Hippocampus facilitates reinforcement learning under partial observability

Abstract

Mastering navigation in environments with limited visibility is crucial for survival. Although the hippocam-pus has been associated with goal-oriented navigation, its role in real-world behaviour remains unclear. To investigate this, we combined deep reinforcement learning (RL) modelling with behavioural and neural data analysis. First, we trained RL agents in partially observable environments using egocentric and allocentric tasks. We show that agents equipped with recurrent hippocampal circuitry, but not purely feedforward networks, learned the tasks in line with animal behaviour. Next, using dimensionality reduction, our agents predicted reward, strategy, and temporal rep-resentations, which we validated experimentally using hippocampal recordings. Moreover, hippocampal RL agents predicted state-specific trajectories, mirroring empirical findings. In contrast, agents trained in fully observable en-vironments failed to capture experimental observations. Finally, we show that hippocampal-like RL agents demon-strated improved generalisation across novel task conditions. In summary, our findings suggest an important role of hippocampal networks in facilitating reinforcement learning in naturalistic environments.