Soft Matter and Complex Systems Seminar

Several recent cage designs support studies of multiple animals housed for weeks with minimal human intervention in a single or multiple compartments where they can interact with cage elements and with each other, and their behavior can be tracked in various ways. Here we focus on Intellicage system where up to 14 female mice housed together can be identified with an RFID transponder interacting with intelligent corners providing reward, and the behavior is described in terms of discrete events. We present a general conceptual, analytical and computational framework for stochastic description, analysis and modeling of data from such cages. This framework combines the theory of point processes (as used in spike train analysis) with reinforcement learning models. We demonstrate how individual and social aspects of learning can be identified within the data, and show different specific approaches which facilitate study of effects of the whole group on an animal or formation of a hierarchy of social effects in group learning. The results of the analysis are validated with equivalent simulated data.

To illustrate this conceptual framework and our analytical approach we designed an experimental paradigm where rewards are offered depending on an arbitrary assignment of an animal to one of two groups, “majority” or “minority”. The two groups were assigned different locations with reward availability, changing in consecutive phases of the experiment. We show that the data support importance of the social effects in animal learning of the reward and may also be used to identify a social structure within the group. Corresponding generative models can be used for validation of various analytical methods and for prediction of mice behavior.