Analogy, Mental Models, and Conceptual Change

Sponsor: Cognitive Science Division, Office of Naval Research

Principal Investigators: Dedre Gentner, Kenneth D. Forbus

Project Summary: This project explores the roles analogy and similarity processes serve in human learning and conceptual change. We focus on the learning of mental models because we believe that they are essential for understanding many phenomena that are the hallmarks of human cognition, such as the ability to reason about complex physical systems, to make and articulate predictions about the world, and to discover causal explanations for what happens around us. In an increasingly technological society, understanding the nature of mental models for complex physical systems and how to help people learn them better could provide substantial benefits. Furthermore, we believe that the deep understanding of analogy and similarity we seek can ultimately also lead to other benefits as well, including improvements in human-computer interaction (by better understanding the propensities of the human partners) and in designing better training materials.

Our goal is a psychological account. We use the vocabulary of AI and computation to gain representational precision and facilitate the development of explicit, testable process models. We test our ideas through a combination of psychological experimentation and computer simulation, building on and extending the tools and techniques we developed in our previous ONR-sponsored research.

Analogy and similarity play a central role in conceptual change. Analogy allows the application of preexisting conceptual structure to new problems and domains, and hence supports the rapid learning of new systems. Of all the learning processes, analogy is the only one that offers a mechanism for the acquisition of substantial knowledge structures in a brief span of learning. By contrast, other learning processes, such as generalization, differentiation, accretion, or compilation, offer ways of refining, adding to or consolidating an existing system of beliefs. Further, similarity is a primary factor in categorization, retrieval, and organization of conceptual structure. No theory of learning and concepts or of conceptual change can be complete without a clear account of the roles played by similarity and analogy.

Analogical learning is particularly central in the learning of complex systems. Analogies are commonly used in science instruction -- for example, when an electrical circuit is compared to a water-flow system. Learners' models of a domain are shaped by such instructional analogies. Analogy is used to apply arguments used in examples to new problems, and similarity is used to extract principles from repeated, overlapping experiences. Even without explicit teaching, learners' naive mental models are often formed around their own spontaneous analogies and implicit similarity comparisons. Sometimes these comparisons lead to models that are wholly or partly misleading. Thus, a better understanding of the psychology of analogy and similarity would not only improve our theories of learning and transfer but would also allow us to anticipate and perhaps prevent many conceptual errors.

Back to Structure-Mapping: A computational model of analogy and similarity page | Back to Qualitative Process Theory page | Back to MAC/FAC: A model of similarity-based retrieval page | Back to Previous Projects page | Back to QRG Home Page