System Dynamics Understanding: Part II - Connectivity-groupthink or heightened completeness

Connectivity-groupthink or heightened completeness?

Complex systems-thinking impacts on SCM system implementation success

Dr. Elliot BendolyGoizueta Business School

Emory UniversityAtlanta, GA

Bendoly - Connectivity-groupthink or heightened completeness?

Distinct… But distinctions that have seldom been made in the OM literature. As normative researchers pursue ways to incorporate human behavior into models, this lack of distinction may prove misleading.

Accepting information from others (sharing information).

Accepting the conclusions drawn from that information (drawn by others). In some cases ceding interpretation to others (groupthink).

f(a, b) f(a, b)Accepting approaches to thinking about information and drawing conclusions (e.g. sharing similar systems views on the nature of information and connections).

(Some) Forms of sharing in team problem solving


Fortunately recent methods for measuring an individual’s strength in a particular approach to thinking have been applied to empirically demonstrate its value above and beyond other forms of expertise (Bendoly 2013).

Specifically the extent to which individuals share an understanding of system dynamics principles has been shown to yield performance benefits in team project settings. Competing against this view are recent normative models such as that of LiCalzi and Surucu (2012) in which such a lack of diversity contributes to phenomena such as groupthink, thus limiting the ability of a team to success in project work.

Bendoly 2013LiCalzi and Surucu 2012


LiCalzi and Surucu (2012) start with the assumption that given a problem (or project) assigned to a team of workers, the graph G(p) represents the landscape of possible solutions (eg. implementation tactics) individuals on that team may be aware of. The term “p” represents the probability that an individual fully aware of one of these solutions, and its likely performance implications, is simultaneously aware of another of the solutions on this graph. It is essentially a “completeness-likelihood”, with regard to connections made across the space of the graph. Clearly the larger the number of possible solutions, which is often massive for many real world team problems, the more complex the graph and the less likely (given 0 < p < 1) any one individual would be able to come up with the best solution on his/her own. A greater completeness-likelihood (p) certainly helps, but for highly complex problems having a sufficiently high value of would also seem increasingly hard to come by (LiCalzi and Surucu 2012). This is where the power of the team can come in… sometimes, as LiCalzi and Surucu argue (2012).


Aside from scale (eg. group size, m), the two main factors which impact the likelihood of team success in the LiCalzi-Surucu model include the graph completeness-likelihood (p), and what we can refer to as the level of connectivity-groupthink (d). As implied, this is fairly specific form of groupthink. In a sense it represents a correlation in p across team members. According to LiCalzi and Surucu (2012), as the strength of this groupthink increases across team members, each member’s awareness of possible solutions becomes increasingly similar (John only knows of the solutions 5, 12 and 13; his colleagues Kim and Andrew also know of 5 and very likely also know of 12 and 13). Based on recent discussions of barriers in group dynamics such as apprehension and production blocking, it is also possible for connectivity-groupthink to represent the tendency for team members to inhibit ideas from entering the team’s set of solution (Kavadias and Sommer 2009). At an extreme (d=1), according to the argument of LiCalzi and Surucu (2012), no amount of team growth will increase its ability to succeed. That is, the team’s awareness of solutions is no better than any one member thinking about tactics independently, regardless of how large the team. More moderate levels of d>0 allow some benefits of group size to emerge.


Extending the 4th theorem of LiCalzi and Surucu (2012) to a solution space of extensive

size (complex SCM implementations), the likelihood of team success can be captured by

the following equation:

(1)

Limits: 0 and 1, given the fractional values attributed to the factors d and p.

Main Question: Given team size m, how to capture the latent concept of p and d to best

capture empirically observed rates of team success?

Secondary Question: If p and d are latent, can they nevertheless be assessed as functions

team attributes we observe? If they are jointly dependent on certain attributes, what would

that mean for managerial prescription?


Figure 1. General Predictive Form of the LiCalzi-Surucu Model, with m=5


Table 1. Possible roles of shared system dynamics understanding (SSDU) in the LiCalzi-Surucu model

H1: Positive b on SSDU

MODEL 1: No Impact

SSDU does not heighten graph completeness-likelihood among team members, but also does not encourage connectivity-groupthink

MODEL 2 : Only Costs SSDU does not heighten graph completeness-likelihood among team members, but encourages connectivity-groupthink

H2: Positive b on SSDU

MODEL 3 : Only Benefits SSDU heightens graph completeness-likelihood among team members, but does not encourage connectivity-groupthink

MODEL 4 : Mixed Effects SSDU heightens graph completeness-likelihood among team members, but also encourages connectivity-groupthink


chosen population for examination in this study involved collocated teams charged with SCM software implementation projects, as extensions to pre-existing enterprise system back-ends. The software developer, one of the largest vendors in that global market space, provided company contact information on 5,514 of its clients. Starting with this data, and through coordinated efforts with the leading members of the Supply Chain Council, APICS, and PMI, we obtained contact information for project team leaders of 4,175 implementations carried out in the US between 2003-2012. These team leaders played a critical role in both serving as the source of objective data on the nature of these projects as well as a source for additional contact information for the team members. Through those leaders access to a total of 856 (20.5%) possible project data points was made possible. The team member contacts associated with these projects included 5,754 individualsto ensure faithful representation of team dynamics within projects, only a subset of interview and survey data would be used in the final analysis. To strengthen the interpretability of the final sample, we applied a minimum project inclusion criterion of 50% team representation, similarly employed by Haas (2006) and based on Hackman (2002). Among those represented by at least one interviewed team member, 279 projects met this criterion. This final number represented 32.6% of the original 856 prospective projects in the target sample. The final count of project team members used in analysis was 1,344 (23.4% of the original team member list provided by the team leaders).


