Neural Prioritisation of Past Solutions Supports Generalisation

Abstract

AbstractGeneralisation across tasks is an important feature of intelligent systems. One efficient computational strategy is to evaluate solutions to earlier tasks as candidates for reuse. Consistent with this idea, we found that human participants (n=40) learned optimal solutions to a set of training tasks and generalised them to novel test tasks in a reward selective manner. This behaviour was consistent with a computational process based on the successor representation known as successor features and generalised policy improvement (SF&GPI). Full model-based control or model-free perseveration could not explain choice behaviour. Decoding from functional magnetic resonance imaging data revealed that solutions from the SF&GPI algorithm were activated on test tasks in visual and prefrontal cortex. This activation had a functional connection to behaviour in that stronger activation of SF&GPI solutions in visual areas was associated with increased behavioural reuse. These findings point to the neural implementation of an adaptive algorithm for generalisation across tasks.