The use of knowledge in analogy and induction

This is Stuart Russell's doctoral dissertation, completed at Stanford under Michael Genesereth and published as a book in 1989. It examines how a reasoning system can use existing knowledge to draw valid analogies and make sound inductive generalizations, asking what justifies concluding that two situations are similar enough for a solution to one to carry over to the other. The work builds a formal framework for analogical and inductive inference grounded in what the system already knows, rather than treating each new case in isolation. Published in Pitman's Research Notes in Artificial Intelligence series, it is narrower in scope than Russell's later books, but its concern with bounding and guiding inference using prior knowledge recurs throughout his subsequent research, from bounded rationality to the problem of keeping AI systems aligned with human intentions.