORMING
OR
IDEA
GENERATION
UNSTRUCTURED
GAMR-Based GDSS
DATA AND
KNOWLEDGE BASES,
EXPERT SYSTEMS,
STATISTICAL,
OPERATIONS
RESEARCH AND/OR
DECISION THEORY
MODELS
STRUCTURED
GAMR-Based GDSS
Ends
Ends
Set of Means
-
14
Based on figures 2 and 3, we can suggest a basic architecture of
a Generalized GDSS for group problem solving, as it is shown in
figure 4, where present types of GDSS has been extended and
complemented by our Mathematical Solution to the Voter Paradox, as
it has been describe above and summarized in figure 1.
Figure 4
Electronic
Brainstorming or
Idea Generation:
EBIG
Generation of a
Qualitative
Ends/Means
Matrix
Generation of a
Quantitative
Ends/Means
Matrix
Our Solution
to the
Voter Paradox:
GAMR
GAMR/Delphi
based
on
EBIG-GDSS
Ordinal
Scales Delphi
Operations Research
Modelling (Integer or
Mixed Programming)
for Selecting Best
Means Subset
GDSS based on
Utility Theory,
Policy Capturing or
Ordinal (or
Cardinal)
Preferences
Ordinal and/or Cardinal Delphi
via GAMR
1. Optimal Solution 2. 1st Sub-Optimal 3. 2nd Sub-Optimal ⢠⢠â˘
n. (n-1) Sub-Optimal
GAMR/Delphi
based GDSS
GAMR/DELPHI-based EBIG-GDSS GAMR/DELPHI-based STRUCTURED
GDSS
FINAL GROUP
DECISION
PROVISIONAL
GROUP DECISION
Ends/Objectives
Means Generated
Means Pre-Selected
Individual Pre-Selection
Group Pre-Selection
Pre- Selected Means Set
Group Ideas
Group Ideas
Individual Opinion
Group Opinion
Qualitative Ends/ Means Matrix
-
15
The basic architecture shown in figure 4 could be clarified by
means of describing the fundamental sequence of tasks required for
a typical group problem solving process. This sequence might be as
follows: 1. Given a set of ends/objectives, we need to generate
viable means to attain them. This could be
done supporting a GDSS for Electronic Brainstorming or Idea
Generation (EBIG-GDSS) by our mathematical solution to the Voter
Paradox, in the context of an Ordinal Delphi Method. This the
GAMR/Delphi-based EBIG-GDSS could be implemented following the
basinc main steps:
1.1. Group members interact with an EBIG kind of GDSS,
generating ideas about potential
means to achieve the given objectives. The use of this kind of
GDSS has been reported by several authors and diverse products have
been described in the literature.
1.2. The potential means generated in 1.1, are:
1.2.1. ordered by each group member into individual preferences;
and 1.2.2. ordered into a group preference by means of our Solution
to the Voter Paradox,
i.e. GAMR.
1.3. Group ordinal preference among alternative means is
feedbacked to each group member and he/she is allowed, then, to
change his individual preference, in a Delphi loop until no more
changes are required. Then the group preference given by GAMR will
be the Provisional Group Decision.
2. The set of pre-selected means (through the provisional group
decision) is input to
GAMR/Delphi-based structured GDSS, which output is the Final
Group Decision. This might be achieved through the following
steps:
2.1. An ends/means matrix is generated. 2.2. Each cell of this
matrix is filled with group members opinions about how good or bad
is
each pre-selected means in relation to each objective. A
GAMR/Delphi-based EBIG-GDSS (as the one described in step 1) might
be used to fill each cell with each group member opinions. In this
way we get what it might be called qualitative ends/means matrix,
where columns represent the ends, the âwhatâ should be achieved;
rows represent the means, the âhowâ they might be attained, and the
cells the âwhyâ each means is good (or bad) in attaining each end,
according to group members opinions.
2.3. The qualitative matrix, is transformed to a quantitative
ends/means matrix where:
2.3.1. Objectives are weighed by group members via cardinal
and/or ordinal Delphi.
-
16
2.3.2. Usefulness of each means in relation to each end is
identified by Utility Theory, or Policy Capturing Technique, or any
other adequate technique, at the individual level, and through
Delphi Method at the group level. Ordinal Delphi will be supported
by our Solution to the Voter Paradox (GAMR).
2.4. The quantitative ends/means matrix has all the parameters
required to formulate the
problem at hand, as an integer (or mixed) programming one, or
through other kind of Operation Research modeling.
2.5. Solving the problem, mathematically formulated in 2.4.,
gives us the optimal solution,
as well as 1st, 2nd, ..., nth sub-optimal ones. 2.6. The set of
optimal and sub-optimal solutions (along with sub-optimality
distance from
the optimal solution) are input to a final group decision
process, where the set of solutions are ordered by each group
member and group ordering is achieved via our solution to the Voter
Paradox (GAMR), complemented by an Ordinal Delphi. In this way each
member will have three kinds of information to make his
ordering:
a) The rational results given by the Operational Research model;
which parameters are
based on group judgments.
b) His/her intuitive judgment about such results, which might
contain some important, but not quantified variables.
c) His/her groupâs rational-intuitive results, given trough the
application of GAMR to
individual intuitions with O.R. model rationality.
The result of 2.6 is the final group decision. 1.4.
