•  
  •  
 

Abstract

We consider group decision-making on an optimal stopping problem, for which large and stable individual differences have previously been established. In the problem, people are presented with a sequence of five random numbers between 0 and 100, one at a time, and are required to choose the maximum of the sequence, without being allowed to return to earlier values in the sequence. We examine group decision-making on these problems in an experimental setting where group members are isolated from one another, and interact solely via networked computers. The group members register their initial accept or reject decision for each value in the sequence, and then providing a potentially revised decision having viewed the recommendations of the other group members. Group decisions are made according to one of three conditions, requiring either consensus to accept from all group members, a majority of accept decisions from the group, or the acceptance of an appointed group leader. We compare individual decision-making to group decision-making under these three conditions, and find that, under some conditions, groups often significantly outperform even their best members. Using a signal detection analysis we provide an account of how the group decision-making conditions differ from one another, and from individual decision-making. Key findings are that people do not often revise their decisions, but, in the consensus and leadership conditions, are more conservative in their initial decisions. This conservatism removes the individual bias towards choosing values too early in the sequence, allowing the groups to perform better than their individual members. In the majority condition, however, people continue to behave as they did individually, and the group shows the same bias in decision-making.

Download

Share

COinS