On the use of computerised decision aids:an investigation into the expert system as persuasive communicator

The results of this study show that an expert system is a persuasive message source and that users tend to follow the advice of an expert system without critically examining it. Accordingly, users will probably not notice when an expert system gives incorrect advice. These facts comprise a serious problem, because expert systems can make mistakes and have been developed on the implicit assumption that the user will judge the applicability of the expert system's advice. The Elaboration Likelihood Model (ELM; Petty & Cacioppo, 1986a, 1986b), a social psychological theory of persuasion, helps to explain what makes people accept or reject advice. The theory posits that people evaluate messages by either careful examination of the message (the central route to persuasion) or by using simple decision rules without actually studying the message (the peripheral route to persuasion). It is the latter behaviour that can lead users to misjudge the advice of an expert system. An expert system is a computer program that can solve a restricted set of problems using specific knowledge of these problems. Chapter 2 provides a description of an expert system and gives insight into the limitations of expert system advice. An expert system has a knowledge base, an inference engine, and an interface. The knowledge base contains the knowledge model that the inference engine uses to deduce the advice. Knowledge acquisition, the obtaining and construction of the expert system's knowledge model, is the most difficult part of expert system design (Waern, 1989). Some of the problems of knowledge acquisition are the definition and restriction of the knowledge domain, the limitations of models, the consideration of different experts' opinions, and the use of common sense knowledge. As a result, the expert system's knowledge model is probably incomplete and possible incorrect. Consequently, an expert system ought not be treated as an oracle (Boden, 1990). It is an advisory decision aid and the user should evaluate the applicability of its advice. Adequate explanation facilities and a user friendly interface are necessary to provide the user with sufficient information to to be given to the expert system, and the user should do the actual problem solving. That is, expert systems should be designed as a tool for decision making in which its user is critically placed in the centre of the decision making process