Conclusions
We tried to make, in this section, a general definition of
âproblemâ, which allowed us to identify problems types and to
relate them to Synectics macro-phases, to Systems Design and to
GDSS types found in the literature. This, in turn, showed us the
basic blocks of the system architecture we are looking for. We
stressed the paradoxical fact that Group Decision Theory is not
found integrated in the GDSS literature. Consequently we proposed
our solution to the Voter Paradox as a way to make this integration
and to extend conventional GDSS, as to include the Ordinal Delphi
Method. Then we presented the general architecture we are proposing
to make the mentioned integrations. In our consulting activities,
we applied the architecture suggested here to a considerable
diversity of GDSS, with different levels of automation. We applied
our proposed GAMR/Delphi-based GDSS to Strategic Planning,
Managerial Control, Curricula Design, Investment Projects
Selection, Technological Innovation Projects Selection, National
Constitution Creation, Marketing, Logo Selection, Reengineering
Projects Methodology Selection, Software Selection,
-
17
etc. In this paper, we are trying to apply our proposed
GAMR/Delphi-based GDSS to Systems Design. Right now, general
purpose software is being developed for the Generalized GDSS
described here. The Venezuelan equivalent to the National Science
Foundation (ComisiĂłn Nacional de Ciencias y TecnologĂa: CONICIT) is
financing partially the project. Private corporations are
co-financing it. An initial version, or a working prototype, is
being beta tested now, through the use if the software to different
problem solving processes. In order to show other potential
applications of the GAMR/Delphi-based GDSS proposed here, and its
respective software, to the domain of Systems Design, it is very
desirable to analyze the concept of design. This will be done in
the next section. 2. The concept of Design A systemic notion of
âdesignâ would be an integrated and an integrative one. To be
integrated, the notion of design should include its contextual
relations, and to be integrative it should be a comprehensibly
unifying one. In this section, we will attempt to identify an
initial semantic comprehensive structure, which will lead us to a
hypothetical conceptual infrastructure, which, in turn will give us
the pointers to the contextual relations and the other notions
strongly related to the notion of design. After examining these
contextual notions, we will check the validity of the hypothesis
made about the conceptual infrastructure of the notion of âdesignâ.
Banathy, et. al. (1979) made a very wide exploration on the
different meanings/definitions given to the notion of âdesignâ and
extracted the commonalties they found among the high diversity they
found. This is a synthetic-inductive approach to the problem. We
will try an inductive/deductive approach: a very small semantic
synthetic/induction first, concluding in a hypothesis formulation,
followed by an analytic/deductive process oriented to the
conceptual context. We will also relate our conceptual finding to
the main pragmatic purposes of this paper. 2.1. A Semantic
Approach
From a thesaurus we observe that there are seven groups of
synonyms of âdesignâ. Three of these are verbs and four are nouns.
So, we might make a preliminary conclusion that there are two
macro-senses in the meaning of âdesignâ: as a process and as a
product, in a temporal existence and in an âatemporalâ one, a
chronological sense and a logical one.
Etymologically, âdesignâ derives from the latin term designare
(to mark out), and this word, in turn, derives from signum (sign).
Peirceâthe founder of semiotics: the science of signsâgives many
definitions of âsignâ, the most referenced one is âa sign ... is
something that stand somebody for something in some respect or
capacityâ(Collected Papers, vol.II, Par. 228; emphasis added) The
notion of âsignâ as âsomething standing for somethingâ has been
very used through history, and it could be associated, by
analogical thinking, to the notion of âre-presentationâ. Hence, the
notion of âdesignâ, as a process, could be thought as âmarking
outâ, âgenerating a signâ, âproducing a representationâ; and as a
product could be thought as the ârepresentationâ produced, the
âsignâ generated. But, in the sense associated with âsystem
design,â it is not any kind of sign or representation; it is not a
fantasy for example. It is a special
-
kind of representation. Representation is the genus of
design.representation is not necessarily a designdifferentiate
design from other species belonging also to the genus of
representation. Up topresent we only need to know that
designprincipal kinds: mental and physical âdesignâ: as a mental
and as a physical representation. Some grthe sense of mental
representation, such as, for example, âintendâ, âaimâ,
âcontemplateâ, âpurposeâ. Other groups are related to the sense of
physical representation such as, for example, âblueprintâ, âchartâ,
âlay outâ, âma A first semantic/conceptual framework could be
derived by crossing the two semantic dichotomies found: âdesignâ as
process and as product, and âdesignrepresentation. In this way we
will get four senses (or the term âdesignâ, as it is shown in table
1the 2x2 matrix.
18
ntation is the genus of design. Design is a representation, but
a necessarily a design. We will try to find the specific
characteristics that
differentiate design from other species belonging also to the
genus of representation. Up toent we only need to know that design
is a representation and, as such, there could be two
physical representations. So, we have another two macro: as a
mental and as a physical representation. Some groups of synonyms
are related to
the sense of mental representation, such as, for example,
âintendâ, âaimâ, âcontemplateâ, âpurposeâ. Other groups are related
to the sense of physical representation such as, for example,
âblueprintâ, âchartâ, âlay outâ, âmap outâ, âset outâ.
A first semantic/conceptual framework could be derived by
crossing the two semantic as process and as product, and âdesignâ
as mental and as physical
representation. In this way we will get four senses (or
sub-notions, or sub-concepts) associatedâ, as it is shown in table
1, where synonyms are distributed in the four cells of
Table 1
Design is a representation, but a We will try to find the
specific characteristics that
differentiate design from other species belonging also to the
genus of representation. Up to the is a representation and, as
such, there could be two
representations. So, we have another two macro-senses of oups of
synonyms are related to
the sense of mental representation, such as, for example,
âintendâ, âaimâ, âcontemplateâ, âpurposeâ. Other groups are related
to the sense of physical representation such as, for example,
A first semantic/conceptual framework could be derived by
crossing the two semantic as mental and as physical
concepts) associated , where synonyms are distributed in the
four cells of
-
19
The term âplanâ appears in each one of the four cells. Hence,
the notion of âplanâ is completely included in the notion of
âdesignâ and has the same four senses given in table 1.
