Question-Generation in Constraint-Based Expert Systems

If-then rules are the core knowledge representation technology in currently deployed expert systems. However, if we replace rules by constraints, we can greatly extend both the expressive and reasoning functionality of such systems. An open issue, however, is how best should constraint-based expert systems decide which questions to ask their users during a consultation – how can an interactive constraint-based expert system extract needed data from its users while imposing the smallest possible burden on those users. In this paper we consider three types of approach to the task.