Reliable reasoning : induction and statistical learning theory /

This is an engaging four-chapter book whose authors (both, Princeton) essentially challenge the privileged position of deduction over induction as a method of reasoning. Unlike previous scholars who have treated the subject, e.g., Max Black, the authors characterize deduction simply as a rule-following procedure that tells one "what follows from what" and so does not enlarge one's knowledge. Induction, by contrast, enlarges knowledge by forcing one to give up old beliefs that are found to be inconsistent with new data, or to adopt new beliefs about a given phenomenon in light of new data that better explain the phenomenon. This epistemic value of the inductive method is what makes it a genuine form of reasoning, and hence raises significant questions about its reliability as method. Drawing upon statistical learning theory that highlights probability, and using "transduction," described as "a method that ... infers directly from data to the classification of new cases as they come up," the authors make a case for the reliability of the inductive method. Their treatment of the subject is as illuminating as it is provocative, and this book should appeal to a wide readership in the sciences and the humanities. Summing Up: Highly recommended. Lower-level undergraduates through faculty/researchers. C. S. Johnson Middle Tennessee State University