Dynamics of similarity
Similarity judgments are involved in various kinds of cognitive
activities (categorization, problem-solving, learning). Consequently,
they must be very flexible. In other words, we have the capability to
compute different kinds of similarities, depending on activities' goals and
contexts. This flexibility cannot be achieved by cognitive mechanisms that
simply compute the degree of overlapping features.
Analog Retrieval and Similarity
Not a few researchers agree that retreival from LTM is governed by the
similarity of a given cue to items in LTM. We store infinite pieces
of information in LTM, and have to retrieve relevant one from it when
making analogies. The relevance of stored information to the current
situation is closely related to the goal that they share.
Goal-sensitivity of human similarity judgment
Thus, a challenging question to establish a theory of analogy is
whether our mechanism for similarity judgments is goal-sensitive.
We tested this hypothesis by comparing similarity ratings of the goal
state and others done by those varying degrees of expertise in the
Tower of Hanoi puzzle.
Main findings are that experts' judgments were function of the
distance of states to the goal, while naives' ones were function of
the number of shared features. This suggests that experts' judgments of
similarity reflect the goal-ralated features.
Related Papers
- Suzuki, H., Ohnishi, H., & Shigemasu, K. (1992) Goal-directed
processes in similarity judgement. In Proceedings of the
Forteenth Annuanl Conference of Cognitive Science Society, 343-348.
- Ohnishi, H., Suzuki, H., & Shigemasu, K. (1993) A
context-sensitive model of human similarity judgment. Japanese
Psychological Review, 36, 633-649. (Japanese with English abstract).
- Ohnishi, H., Suzuki, H., & Shigemasu, K. (1994) Similarity by
feature creation. In Proceedings of the
Sixteenth Annuanl Conference of Cognitive Science Society, 687-692.