Donald Bren School of Information and Computer Sciences

Analogical reasoning has a seductive history in artificial intelligence (AI) because of its assumed promise for the acquisition and effective use of knowledge. This paper, the second of a two part series, reviews a vartiety of efforts at capturing some of this promise in the existig AI literature. Studies are reviewed chronologically within broadly defined task domains, and a simple framework for examining components of analogical reasoning is developed for comparing computational studies. This framework, and many of the ideas surrounding it, are related to important but often overlooked contributions from other academic disciplines reviewed in the first paper of this series. Particular emphasis is given here to the role of analogical reasoning in machine learning.