In order to convert from the individual ISDU scores to a team-wide shared system

dynamics understanding (SSDU) score, a simple Euclidean distance approach was adopted

(c.f. Harrison and Klein 2007; Tsui et al. 1992).

In the above, I is the total number of team-members for which data was available. The

denominator scales the calculation to account for the number of comparisons made for

each team member (I-1) and the maximum score on the S-BI protocol (MaxISDU=56).

The result is a variable that ranges from 1 (perfect similarity) to a minimum level of 0

(maximal distinctness).


Table 2. NLMIXED Maximum Likelihood results for Schedule Adherence and Estimated Antecedent Structures for Variants of the LiCalzi-Surucu Model


Figure 2. Estimated Effects of Shared System Dynamics Understanding (Model 4)

Very initially completeness-likelihood (p) gains from increased SSDU appear to offset groupthink.

As SSDU increases, groupthink losses begin to match completeness-likelihood gains (those still at the lower end of the ISCU scale acquiesce?). Further on, individuals are more inclined to share information and ideas (not just adopt those of others), increasing typical completeness-likelihoods held by team members, dominating the detractions of groupthink for additional gains in SSDU.


Implications for Research

Similarity in System Dynamics Understanding:

• Proves a relevant predictor of theorized latent variable effects of connectivity-groupthink (d) and graph completeness-likelihood (p) [Absence of SSDU in predictive models reduces model fit by more than 40% of Ngl-R2 0.410]

• The strength of the SSDU effect is only realized in the context of the LiCalzi-Surucu model, and appears appreciably reduced in less sophisticated models (eg. simple Logistic) that pick up less fit to empirical observation (Ngl-R2 0.338).

• Use of sophisticated normative approaches (LiCalzi-Surucu model) in conjunction with novel and relevant empirical metrics (SSDU) prove useful --- suggest a solid path for future related investigation.


Implications for Practice

• Cause to promote not only individual-level SDU but also SSDU within groups. Training programs, reassessments incorporated in group design, etc.

• Temper expectations of direct benefits: Incremental adjustments to SSDU for teams already in a performance inelastic region may see no benefit until completeness-likelihood gains once again outstrip connectivity-groupthink losses.

• Recast managerial views of groupthink in general. If it turns out that groupthink is often accompanied by other phenomena that provide net gains to performance, efforts to minimize groupthink (eg. limiting certain kinds of team interactions) may be counter-effective. Context-specific.


Questions?


APPENDIX: Team Level Descriptive Statistics and Inter-correlations


APPENDIX: Non-LiCalzi-Surucu Benchmark Model: Logistic Regression


APPENDIX: Adjustments to S-BI Protocol used Scenario B:

Scenario B: Project Leader Perception / Project Team Worker Self-Perception SAY: Okay, let’s now look at this issue: The way a project team worker sees/perceives him or herself, and the way a project leader sees/perceived that worker.LEVEL 1: Begin by asking a very open-ended question:SAY: “I’d like you to tell me one possible story about the relationship between (1) the way a project team worker sees themselves (in terms of how good or not they are as project workers in the team context) and (2) the project leader’s perception of that worker (or how the project leader sees that worker)”LEVEL 2: SAY: “How might these two issues be interrelated? What do you think happens to both the project worker’s and the project leader’s perceptions over time? Can you draw that (e.g. in a object-arrow diagram form)?LEVEL 3: (Unless these issues are already elaborated on by participant in previous points, continue…)SAY: “What happens next?” or, “So, if a project leader thinks a project group member is not a good worker, what may happen to the worker’s view of his/her own abilities?” “How long do you think it takes for this change in perception to occur?”LEVEL 4: SAY: “Can you think of other situations that feel the same as this (improving or worsening over time) behavior?” Scenario F: Red Flags / Fixes / Team Leader Attitude

SAY: Does the project leader often ask you to fix work done on the project (e.g. red-flagged items)?LEVEL 1: SAY: “Tell me a story about a possible relationship between the quality of progress on a project (how many red-flags project work seems to give rise to) and the level of happiness of the project team leader.”LEVEL 2: SAY: “How might the (1) total number of project red-flags and (2) your project leader’s attitude be connected? What do you think happens to these two issues over time?”LEVEL 3: (Unless these issues are already elaborated on by participant in previous points, continue…)SAY: “What happens next?” or, “So, if a lot of red-flags are raised, then what happens?” (note, first response may be something like “Project leader gets upset”). “Okay, So what happens next?”LEVEL 4: SAY: “Can you think of other situations that feel the same as this? Can you think of examples that have this same kind of up and down behavior?”

System Dynamics Understanding: Part II - Connectivity-groupthink or heightened completeness

Business

connectivitygroupthink

bendoly connectivity

team members

heightened completeness

graph completenesslikelihood

likelihood of team success

team of workers

team growth