Analogy and Similarity in Scientific Reasoning

Abstract

This paper argues that both human and artificial intelligence are in many situations bound to use analogical inference for heuristic and justificatory purposes. by employing a quantitative concept of similarity, formal systems for noninductive and inductive inference by analogy are constructed (developing the earlier work of carnap, niiniluoto, spohn, kuipers, and costantini). it is further suggested that, instead of one system of inductive analogy, frameworks with different characteristic properties are needed in various types of applications.