Three-term relation of analogy

Analogy has been considered to be the mapping between a base and target. But, this framework requires several unrealistic processing in modeling analogies, such as exhaustive mappings in the initial stage of analogy.

My idea is that what is really computed in human analogical reasoning is the equivalence (or sameness) between a base and target. In order to establish the equivalence, one has to postulate a superordinate class (hereinafter abstraction) that subsumes both a target and base. Therefore, analogy should be regarded as a three-term relation (base, target, and abstraction), rather than the direct mapping between a base and target.

Quasi-abstraction

What kind of abstraction is involved in human analgical reasoning? One of the most rubust findings in human learning is that people have considerable difficulties in learning and using abstract knowledge. On the other hand, some linguists (Lakoff and his colleagues) and develpmental psychologists (A. Brown and Goswami) have found that people, even 3-year-olds, can use, detect, and transfer the abstract relations. These contradictory results can be reconciled by carefully analyzing the nature of abstraction. Abstraction people can use have following properties: Such abstractions differ from those generated by random deletion of features of concrete examples and from hard abstract rules and principles that scientists and logicians produced by entity-variabilizing. In this sense, human abstraction can be called "Quasi-abstraction."

Related papers