Identifying Cost Effective Boundaries of Operationality

Abstract

Operationality is concerned with the cost of testing the relevance of a concept description in a new situation. A boundary of operationality can be viewed as delimiting a set of predicates with ‘acceptable’ test costs by some criteria. Very often though, these criteria do not take into account the ability of the predicate's truth value to discriminate between competing concept descriptions for one which is most relevant. Thus, in isolation a predicate may have a moderate test cost, but the benefits of this test in terms of continued search may outweigh the costs. This paper describes EXOR, which is an inductive concept formation system that abstracts and organizes explanatory knowledge for efficient reuse. In particular, we will focus on EXOR'S identification of an appropriate boundary of operationality, which we interpret as maximizing a measure of cost effectiveness.