FUZZY COGNITIVE MAPS: BRIDGING THE GAP BETWEEN COGNITIVE TASK ANALYSIS AND SYSTEM DESIGN

摘要

There has been a renewed interest in using cognitive task analysis to support design of better systems. However, as Roth, Gualtieri, Easter, Potter, and Elm [2000] note, as useful as CTA has been in systematically describing complex human performance, all too often, the results of CTA are only weakly linked to design. In large measure, this can be attributed to the lack of a formalism for representing CTA results. Creating an abstract and algorithmic model of intelligent systems based on CTA results has proven to be difficult given that most types of CTA rely on narrative and text-based methods of data collection. Fuzzy Cognitive Maps (FCM) are signed directed digraphs that are comprised of two elements, concepts and causal relationships. Concepts are nodes that represent fuzzy sets or fuzzy events that have certain degrees of activation, and causal relationships are the fuzzy rules for the mappings between concepts. The Fuzzy Cognitive Map presented is actually a hybrid of several different established artificial intelligence constructs that include fuzzy logic, neural networks, finite state machines, and genetic algorithms. This paper presents a method in which data collected using a CTA is applied to the construction of a Fuzzy Cognitive Map for the use in the design and implementation of an intelligent system. We will demonstrate how CTA data can be used to derive FCMs. In addition to a specific application of this technique, discussion will also focus on how FCMs might benefit modeling of intelligent entities.