Consequently, it is adequate to differentiate, in the notion of
âdesignâ, between âplanâ and the âobjectâ planned, i.e. the purpose
sought, the aim quested, the intention wrought, the intention to be
achieved by mental effort and/or physical labor. Thereupon, we will
have a semantic/conceptual framework of 2x2x2 matrix, based on
three dichotomies, i.e. process/product, mental/physical
representation, and object sought/plan to achieve the object.
Therefore, eight sub-notions or sub-concepts form the conceptual
infrastructure that supports the notion or the concept of design.
In a future paper we will try to organize the high diversity found
in the literature about the definition of design, according to the
conceptual infrastructure found here. Our pre-hypothesis is that
most authors emphasize into some of the eight sub-notions or
sub-concepts found here, stressing on one characteristic or the
other, as it is more adequate to the case they are sealing with.
For the present purpose, let us assume this pre-hypothesis and
continue analyzing the sub-concepts found, especially those that
provide the contextual relationships we are looking for. Let us
suppose we have an integrative notion of âdesignâ and let us now
direct the search toward making such a notion an integral part of
its conceptual context. In order to do so, we will briefly analyze
the concepts of representation, intention and plan. The first two
concepts have been largely treated in the philosophical literature.
Hence, we will provide here a much resumed treatment as a first
step in this direction. Notice that a collective definition of
âdesignâ might be achieved applying the structured
GAMR/Delphi-based GDSS to the 2x2x2 matrix. In this way, each cell
in the 2x2x2 matrix will be weighted through an Ordinal Delphi
process (which is feasible only by applying GAMR, our solution to
the Voter Paradox). In this way we will have a fuzzy notion, or
concept, which defining fuzzy set is given by the 2x2x2 matrix in
which each cell is weighted by a collective decision making
process.
2.2. Design as Representation
In terms of traditional logic, we identified, so far, the genus
of the notion of âdesignâ and the sub-species of this notion.
âRepresentationâ is the genus of âdesignâ, thus, to define âdesignâ
we should analyze its genusâ comprehension (Port-Royalist) or
connotation (Mill), and its differentia as specie. We have already
identified the eight sub-species of âdesignâ and their respective
differentia as such. Thus, the next step is to identify the
predicates of ârepresentationâ, since what is predicable from the
genus (representation), is also predicable from the specie
(design). After this step we will try to identify the differentia
of the specie âdesign.â As we said before, the notion of
ârepresentationâ has been largely treated along the history of
philosophy: so, all what we will do here is to present a very brief
summary of the features that we think are relevant to our inquiry.
It will be a very first step that could be followed, in the future
by a more explanatory study.
-
20
The etymological meaning of ârepresentâ is to bring into
presence, hence to make clear, demonstrate, symbolize, stand in
place of (Weekley, 1967). From a psychological perspective, we can
distinguish among several kinds of ârepresentationâ, as
follows:
1. Herbert Spencer (1855) distinguished different kind of
cognitions: Presentative, Presentative-Representative,
Representative, and Re-representative, but âwhen simplified, marks
two general classesâpresentative and representative.â (Thomson,
1878) (Presentative and Representative Cognitions (pp. 270-276).
The Presentative cognition, which is presented immediately, is of
two kinds: 1) that which is directly delivered by our
sense-perception, and 2) that which presented in
self-consciousness. Presentative cognition is the apprehension of
an object effectively present, as in perception; or in the
âpresentativeâ knowledge where the related terms correspond to
present and existing objects (Spencer, 1855). Consequently, design
requires the apprehension of an object effectively present. Needs
and design requirements apprehension should be part of the
designing process. Design includes requirements identification,
which are not a given initial condition, as several authors
explicitly say or implicitly stand for. This fact will be retaken
in the last section of this paper. Since requirements
identification imply decisions about what are requirements and what
are not, and for differentiating between necessary and desirable
requirements, as well as for setting the priorities among desirable
requirements, and since these decisions are not always individual
decisions, but are collective, or group ones (as it is the case of
complex systems design serving a variety od users), then
requirements identification, classification and priorities setting
need GDSS to be generated, especially our proposed GAMR-based
GDSS
2. The representation in the mind of past perceptions, i.e.
memory representations, remembrances. When design depends on group
perceptions or collective memory, a GAMR-based GDSS is again a very
desirable feature.
3. The anticipation of future happenings by means of a
combination of past perceptions; be it reproductive or productive.
In this sense, representation is frequently equated to imagination.
Anticipation of undesirable effects of different design options, as
well as its desirable ones, are better identified by a group than
by an each individual of the group, because the techniques of
Judgmental Forecasting. Hence, GAMR-based GDSS is almost a
necessity for this aspect of design.
4. The mental union of several perceptions (not present, nor
past, nor anticipated). In this sense representation is paralleled
with imagination, fantasy, or hallucination. Here, design is a
representation paralleled with imagination, but not fantasy or
hallucination. This characteristic differentiates design from other
representation species. Design is a representation of desirable and
feasible state of affairs, product, process, or system.
Non-feasible or non-possible representations are not designs. Both
desirability and feasibility (possibility) are better assessed by a
group than by any of its individual members, especially in systems
design where the users and those affected by the design are
multiple persons, a collectivity or a community. In such a case a
GDSS is probably a must for an adequate design, and a GAMR-based
GDSS would be highly desirable.
