– 1 – Assumptions on the VT task model in DESIRE Frances M.T. Brazier, Pieter H.G. van Langen, Jan Treur, and Niek J.E. Wijngaards Artificial Intelligence Group Department of Mathematics and Computer Science Vrije Universiteit Amsterdam de Boelelaan 1081a 1081 HV Amsterdam E-mail: {frances, langen, treur, niek}@cs.vu.nl EXTENDED ABSTRACT This paper presents assumptions underlying the modelling and specification of the VT task in DESIRE. These assumptions are about the modelling of tasks in general, the specification of task models in DESIRE, the modelling of the VT task in particular, and the problem solving method developed for the VT task. By an explicit formulation of such assumptions, different problem solving methods for one task can be compared and the reuse of problem solving methods is facilitated. 1 Introduction The VT task is an elevator configuration task, described by Marcus, Stout and McDermott (1988), Marcus and McDermott (1989), and Yost (1994). The modelling and specification of the VT task by several AI groups throughout the world serves as a valuable test case for understanding, comparing, and evaluating different approaches in knowledge engineering. In January 1994, eight AI groups presented their approaches and solutions to the VT task in Schreiber and Birmingham (1994). Some of these approaches and solutions have been further discussed in Schreiber and Birmingham (1995). In order to compare the solutions and methodologies from the different AI groups, the assumptions underlying the modelling and specification of the VT task have to be made explicit. The identification of assumptions is also useful for the selection and retrieval (i.e., reuse) of problem solving methods in general. In this paper, the assumptions underlying the VT task model as described in Brazier, van Langen, Treur, Willems and Wijngaards (1994) are presented. The VT task has been modelled and specified using the DESIRE framework, described in Langevelde, Philipsen and Treur (1992) and Brazier, Treur, Wijngaards and Willems (1995). Assumptions underlying the modelling and specification of a task can be divided into several levels, where assumptions at one level establish a point of departure for assumptions at lower levels. In this paper, four levels are identified: (1) assumptions on task modelling, (2) assumptions on specifying task models in DESIRE, (3) assumptions on modelling the VT task, and (4) assumptions on the problem solving method developed for the VT task. The paper is structured accordingly.