Micro-foundations of Groupthink – A Lab Experiment

The notion of “groupthink” – coined by Irving Janis – has been identified by both academics and policymakers as a central cause of major institutional and organizational failures. The main behavioral claim underlying... [ view full abstract ]

The notion of “groupthink” – coined by Irving Janis – has been identified by both academics and policymakers as a central cause of major institutional and organizational failures. The main behavioral claim underlying groupthink is the tendency of group members to self-censor views that threaten the group’s consensus, leading to sub-optimal performance, and sometime major organizational failures. However, while many case studies identify elements of groupthink as antecedents of poor decisions, and experimental studies provide mixed support for the model’s predictions, no study has so far demonstrated the causal effect of group cohesiveness on group performance. Moreover, despite the fact that a central feature of the historical cases that provided inspiration for the theory (e.g. The Pearl Harbor attack) is the failure of groups to address changes in their environments, this particular element is absent from previous attempts to test the theory experimentally.
To test these theoretical relationships we rely on two laboratory experiments (N=240). Participants were tasked with an experience-based decision making assignment (Erev & Roth 2014). Such assignments provide environments whose structure is not explicit nor directly observable, thus decision makers can learn about the properties of their options only by experiencing the outcomes of their choices. Participants were assigned to one of three conditions: (1) Individual (N=30); (2) Group (N=90: 30 groups of 3); and (3) Cohesive-group (N=90: 30 groups of 3). Participants played a computer game that consisted of 100 choices between two alternatives presented as two unlabeled buttons on a computer screen. On each trial, participants clicked on one of the two buttons and either won or lost points. After clicking, participants saw the outcomes of their choices. Their objective was to maximize the number of points obtained over the 100 trials. The game had two stages (unknown by the participants): a stage where one option dominated the other and a stage where the dominant option switched to be the other one. These two stages simulate a stable environment (rounds 1-60), and a dynamic environment (round 61 onward). All the participants performed the experiment on their own computer screen, without seeing any of the other participants. Participants in the group condition were informed that they are part of a 3-member group, and after they made their individual choice, the actual decision was a group-decision (majority). The cohesive-group condition was the same as in "group" with the addition that participants whose individual choice was different from the subsequent group decision (i.e. decenters) incurred a small deduction from their payoff.
Our results show that cohesive-groups performed slightly better than individuals and groups in a stable environment, but were markedly slower in recovering from a change in the environment compared to both individuals and groups. These results provide direct support for the effect of group cohesiveness on group performance, and further specifies its detrimental consequences as applying to dynamic environments. The second experiment (N=30) addressed a competing explanation, and the paper further discuss the implications of the findings to public administration.