-
21
Although all four kinds of representations, briefly described
above, characterize design as well, the third and fourth are the
most related to it. Design is the combination of reproductive and
productive perceptions oriented to a future existence, and directed
by an intention. It is the imagination of what it is possible and
desirable according to an intention. It is to imagine what it could
exist by means of existing objects. It is a âpoieticâ (productive)
imagination, intentionally directed, and action oriented. From an
epistemological perspective, ârepresentationâ has been conceived by
the scholastics as âsimilitudeâ. âTo represent somethingâaccording
to Aquinasâis to contain the similitude of the thingâ(On the Truth
of the Catholic Faith; q.7, a.5.) To know is to represent an
existing object (of knowing) by a simile; it is to simulate an
existing object by a resembling idea. In this sense we could think
of âdesignâ as the knowledge of a pre-existence, a pre-figuring; it
is a âpre-knowledgeâ, a pre-cognition, of something that it is to
come. It is a âpre-knowingâ by means of what is known. When âwhat
is knownâ is distributed among different persons, the pre-knowing
is also distributed among them. So, no individual person should
take design decision in such a case, if an adequate, feasible,
right and just design is minded. Designers should care, not just
for doing the design right, but also for conceiving the right
design. Experience and knowledge in the respective domain is a must
for doing the design right, but group, or collective judgment and
decisions are what it is required for conceiving the right design.
In the late scholastics, the senses of âimagesâ and âmeaningâ were
added to the significance of ârepresentationâ. Descartes emphasized
on its sense of image and Kant generalized its meaning as to
signify: (1) any cognitive act or content no matter if they are
similitude of a knowing object or not; and (2) any non-private,
public structure, frames, models or scheme that is cause or effect
of such cognitive acts or contents. In this sense, the notion of
âdesignâ would refer to cognitive acts or contents, and/or their
public cause or effect, all of which are future-oriented,
representing non-existing physical objects. The epistemological
value of these private acts/contents and its public cause/effect
depend on the feasibility and desirability of their physical
existence and on the accompanied intention to make them come true.
It is evident here the epistemological value (and not just the
utilitarian one) of group, or collective, judgments and decisions.
Hence, it is clear the importance and the instrumental value of
GDSS (and GAMR-based GDSS) for the generation of epistemological
value in design processes. Kant differentiated between reproductive
and productive imagination (Critique of Pure Reason; A79/B104,
A123.) The role of the âproductive imaginationâ is not limited to
the pure reason; Kant extended its role to the practical reason and
to judgment (Critique of Judgment, 17). Since design is based on
judgment and practical reason, Kantâs differentiation is an
adequate one to use here. An analogous differentiation had been
made by Christian Wolff (Psychologia Empirica, Par. 92.)
Accordingly, âdesignâ could be conceived as a turnout of a
productive imagination, oriented to the future existence of an
object and, hence, accompanied by the intention that makes the
objectâs existence come true. It is a kind of âa priori synthesisâ
with the consorted intention to
-
22
bring it to physical existence. The productive imagination is a
process/product of image dividing and combining. Because what we
concluded above from Kant, this process/product dividing and
combining should be done by the group, or the collectivity, who is
the cause the cognitive design representation or who are receiving
the effects of such a representation or design. Again, it is
evident the importance and the instrumental value of GDSS, and of
GAMR-based GDSS, in designs affecting more persons than just the
designer(s). In the case of designing, our productive imagination
generates a mental âa priori synthesisâ, a mental image or
representation of a ânon-existing-yetâ physical object. This mental
representation might, in turn, be physically represented though a
drawing, a diagram, a visual schema, a material model, etc. This
physical representation is done in order to communicate the mental
image to other person(s), i.e. to cause a mental representation in
other(s). The physical representation is an effect of the original
mental image, and a cause of other mental images. Private mental
designs could be made public by means of their physical
representation. This would create mental representation in the
other person, according to which the design might be modified in a
process of participative design. In such a case, group judgments
and design decisions might be needed, which require GDSS,
especially GAMR-based GDSS. The starting point and the essence of
the design process is a mental one and, as such, it is necessarily
intentional. According to Brentano, mental (psychic) phenomena
possessâunlike the physical onesâintentionality, i.e. they refer to
an object. A perception is always a âperception of âsomethingâ, a
conscienceâas Husserl (1900) emphasizedâis always a âconscience of
âsomething. Mental phenomena, unlike physical ones, exist always in
the mind. This is why the scholastics called them âinexistenceâ
which should not be confused with ânon-existenceâ or absence of
existence. In its scholastic sense, âinexistenceâ means
âexistent-inâ other thing (Boudry, 1980). Brentano emphasized this
scholastic sense: âthis intentional existenceâhe wrote âis
exclusively characteristic of mental phenomena. No physical
phenomenon manifests anything similar. Consequently, we can define
mental phenomena by saying that they are such phenomena as to
include an object intentionally within themselvesâ (Bruce, 1967).
In this sense, design could be conceived as a pre-existent
intentional inexistence. 2.3. Design as Intention
The term âintentionâ refers to the act and the effect of tending
toward something. In this sense, with an etymological flavor, the
term intentio was defined by Aquinas (Summa Theologica; Ia-IIa,
qXII, a1). The notion of âintentionâ relates the knower with the
known, the perceiver with the perceived, i.e. the subject with the
object. This is why âintentionâ is a central notion in
phenomenology, which opposes strongly any kind of reductionism, and
any way of isolating the subject from the object. The
perceiver/knower is always perceiving/knowing something. To be a
perceiver/knower is to be related to what it is perceived/known. To
be a subject is to be necessarily related to an object; and to be
an object is to be related to a subject. There is neither an
isolated subject, as such, nor an isolated object as such.
