A Framework for Model-Based Adaptive Training

Scheduling System - Associative Knowledge

An alternative and more general approach to procedural knowledge is to use situation dependent knowledge. In many problems it is more flexible to recognise a particular situation and to use a pre-compiled or empirically derived set of rules to determine the solution. This of course relies on the existence of the appropriate rules; if the situation is abnormal or at least outside the previous experience of the expert, no solution is possible, and the trainee must resort to the less efficient but more general principles.

The associative MOBIT expert has internal rule slots (i.e., variables) which are filled from a database [MOBIT 1994]. This form of expert is created for the purpose of demonstrating to a trainee just how an expert deals with simulations which are not simply following a fixed procedure.

The MOBIT expert does not support the concept of problem solving methods; there are no conceptual problem solving control structures imposed on the associative model. In Scheduling-MOBAT, the associative knowledge model is extended with the RIME methodology [van de Brug et al. 1986], and incorporates reasoning capabilities in the form of an expert system including an explanation facility.

Examples for the representation of associative knowledge in Scheduling-MOBAT have been presented earlier in this chapter.

The extended associative model provides more flexibility in using the expert system within the training system architecture:

  1. to provide demonstrations to a trainee;
  2. to provide flexible step-by-step reasoning methods which can be controlled by the trainer agent; and
  3. to provide an independent agent acting as a task-specific explanation facility
Flower Show
© | | Sitemap