Abstract

Network Model-Guided Expert Knowledge Elicitation

Madni, A.M. and Chu, Y.

Abstract

Expert protocol analysis consists of formal and informal methods for collecting and associating various levels of expert performance data related to the enactment of specific tasks. Conventional analysis methods, however, tend to be ad hoc. To date, few rigorous and consistent approaches exist. Using these existing approaches, it is difficult to uncover concurrent processes and decisions that contribute to superior expert performance and unacceptable novice performance. To this end, this paper presents a knowledge representation framework based on Modified Petri Nets for modeling concurrent tasks and suggesting methods for expert protocol elicitation and analysis. This framework provides a model-guided technique for eliciting expert knowledge and encoding task protocols for complex problem solving and decision making tasks. The model has been expanded to represent a human operator's behavior as a focused control of production rule-based sub-processes that include problem solving, pattern matching, and psychomotor activities. Specific aspects of task performance such as procedural skills, task concurrency, and errors/slips can be represented in this framework as loci of dynamically moving tokens.

From:

Madni, A.M. and Chu, Y., Proceedings of 1985 IEEE International Conference on Systems, Man, and Cybernetics, Tucson, Arizona, November 1985, pp. 10-13.