Integrating Abduction and Induction in Machine Learning

Abstract

Abduction is the process of inferring cause from effect or constructing explanations for observed events and is central to tasks such as diagnosis and plan recognition. Induction is the process of inferring general rules from specific data and is the primary task of machine learning. An important issue is how these two reasoning processes can be integrated, or how abduction can aid machine learning and how machine learning can acquire abductive theories. The machine learning research group at the University of Texas at Austin has explored these issues in the development of several machine learning systems over the last ten years. In particular, we have developed methods for using abduction to identify faults and suggest repairs for theory refinement (the task of revising a knowledge base to fit empirical data), and for inducing knowledge bases for abductive diagnosis from a database of expert-diagnosed cases. We treat induction and abduction as two distinct reasoning tasks, but have demonstrated that each can be of direct service to the other in developing AI systems for solving real-world problems. This chapter reviews our work in these areas, focusing on the issue of how abduction and induction is integrated.¹