Influence Structures and Information Aggregation in Groups

Group decision-making in organizations often occurs in the context of influence relationships. We develop a theory anchored in a computational model to analyze the conditions under which social influence structures aid or hurt information aggregation. We find that a group is most likely to achieve high accuracy through information aggregation when its overall influence structure (which at the dyadic level may have both informational and normative effects) resembles a single non-clustered component. When clustering is low while preserving a chain of connectivity throughout the group such that there are few isolates, information distributed in the group can be aggregated effectively. This “wisdom of communities” through social influence can be superior to the “wisdom of crowds” obtained by pooling individual estimates. We also discuss why laboratory studies may systematically underestimate the ability of groups with social influence to obtain accurate aggregate information, as well as some interventions to improve group accuracy by manipulating the influence structure indirectly via group composition