A critiquing model of flexible constraint evaluation for a scheduler's workbench

Scheduling complex tasks is a difficult and ill-structured problem. Totally automated solutions to certain scheduling problems have certainly been achieved; however, other types of scheduling tasks do not yield easily to traditional solution methods. The latter tasks often involve both quantitative and qualitative constraints as well as changing preferences and subjective judgement. Consequently, it is sometimes impossible to take the human element out of the loop. Faced with similar problems, research in medical artificial intelligence has yielded a model of advising, called critiquing, which can be made to be comprehensible to, and consistent with, the decisionmaker's methods. In this paper we describe a project which incorporates a version of the critiquing model within a hybrid artificial intelligence/analytical-based scheduler's workbench, called MRL.