Proceedings of the Annual Meeting of the Cognitive Science Society

Humans construct mental models to navigate novel situations, yet how they do so efficiently remains poorly understood. Recent resource-rational accounts have explored what representations should be constructed under computational constraints, but leave open how humans actually discover them – and how prior solutions might be reused in the process. We argue that analogy serves as a core mechanism, enabling people to reuse solution-relevant structure from past experience. Formalising analogies as partial homomorphisms between decision problems, we sketch a framework in which abstract modules, extracted from previous experiences, serve as composable building blocks for new mental models. When a child learns that a password is "like a key," they import not just surface similarity but functional structure: transition dynamics, possible actions, and effective strategies. This modular reuse points toward a process-level account of how humans construct situationally appropriate representations, and connects classic work on analogy with computational approaches to bounded rationality.