An agent-based model of consensus building

Abstract

Our model, CollAct is built around the question how people gain a shared understanding and reach consensus in an interactive group setting. This is an important question which is rather difficult to analyze within case studies. We model agents in a cognitive way, including substantive and relational knowledge in mental models, which may change through learning. The agents in CollAct discuss with each other and produce a group model (consensus). Factors identified to have an important influence on the results of a group discussion include group size, the level of controversy within the discussion, cognitive diversity, social behavior in form of cognitive biases (Asch and halo effect), and, depending on group size, the existence of a leading role at the beginning. Furthermore, the integration of topics into the consensus follows a saturation curve, thus the ending time of discussions should be carefully chosen to avoid a loss of information.