This paper presents a logical system for representing knowledge about defaults and prototypical properties. A proof theory and semantics are provided for the system, and the soundness and completeness of the system are proved. The logical system introduces a variable conditional operator that allows the representation of typical properties of classes of individuals as well as properties of exceptional individuals without giving rise to inconsistency. It is also possible to strengthen the antecedent of the variable conditional operator and to negate the consequent without giving rise to inconsistency. The variable conditional operator is not transitive. These properties of the variable conditional operator allow the logical system successfully to handle common examples of reasoning about defaults and prototypical properties without giving rise to inconsistency or intuitively incorrect results. However, the logical system presented in the paper is monotonic and does not have modus ponens as an inference rule. These are regarded as serious shortcomings by Reiter, who also provides a detailed comparison of several approaches to reasoning about defaults and prototypical properties [1].
This well-written paper is a revised version of a paper that won the Artificial Intelligence Journal Award at the 1986 Canadian AI Conference [2]. It will interest ,esearchers and graduate students in artificial intelligence.