-
23
The characteristic of âintentionâ, relating subject and object,
generates ambiguities in the meaning of the term, which sometimes
is used to refer to the subjectâs mental potential, act or content;
and other times is used to refer to the object or to the
circumstances or conditions. This equivocalness of âintentionâ has
been recognized since the scholastics (as, for example, in St.
Thomas Aquinas and St. Bonaventure), up to the present (Aune,
1967). Being the notions of design and intention so conjoined, it
is no surprise that the senses of the notion of design, identified
at the beginning of this paper, are analogous to the senses, which
seem to have been identified for the notion of intention. The
senses of process and product, we found for the notion of design,
seem to correlate with the notion of act and content; and mental
and physical design seem to correlate with subject and object.
Collaborative Design (frequently used in complex systems design) is
produced by Collective Intentionality which might be supported and
made explicit by Group Communication Support systems, in general,
and by GDSS, or more specifically, by GAMR-Delphi based GDSS.
Individual intentions, or purposes, form parts of a group or
collective intentionally. These individual intentions usually
differ from each other and, consequently, a consensus should be
identified with regards to the collective intention. The consensus
identification process requires sharing information about the
respective individual intentions, making individual decisions about
what should be the collective intention, identifying the collective
decision as input to the Delhi process required to increase the
consensus level, and to make the final collective, or group,
decision; which is a necessary step in the designing process. 2.4.
Design and Action
Important essentialities of the notion of intention are
constructive of the notion of design. The first of these
essentialities is the inherent disposition to action (Aune, 1967,
p. 198) of intention and, hence, of design. When there is no
disposition to action there could not be any intention, or design.
A desire might conflict with another desire and not be followed by
action, but intention and, hence, design generates action.
Intentions and desires are both pro-attitudes, butâas Bratman
emphasizedâjust intentions are conduct-controlling pro-attitudes.
Desires are potential influencers of action, while intentions are
actual influences of actions. (Bratman, 1987, p.16) Consequently, a
design is also a conduct-controlling pro-attitude. If a design
generates no action it is not a real design, it is a virtual one,
it is a desire which, when contrasted by other desires, refrained
from action, is an option not an intention; hence not a design. To
have an epistemological and practical value, design should generate
action that would produce the object designed; it should bring to
physical existence the pre-existent, the ânon-existent-yetâ
physical object of the design. Else, what is the use of design?
What is its reason of being? 2.5. Design and Practical
Reasoning
Not any action would bring a pre-existent intentional
inexistence (the design of a non-existent-yet physical object) into
existence. Just an adequate and an effective action would do it. To
do so, we need a practical reasoning. This is the second
essentiality of the notion of intention and, hence, of design.
Apart from having an intention in the âdisposition to actionâ
sense, âit is also possible to intend in an occurrent,
nondispositional sense -that is, to engage in âactsâ of
-
24
intending. This is possible because resolving is an âactâ that
counts as a special case of intending- namely, intending as an
immediate consequence of deliberation or choiceâ (Aune (1967), p.
200). Consequently, design is also an immediate consequence of
deliberation or choice. Group design would be requires group
deliberation and choice. A GDSS, as the one described above, would
be a very desirable â if not necessary â support to such a group
deliberation and choice. As a mere disposition to action
âintentions may form themselves as effortlessly and as
unconsciously as beliefs, which they resemble; but sometimes, as in
deliberation or choice, one forms an intention explicitly,
consciously, and occurrently -in which case oneâs intending may
have a character of a resolve... Here oneâs intending, as act, is a
âpracticalâ thought, serving as the conclusion of a line of
practical reasoningâ (Aune, 1967, p. 200). Consequently, Group
design requires Group Practical Thought, which, in turn, requires a
GDSS, or a GAMR-based GDSS, like the one described above. If our
actions were influenced by deliberation only at the time of action,
the influence of such deliberation would be rather minimal, since
deliberation requires time, effort and other limited resources, and
there is an obvious limits to the extent to which one could
successfully deliberate at the very time of action. Consequently,
we need some ways by which deliberation and rational reflection
will be allowed to influence action and to take place before the
actionâs time. Consequently, plans are a must for an opportune
deliberating and rational reflection, and a Generalized GDSS for
Group Problem solving, would be of a great help, for the generation
of these plans. 2.6. Design and Planning
Plans are also required for intra-personal and/or inter-personal
coordination. By constructing plans for the future, we facilitate
coordination in both our activities over time, and our activities
with the activities of others. By setting plans, we enable our
present deliberation and practical reasoning to influence our later
conduct, extending the influence of our deliberation beyond the
present moment and beside ourselves. As a design gets complex
requires us to go beyond the present and beside ourselves,
consequently, it will require plans. A plan (or various plans) is
(are) usually required for achieving the âpre-existent intentional
inexistenceâ, and a plan (or various plans) is (are) required for
bringing to existence the mental and/or the physical representation
of such âpre-existent intentional existenceâ. Butâas Bratman (1987)
assertedââwe do not, of course, promote coordination and extend the
influence of deliberation by means of plans that specify, once and
for all, everything we are to do in the future. Rather, we
typically settle on plans that are partial and then fill them in as
need be and as time goes by. This characteristic incompleteness of
our plans is of the first importance. It creates the need for a
kind of reasoning characteristic of planning agents: reasoning that
takes initial, partial plans as given and aims at filling them in
with specifications of appropriate means, preliminary steps, or
just relatively more specific courses of actionâ (emphasis added).
