![Page 1: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/1.jpg)
Groupthink: Theory and Evidence
ETHZ, Game Theory and Society, July 27-30, 2011
Christopher Baker Harvard University
Hanja Blendin & Gerald Schneider
Universität Konstanz
![Page 2: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/2.jpg)
Motivation Thomas C. Schelling 2006, An
astonishing 60 years: The legacy of Hiroshima [Nobel Prize Lecture]: “The most spectacular event of the past half century is one that did not occur. We have enjoyed 60 years
without nuclear weapons exploded in anger. ...we may come to a new
respect for deterrence”
![Page 3: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/3.jpg)
What is groupthink?
![Page 4: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/4.jpg)
The Decision to Invade Iraq Senator Pat Roberts, Chairman of the U.S. Senate Intelligence Committee, “…the intelligence community was suffering from what we call a collective groupthink”
Source: F. Zagare 2004. Reconciling Rationality with Deterrence. Journal of Theoretical Politics
C=Conceding D=Demand
![Page 5: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/5.jpg)
• Definition of Groupthink: „A mode of thinking that people engage in when they are deeply involved in a cohesive in-group, when the members' strivings for unanimity override their motivation to realistically appraise alternative courses of action.“ (Janis 1972, 9)
• Problems:
![Page 6: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/6.jpg)
![Page 7: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/7.jpg)
Problems - Many citations (>2500), few recent experimental tests - Limited micro-foundations Our approach - Formal model of group- think - Focus on stress and cohesion
Source: Parks (2000)
![Page 8: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/8.jpg)
Antecedents Black box Decision Disaster
Stress
Group cohesion
Deflated self- confidence
Concurrence seeking
Reduced decision quality
![Page 9: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/9.jpg)
The theoretical foundations: an adapted version of the Condorcet Jury Voting
Model Condorcet Jury Theorem: If the probability of “voting” for the right decision exceeds 0.5, then larger groups make a more correct version than smaller ones. Austen-Smith/Banks (1996) and Feddersen/Pesendorfer (1998) show how strategic voting undermines this optimism. We use the latter model to show some conditions under which „irrational believes“ lead to concurrence seeking and poor decisions.
![Page 10: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/10.jpg)
Crisis cabinet of a country (Janistan) has to decide if to escalate in a conflict or not
Correct decision depends on the opponent (Whyteland)
q = members‘ escalation threshold Probability that signal is correct: Probability that signal is wrong:
Pr(si = h H ) = Pr(si = h H ) = c
Pr(si = h H ) = Pr(si = h H ) = 1− c
Outcome(decision = escalate | Whyteland = hostile) = 0O(e | H ) = −qO(e | H ) = 0O(e | H ) = −(1− q)
![Page 11: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/11.jpg)
Assumptions • -1- ministers have identical competence levels • -2- ministers update their beliefs based on information of the others • -3- prime minister casts his/her vote first, the others then simultaneously • -4- prime minister votes informatively, this is common knowledge
Equilibria • unique symmetric response equilibrium: i votes informatively, when
• pessimistic mixed strategy for
• optimistic mixed strategy for
β(k̂ −1,n) ≤ q < β(k̂,n)
0 < σ (si = h ) < σ (si = h) = 1
0 = σ (si = h ) < σ (si = h) < 1
![Page 12: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/12.jpg)
Variation 1: expected competence of leader exceeds actual competence
c1 > ci≠1
Equilibrium ranges between: always copying the prime minister‘s vote and the equilibrium of the base scenario
Variation 2: Ministers are under-confident
c > ci
The smaller ci the more a minister tends to ignore the own signal and to copy the prime minister‘s vote
![Page 13: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/13.jpg)
Experiment • Conducted 2010 in the lakelab of the University
of Konstanz (programmed in z-tree) • 104 subjects of all faculties, mainly males • 2(3) treatments: time pressure, cohesion • 3 dependent variables
- concurrence seeking (change of opinion after decision) - self-confidence after knowledge test - „wrong“ decisions (100 balls in jar, guessing the dominant color)
![Page 14: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/14.jpg)
Of which color are there more balls in the jar?
