Decoding behavior with minimal and interpretable agent models

Understanding how living organisms process sensory information from their surroundings and translate it into decisions is a fundamental problem across biological scales – from biochemical signalling in single-cells to neural computations in animal brains. In this work, we address this challenge by introducing a method to reconstruct general decision processes directly from behavioral observations alone. Our approach is applicable to any biological agent and does not require prior knowledge of its internal mechanisms or its environment. Our agent model is defined by a recurrent dynamics over a discrete set of internal states which encode and process sensory information, and dictate which actions to execute. We validate our method on synthetic agents and demonstrate that we can exactly recover the agent’s behavior for non-trivial tasks. Then, we infer agent models from experimental data of rats performing evidence accumulation and of mice making decisions under uncertainty and in changing environments. In both cases, very few internal states suffice to reproduce the observed behavior with high accuracy. Crucially, the immediate interpretability of the inferred dynamics allows to understand the computational process underlying decision-making.Author summary: Living organisms constantly process information to make decisions. Which are the computational processes that translate sequences of sensory cues into courses of actions? Since the internal states of an organism typically cannot be observed, it is difficult to answer this question. Even when measurements are possible – such as in large-scale recordings of neural activity – extracting computational principles from a high-dimensional dynamics still poses significant challenges. Here, we introduce a method to infer interpretable models of the decision-making process directly from behavioral trajectories, without making assumptions about the environment. Despite their simplicity, these agent models are expressive enough to accurately describe experimentally observed behaviors and reveal the structure of the underlying neural computational processes.