Contextual inference through flexible integration of environmental features and behavioural outcomes

The ability to use context to flexibly adjust our decision-making is vital for navigating a complex world. To do this, the brain must both use environmental features and behavioural outcomes to distinguish between different, often hidden contexts; and also learn how to use these inferred contexts to guide behaviour. However, how these two interacting processes can be performed simultaneously remains unclear. Within the brain it is thought that interaction between the prefrontal cortex (PFC) and hippocampus (HPC) supports contextual inference. We show that models using environmental features (similar to those proposed to be implemented in hippocampus) readily support context-specific behaviour, but struggle to differentiate ambiguous contexts during learning. In contrast, models using behavioural outcomes (similar to those proposed in PFC) can stably differentiate contexts during periods of learning, but struggle to guide context-specific behaviour. We show that supporting feature-based with outcome-based strategies during learning overcomes the limitations of both approaches, allowing for the formation of distinct contextual representations that support contextual inference. Moreover, agents using this joint approach reproduce both behavioural- and cellular-level phenomena associated with the interaction between PFC and HPC. Together, these results provide insight into how the brain uses contextual information to guide flexible behaviour.