This continuous planning and re-planning requires a planning GDSS,
when it is related to complex system design and, hence, to group
decision.
-
25
There are several whys supporting Bratmanâs assertion about the
inherent partiality and incompleteness of plans, especially those
related to complex designs. Letâs enumerate briefly some
fundamental whys: 1. Experiments had shown that people have
perception processes which can handle between
5 and 9 (7+-2) things at once (Miller, 1956). Consequently,
complex situations should be handled by means of different levels
of abstraction, where details are not shown in the highest level of
abstraction. Consequently, the most abstract, or general plan would
not contain the details that will be filled in at lower levels of
abstraction. The general plan will not have the specificities of
the special plans forming parts of the general one. Then, the
general plan will be partial and incomplete, considered in
comparison with the specific ones.
2. Plans need time to be executed. The larger and the more
complex the plan is, the larger
the time required for its execution. And, the larger the
execution time, the larger the probability of modifications in the
initial conditions, and the larger the amount of new relevant
information that will emerge. Consequently, the larger and the more
complex the plan is, the larger the probability that such a plan
will be inadequate at some time in its execution process.
Thereupon, as the plan reaches further in the future, the
probability of change and new information will increase
(exponentially), and, hence, the details will be less relevant.
Consequently, the plan will be more partial and more incomplete, as
it protracts in the future.
3. In an empirical research, Braybrooke and Lindblom (1970)
found that executives and
policy makers, when facing complex problems, try to clarify and
plan with details just the next step, the next planning increment,
leaving the succeeding steps, or increments not so clear and so
detailed, i.e. leaving the following planning increments partial,
incomplete and even obscure.
4. Our experience in designing and implementing complex systems
(educational,
organizational and informational) evidenced the verisimilitude
and the applicability of Braybrooke and Lindblomâs conclusion, as
well as the appropriateness and the relevance of Bratmanâs
arguments. In fact, we have been developing a Methodology for
Systems Analysis and Synthesis, using Braybrooke and Lindblomâs
conclusions and Bratmanâs philosophical perspective among the
foundational bases of our methodological theory construction. We
have already done a general description of such a methodology
(Callaos, 1992a; Callaos and Callaos, 1991), different applications
in the area of educational systems design and implementations,
(see, for example, the design of the Latin-American School of
Statesmen and Executives: LSSE, Callaos, 1992), few applications in
organizational design and implementation, more than 130
applications in information systems analysis, synthesis and
implementation (see, for example, Callaos and Callaos 1992a), and
an application to the design and experimental implementation of
Total Quality Designing System (Callaos and Callaos 1992b).
-
26
2.6. Action-Design
We can conclude without any hesitation that when the designing
process is not simple, plain and facile, (1) it should be done with
successive partial and incomplete plans to be filled in along with
the design activities, as the process progress toward an accepted
pre-existent intentional inexistence, which could physically be
represented as a verbal model and/or a visual diagrammatic
maquette; and (2) the design should be an evolutionary one, and the
designing process should be accompanied from the earliest possible
stage with implementation actions, which will be conducted, in
turn, with successive partial and incomplete plans. In this way,
the design process and the implementing action will be interwoven,
interacting with each other,
with reciprocal loops of feedback and feedforward (figure 6).
The design of the Latin-American School of Statesmen and Executives
(Callaos, 1992b) is an example where details could be found with
regards to the application of the diagram of figure 6 to a very
specific case.
Figure 6
Design is always intentional and action-oriented. The essence of
design is to generate action in some direction and/or for some
creation/production. It should not be isolated from action since it
is strongly related to it. Both are parts of the same whole, both
are members of the same organically dynamic system. Design gives
direction and action gives propulsion to the whole. They are polar
opposites, and as such, they complement and require each other. So,
there is no way in separating them without deteriorating their
essence. Usually, design comes before and is input to material
action. But when we are dealing with a complex system, design and
action should be conducted concurrently, even though design will
initially start alone up till an initial design of the first
prototype, or archetype, of the wanted system is available. From
there on,
INCREMENTAL PLANNING OF THE IMPLEMENTATION
PROCESS
EVALUATION FOR
FEEDBACK AND FEEDFORWARD
PHYSICAL ACTION: IMPLEMENTATION
OF SUCCESSIVE DESIGNS
INCREMENTAL PLANNING OF THE
DESIGNING PROCESS
META-DESIGN
LEVEL
MENTAL ACTION (WITH PHYSICAL REPRESENTATION OF INTERMEDIATE
AND FINAL PRODUCTS)
INTERMEDIATE AND FINAL
DESIGNS (PHYSICAL
PHISYCAL SYSTEMS
IMPLEMENTATION
-
27
design and action should be interwoven, interacting with each
other, by means of reciprocal loops of feedback and feedforward, in
an evolutionary process that could be called action-design, and
which is to be nurtured by the ingredients of action-research and
action-learning. In the case of collaborative design, decisions
with regards to which actions need to be taken should be group, or
collective, decisions. Consequently, GDSS, or more specifically,
GAMR-Delphi GDSS, would effectively support the collective decision
making process.
3. Design as Discovery4
In this section we will try to present a brief synopsis of some
key principles, concepts, and models of system design, presented in
the context title of "design as discovery". The emergence of the
ideas into the present form has occurred while Richard Evan, one of
the co-authors of this paper, being privileged to work with IBM and
NASA over the past four years. (Evans, 2001) But, while there is an
enormous debt to every individual in those organizations for their
generous and insightful participation, it must be stressed that the
thoughts here expressed do not in any way whatever carry any
affiliation, concurrence, or endorsement, etc. of either those
institutions nor any other person. The opportunity to work with
others is only mentioned here to express more focused appreciation.
