The implementation and evaluation of a theory for high level cognitive skill acquisition through expert systems modelling techniques

Abstract

From previous studies by Koubek and Salvendy (1988), it has been established that differences exist in the high level controlled processes between expert (E) and super-expert (SE) subjects on computer program modification tasks. This study examines the implications of this finding by applying modelling techniques via expert systems technology. To examine the hypothesis that the knowledge representation is critical to SE performance, and to indicate how these results might have practical application, two small prototype expert systems were developed using the E and SE knowledge representations respectively. A qualitative analysis reveals significant performance differences between systems attributable to the knowledge representation and suggests a combination of E and SE knowledge be used for construction of a hybrid expert system. The SE knowledge base is a hierarchical structure and is organized under abstract categories, suggesting a breadth first approach. This structure allows for interactions within the program itself and between the program and user environment. The E knowledge base is narrow and task specific, indicating a depth first approach. In conclusion, theoretical and practical implications of this research into SE performance and modelling techniques are discussed.

Keywords:

• Modelling
• Expert systems
• Skill acquisition
• Knowledge representation
• Computer programming