Learning Abductive Search by Analogy in ACT-R
Alexander P.M. van den Bosch · 1996
The general question investigated in this paper is: How can the cognitive architecture of John Anderson's ACT-R theory of the mind be used for modeling the search and evaluation of (scientific) hypotheses? It is demonstrated how inductive search, modeled by Paul Thagard's PI and Herbert Simon's BACON, can both be learned from examples by ACT-R as a form of abductive search. Furthermore, I argue that PI's hypothesis evaluation process 'inference to the best explanation' is subsumed and extended by ACT-R's rational analysis mechanism. 1 Introduction In cognitive science, scientific endeavor itself is being studied as an interesting cognitive phenomenon. One popular view is taking scientific discovery, or the search and evaluation of hypotheses, as just a form of human problem solving (Langley et al. 1987). One of the most complete theories about human problem solving is developed by John R. Anderson. It is called ACT-R, adaptive character of thought - rational version, and is implemente...