![Page 15: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/15.jpg)
• Time pressure increases concurrence-seeking • Cohesion (in the form of building team and
participation in rock-scissor game) decreases self-confidence
Table 1: Influence of treatments on concurrence-seeking (CS) and self-confidence (SC) CS (1) CS(2) CS(3) SC (4) SC (5) SC (6)
Time pressure 1.8* (0.78)
3.33* (2.38)
0.61 (0.26)
1.49 (1.29)
Cohesion 1.23 (0.54)
2.14 (1.39)
0.24*** (0.11)
0.44 (0.26)
Cohesion and time pressure
0.38 (0.35)
0.17* (0.17)
Log-Likelihood -62.45 -63.24 -61.66 -67.02 -62.42 -59.42 % correctly predicted 70.19 70.19 70.19 64.42 64.42 72.12 Notes: N=104. Coefficients are odds ratios, standard error in parentheses. * p < 0.10, ** p < 0.05, *** p<0.01
![Page 16: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/16.jpg)
• Time pressure and self-confidence decrease decision making quality
• Cohesion increases quality of decisions Table 2: Influence of treatments, of concurrence seeking and of self-confidence on decisio quality (DQ) (1) ( 2) (3) (4) (5)
Time pressure 0.40** (0.18)
2.05 (1.41)
0.15*** (0.10)
0.06*** (0.05)
Cohesion 1.78 (0.76)
4.64** (2.76)
Cohesion and time pressure
0.04*** (0.05)
Concurrence seeking
0.86 (0.319
0.36 (0.23)
Concurrence seeking and time pressure
13.49** (13.98)
Self-confidence
0.56 (0.24)
0.18*** (0.12)
Self-confidence and time pressure
18.26*** (19.31)
Log-Likelihood -64.90 -66.16 -59.26 -67.03 -61.50 -66.15 -59.21 % correctly predicted 65.38 65.38 69.23 65.38 65.38 65.38 73.08
![Page 17: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/17.jpg)
• BBS version of CJT provides one mechanism through which groupthink and the consequences of groupthink might be explained
• Time pressure increases concurrence-seeking and decreases decision making quality
• Concurrence seeking results more ambiguous: Cohesion increases self-confidence and not confidence in group, increased self-confidence lowers decision making quality
![Page 18: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/18.jpg)
Appendix β(k,n) = ck (1− c)n−k
ck (1− c)n−k + cn−k (1− c)k
i always votes informatively when:
β(k̂ −1,n) ≤ q < β(k̂,n)
![Page 19: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/19.jpg)
Pessimistic mixed strategy
Unique symmetric equilibrium in mixed strategies such that: σ1(h) = σ i≠1(v1 = e, s1 = h) = σ i≠1(v1 = e , s1 = h) = 1
σ1(h ) = σ i≠1(v1 = e, s1 = h ) = σ i≠1(v1 = e , s1 = h ) = v *
The probability that i votes for „escalate“ is between v* and 1
0 < v* =%c(1+ A
1k−1 ) −1
%c − A1
k−1 (1− %c)< 1 , with
A =
(1− q)(1− %c)n− k+1
q%cn− k+1 Pessimistic strategy
(1− %c)n− k̂+1
(1− %c)n− k̂+1 + %cn− k̂+1< q < β(k̂ −1,n)
![Page 20: Groupthink: Theory and Evidence - ETH Zürich · Groupthink: Theory and Evidence. ETHZ, Game Theory and Society, July 27- 30, 2011. Christopher Baker . Harvard University . Hanja](https://reader030.vdocuments.net/reader030/viewer/2022040311/5d506a1988c993b17e8b63f8/html5/thumbnails/20.jpg)
Mixed strategy
Unique symmetric equilibrium in mixed strategies such that:
The probability that i chooses „escalate“ is between 0 and w*
Optimistic strategy
β(k̂,n) < q <ck
ck + (1− c)k
σ1(h) = σ i≠1(v1 = e, s1 = h) = σ i≠1(v1 = e , s1 = h) = w *
σ1(h ) = σ i≠1(v1 = e, s1 = h ) = σ i≠1(v1 = e , s1 = h ) = 0
0 = σ (si = h ) < σ (si = h) < 1