For similar reasons, while over five hundred texts have contributed
to this work [section 3 of this paper], there are no references
cited in the initial paper where are integrating in this one. The
aim is to minimize the risk, prior to very careful reviews, of
implying any concurrence in these principles, concepts, and models
by any author(s). Subsequent papers will both cite the many
applicable references as well as treat each topic in more
appropriate detail. The work began as sessions in "System
Requirements", and that led to the initial perception that the very
verb "require" and all the attendant implications and applications
of the noun "requirements [as well as requirements engineering,
etc.] was worth reconsideration. Thus the proposal that the very
notion of a whole set of statements, sentences, etc. being forever
separate in nature, purpose, substance etc. as "requirements", from
another set called "design" decisions, is not valid, and thus a
very seriously false premise as that concept is so ubiquitous. To
generate a requirement set require decision to be taken, and these
are part of the design decision, and as such, if they should be
taken be a group, a GDSS, as the one described in the first part of
this paper, would be a very good support for the design process and
its inherent decision making process. Concomitant with the review
of "requirements" per se came the perception that the typical means
for developing design decisions [including design reviews,
independent reviews, Red Teams, etc.] might also be seriously
flawed.
4 The content of this section is the same paper that Professor
Evans presented at the 4th World Multi-conference on Systemics,
cybernetics, and Informatics, which was included in the respective
proceedings (Evans, 2001). Some text were added in order to link
the content of this section to previous two sections of this
paper.
-
28
Design as discovery has also been found to rely in a major way
on the united operation of the three roles [relationships] of
Customer, Builder, and Associate. Likewise, it is suggested that
there are only these three roles, whether for person-to-person or
organization-to-organization relationships. It has further been
seen that design is thinking, and that structured thinking,
specifically three dimensions or 3-D thinking is essential in
enabling the critical expansion and awareness of all dimensions of
design. An example is the three dimensions of systems themselves,
namely the Discovery and Delivery systems that are over and above
the typical one-dimensional idea of just the design of a Delivered
system. Several other design-as-discovery topics are also presented
in summary form in the remainder of the paper. 3.1. Principles,
Concepts, Models
The structure of system design Principles, Concepts, and Models
address, as depicted in the figure, an expansion on the early âtask
or project-centeredâ work of leaders such as Taylor and Galbraith,
and then W. Edwards Demingâs contribution to recognize the need for
more than just optimum projects or tasks--namely for processes. The
recognition is that, even more than processes, the need is to
address the principles, concepts, and models for the application of
processes. Their effect and essentiality might be seen to be
equivalent to the impact of a presumption of innocence versus
guilty on the same judicial processes.
Projects [Tasks]
Processes
Principles
Principles are considered to be accepted truths, judgments,
policies, values, etc. Concepts are means or ways to apply the
principles. Models are structures for the implementation of the
concepts. An example principle is, as with presumed innocent versus
guilty: invitation, nomination, and confirmation rather than
imposition, compulsion, and unilateral investigation. An example of
a concept or "way" is three-dimensional thinking [3-d Thinking]
based on asking the Great Question of "What might be at least three
dimensions of this." An example of model or structure is the
Customer, Builder, Associate model [ABC Model] for the operating
structure of basic three roles/relationships.
-
29
3.2. Ideas, Options, Assessment
Design as discovery is the care and feeding of ideas: both as
options and also their comprehensive assessment. Design is founded
on options, and thus on the quality of the "source" or headwaters
of ideas/options and their assessment is the critical dimension.
System design has as a prerequisite the richest feasible set of
options and their comprehensive assessment. And assessment is
itself dependent on the options considered: optional perspectives,
optional criteria, optional metrics, and etc. Further, as all
design is decision under uncertainty, it is risk-based design and
predicated on the effectiveness of the exploration, the search, the
inquiry, in short: the discovery. Design as discovery is
inaugurated and conceived in the nurturing of ideas: ideas about
the situation; the diagnoses of the existing and desired situation
[state] to identify contributing problems and the
interrelationships of those problems; ideas about possible ways to
address the problems; ideas about assessment of these activities
and results, ideas about the implementation of the selected
solutions; and even ideas about the assessment activity itself.
Ideas [the headwaters of design] are initially as tiny and
essentially hidden and inconspicuous as seeds: as in the phrase the
"germ" of an idea. They are similarly perishable and in need of the
most precise nurture and care to even germinate, let alone mature.
Also, as with seeds, a particular challenge is [as with all
exploration/discovery] to discern the desired from the
inappropriate--the wheat from the tares. Often all seem equally
promising and fitting. The nurture and maturation sufficient to
enable the discrimination and selection of preferred ideas demands
the greatest care: idea greenhouses, thought conservatories,
inspiration nurseries. Design as discovery calls for the design
itself of climate-controlled environments to enable options and
possibilities and alternate views, etc. to be safely and securely
conceived and then be strengthened as they germinate, emerge,
and--when seen to be possibly more on the side of good--continue to
grow. This is the essence of system design: The care and feeding of
ideas. 3.3. Invitation, Nomination, Confirmation
Confirm, never compel or "require" A foundation principle for
the achievement of design ideas is to invite never impose. It is to
introduce and then enrich other's nominations by suggestions,
explanations, persuasion--never by compelling. It is to apply
so-called "independent" resources as ones to show, to teach, to
illustrate, to encourage. It is to help identify optional features
for consideration in self-nominated plans, procedures, design
decision options, assessment plans, etc.--plans and decisions to
then be confirmed. The one to whom delegated fully responsible to
nominate, the delegate fully responsible in confirming. Thus both
united--equally yoked in a common endeavor--one to nominate, the
other to confirm.
