A normative theory of social conflict

Social hierarchy in animal groups carries a crucial adaptive function by reducing conflict and injury while protecting valuable group resources. Social hierarchy is dynamic and can be altered by social conflict, agonistic interactions, and aggression. Understanding social conflict and aggressive behavior is of profound importance to our society and welfare. In this study, we developed a quantitative theory of social conflict. We modeled individual agonistic interactions as a normal-form game between two agents. We assumed that the agents use Bayesian inference to update their beliefs about their strength or their opponent's strength and to derive optimal actions. We compared the results of our model to behavioral and whole-brain neural activity data obtained for a large (n=116) population of mice engaged in agonistic interactions. We find that both types of data are consistent with the first-level Theory of Mind model (1-ToM) in which mice form both "primary" beliefs about their and their opponent's strengths as well as the "secondary" beliefs about the beliefs of their opponents. Our model helps identify brain regions that carry information about these levels of beliefs. Overall, we both propose a model to describe agonistic interactions and support our quantitative results with behavioral and neural activity data.