Inferring collective dynamics in groups of social mice

Social interactions are a crucial aspect of behavior in many animal species. Nonetheless, it is often difficult to distinguish the effect of interactions from independent animal behavior (e.g. non-Markovian dynamics, response to environmental cues, etc.). In this talk, I will address this question in social mice, where we infer statistical physics models for the collective dynamics for groups of 15 mice, housed and location-tracked over multiple days in a controlled environment (the Eco-HAB [Puścian et al. 2016]). We reproduce the distribution for the co-localization patterns using pairwise maximum entropy models, and find that the resulting local fields successfully predict the transition rates. I will also discuss progress towards developing novel inference methods with memory, which while giving consistent equilibrium distribution as the maximum entropy models, also capture dynamical observables, e.g. long-tailed waiting time distributions, and that mice actively follow each other. We show that in the case of social mice, both individuality and interaction with peers are essential to give the observed co-localization patterns.