Proceedings of the Annual Meeting of the Cognitive Science Society

The idea of a “cognitive map” was originally developed to ex-plain planning and generalization in spatial domains througha representation of inferred relationships between experiences.Recently, new research has suggested similar principles mayalso govern the representation of more abstract, conceptualknowledge in the brain. We test whether the search for rewardsin conceptual spaces follows similar computational principlesas in spatial environments. Using a within-subject design, par-ticipants searched for both spatially and conceptually corre-lated rewards in multi-armed bandit tasks. We use a GaussianProcess model combining generalization with an optimisticsampling strategy to capture human search decisions and judg-ments in both domains, and to simulate human-level perfor-mance when specified with participant parameter estimates. Inline with the notion of a domain-general generalization mecha-nism, parameter estimates correlate across spatial and concep-tual search, yet some differences also emerged, with partici-pants generalizing less and exploiting more in the conceptualdomain.