Dynamic Extension of Episode Representation in Analogy-Making in AMBR

Models that rely exclusively on static representations cannot account fully for the flexibility of human analogymaking. More sophisticated models should provide mechanisms for dynamic extension, elaboration, and rerepresentation of episodes. One such mechanism—the instantiation mechanism—is described. It uses the target problem as a template for extending the source and vice versa. These extensions are driven and constrained by semantic knowledge about general regularities in the domain. The instantiation mechanism has been developed within a model of analogy-making called AMBR. It relies on AMBR's support for parallel, decentralized, and interactive computation. The instantiation mechanism runs in parallel with the mechanisms for analog access and mapping. Thus, these latter mechanisms guide the instantiation mechanism as to which facts in the large semantic memory are relevant to the specifics of the current situ-