Dissonance Minimization and Conversation in Social Networks

We study a model of social learning in networks where the dynamics of beliefs are driven by conversations of dissonance-minimizing agents. Given their current beliefs, agents make statements, tune them to the statements of their associates, and then revise their beliefs. We characterize the long-run beliefs in a society, provide the necessary and sufficient conditions for a society to reach a consensus, and show that agents’ social influences (weights on the consensus belief) are decreasing in their dissonance sensitivities. Comparing the outcomes of two models, with and without conversation, we show that conversation leads to a redistribution of social influences in favor of agents with higher self-confidence. Finally, we provide analytical insights for the model where agents minimize dissonance by revising both beliefs and network, and show that an endogenous change of network may prevent a society from reaching a consensus.