Studying the Impact of Minority Views in a Computational Model of Collective Sensemaking: The Role of Network Structure

Abstract—A series of experiments were performed in order to explore the effect of communication network structure on collective sensemaking under a variety of informational conditions. A multi-agent computational model of collective sensemaking was used in which each agent was implemented as a constraint satisfaction network. Within the simulations, agents were tasked with the interpretation of information indicating the presence of a particular object, and they were al-lowed to share information with other agents while performing this task subject to the constraints imposed by the structure of a communication network. In all simulations, a minority of agents (5) received evidence in favor of one interpretation, while a majority of agents (15) received evidence in favor of a conflicting interpretation. Communication networks with four types of topological structure (i.e., disconnected, random, small-world and fully-connected) were used in the experiments. The results suggest that network topology influences the extent to which minority views are able to influence collective cognitive outcomes. In particular, fully-connected networks deliver a performance profile in which minority influence is minimized in situations where both minority and majority groups are exposed to weak evidence. However, the same networks serve to maximize minority influence when minority group members are selectively exposed to strong evidence. These results suggest that fully-connected networks differentially regulate minority influence based on the kinds of evidence presented to both minority and majority group members. Keywords-sensemaking; distributed cognition; social influ-ence; network science; social information processing. I