A normative theory of aggression

Aggression, leading to the formation of social hierarchy, has an evolutionary role in reducing conflict and facilitating the allocation of limited resources. Previous studies in mice have identified the regions involved in the neural circuitry of aggression, yet the mechanisms leading to the emergence of aggressive behavior remain to be poorly understood. To analyze such mechanisms, we used a mouse chronic social conflict paradigm where multiple aggressive phenotypes were formed, for which we obtained behavioral readouts (134 mice) and whole-brain snapshots of neuronal activity (c-Fos; 54 mice). We modeled the paradigm as a normal-form game to define the optimal actions for mice based on their strengths and used the Bayesian inference approach to model the animals’ beliefs about their strengths. Using our model and behavioral data, we formed and validated several hypotheses about the reward schedule, information availability, and evidence accumulation corresponding to aggressive behavior. Using our whole-brain c-Fos data, we found that the beliefs in our model are significantly correlated with neuronal activity in multiple distinct locations throughout the brain. Overall, our resulting model proposes a normative basis for the emergence of aggressive behavior and identifies the brain regions possibly involved in corresponding computations.