-
30
3.4. Teams vs. Teamwork
The term "team" has become applied to formal organizational
entities--with team leaders etc. But this has confused the spirit
of a team with the term team. The spirit of a team is the spirit of
teamwork, that exists by virtue of a common goal, not because of a
common boss. A basketball team on the floor has no single boss.
That form of teamwork is crucial in the formative efforts of option
generation, and thus needs formally organized [the coach controls
who is on the court] but informally operated efforts. The suggested
structure is tables of three. To achieve ideal option generation
[formally organized but informally operated effort] there is never
a âteam leaderâ. A GDSS, and more specifically a GAMR-based GDSS as
the one described in the first section of this paper is a very
useful support for this non-leader system design process. 3.5.
Design Decisions
Decisions are by a single designated individual. âGroupsâ are
only for the generation of options and their assessment A single
responsible individual is the one responsible, most of all, for the
formal organization of the informally operated âteamâ efforts. A
GDSS, as the one described in the first section of this paper,
would support group options generation, via electronic
brainstorming, ideas writing, Nominal Group Technique, etc., and
group assessment via collective judgment (by way of individual
judgments synthesis) If for some reason, some design decision
should be taken by the group, as a group, then, the GAMR-based GDSS
described above will be almost a must for this kind of decisions in
a complex systems design situations. Never a designated âteam leadâ
etc., thus all members feel equally responsible to each other and
for each otherâand for the options and assessments. Design
decisions [that then serve as "requirements" on all subsequent
decisions] address, as shown, the three dimensions of Ratables
[Measurables], Relationships [with all other decisions, etc.] and
the Rationale--to enable not only assurance of full compliance by
all subsequent decisions, but to provide the essential framework
for those successive decisions.
Decisions
Relationships
Ratables
Rationale
There is a hierarchy of decisions, but it is by timing
sequence--a temporal hierarchy--not be so-called system levels,
such as top-level decisions.
-
31
3.6. Three Roles: Customer, Builder, Associate Design as
discovery involves, as illustrated in the figure, only three roles
[relationships], whether person-to-person or
organization-to-organization: Customer, Builder, Associate. Design
relies on the integrity of the formally organized yet informally
operated headwaters of option generation and assessment, effected
in the framework of a united operation of the three [and the only
three] roles: Customer, Builder, and Associate--the ABC
Model--applying Invitation, Nomination, and Confirmation, and ever
asking the Great Question. This approach of formally organized yet
informally operated headwaters of option generation and assessment,
would be very facilitated with an information system support like
the GDSS (or GAMR-based GDSS) described in the first section of
this paper.
Customer
BuilderNomination
Formally organized-- but Informally Operated ]
Confirmation
Associate
Option Generation and Assessment efforts
Headwaters--Design Foundation
Design is dependent on the united accountability of those in
Builder and Customer roles. The ideal is for them to be equally
yoked--united in the pulling, as depicted, of their wagon:
Equally Wagon
Customer
Builder
Yoked
Those in Builder roles are accountable for design options
nominated [with recommendations] for their system. The Customer is
accountable for the confirmation of the design of their system.
Those in Builder roles "Conduct", those in Customer roles
"Preside". Both seek the wisdom of the other--the Customer seeks
the nominations of the Builder, the Builder seeks the confirmation
of the Customer, and both seek the insights of those in an
Associate role. This approach is a very important one for systems
design effectiveness. Several co-authors of this paper, experienced
the pragmatic value of this approach while designing more than 100
information systems. Effectiveness of the design process as well as
the effectiveness of the system designed (after its deployment)
depends highly on the âbuilder-conduct-customer âpresideâ approach.
In complex systems design it is frequent to find situations where
the builder is not just one person, and the customer, or the user,
is neither a person, but in both cases it is frequent to find a
builder group and a customer or a user group of person. In such
casesâ a GDSS would of a great help in the
-
32
respective systems design processes. Those in Associate roles
are altruistically dedicated to the success of the other two;
having no authority in the relationship they have the greatest
potential influence. Design is equally dependent on the integrity
of the operation [as the design headwaters] of the option
generation and assessment effort that is formally organized and
informally-operated by those in Builder roles. 3.7. Delegation:
[Contract for] Design Work, Not for "A" Design Design as discovery
is contracted for as design work, not for "a" design. Customers
select [delegate to] those to serve in their Builder roles based on
their design abilities, not on their proposed design. Contract
award criteria include past performance and current capability to
provide design work, not an actual complete design--prepared in
isolation--as the basis for the contract competition. 3.8.
Self-Assessments
Design as discovery concentrates on self--especially on Builder
"self-assessments" that are builder-nominated and
customer-confirmed: every designer a designer of systems--every
engineer an engineer of systems. Group self-assessment requires the
kind of support that could be provided by the GDSS for Group
Problem Solving described in the first part of this paper.
3.9. Meeting Purpose Purity
Effectiveness in design meetings is dependent on their purity of
purpose. It is suggested that there are three orthogonal design
"meeting" dimensions--each a dimension of purpose: Preparation,
Presentation, Confirmation.
Meeting Purpose
Confirmation
Presentation
Preparation Preparation "meetings" are formally organized but
informally operated [no boss--and ideally sets of only three
participants. These activities are the most crucial: they are the
headwaters, the birthplaces, the nurseries and greenhouses for the
generation and assessment of design options. It is this activity
that needs to be re-enthroned. Presentations are by information
briefings, lectures, conferences, and etc.
-
33
Confirmation meetings as the third an