Models of Categorization

John K. Kruschke
Outline of Guest Lecture for Q500, 27 September 1995

http://www.indiana.edu/~jkkteach/Q500/categorization.html

Goals of Cognitive Science:
- Explanation at both functional (software) and implementational (hardware) levels.
- Explanatory principles and formalization.
  - Structure of modeling.
  - Formalization yields precision, explicitness, and theorist-independent predictions.
  - Why do we all know the legend that an apple fell on Isaac Newton's head?
    (A: Not because of the explanatory principle, but because of the formalization.)
- Caveat: Shun ontological reification!
  E.g., gravity. Representations and processes are in the models, not necessarily in the mind/brain.
Overview of My Research:
- Categorization.
  - The laboratory learning paradigm.
  - Types of empirical phenomena.
- Explanatory Principles.
  - Two-pronged approach: 1. Empirical testing. 2. Modeling. Logic is the same for both.
  - Some specific principles.
Principle of Selective Attention to Relevant Dimensions.
- Empirical phenomenon: Filtration vs. Condensation.
- Formalization: ALCOVE (cf. vanilla backprop).
Principle of Rapidly Shifting Selective Attention.
- Empirical phenomenon: Inverse base rate effect.
- Formalization: ADIT (cf. vanilla linear associator).