Content area
Abstract
Expertise constitutes the foundation of an organization's ability to benefit from competitive and comparative advantage. Expert systems are the principal form of technology promising enhanced propagation of expertise, yet have been accused of fostering its extinction. The critical feature distinguishing expert systems from other computer-based information systems is their potential to furnish explanations.
Prior studies demonstrated that explanations enhance user confidence in expert system recommendations, discovering that explanation presentation affects user acquisition of several forms of domain knowledge. This study extends prior findings by integrating earlier results within a proposed framework for expert system explanations, the core of which was empirically tested and confirmed. The framework establishes a link among explanation content and intended function.
Relationships among explanation content, presentation format, acquisition of expertise, invocation mode, and various user perceptions regarding the system and explanations were hypothesized. A simulated expert system interface, with variations in explanation features, was written to conduct a controlled experiment employing 260 student subjects. The problem task was product/process matching, within the domain of manufacturing strategy.
The principal content/function associations stipulated in the framework were confirmed: Facts content promotes clarification, trace content promotes duplication, and justification content promotes ratification.
The critical importance of justification explanations was highlighted through numerous significant findings. The impact of the presence or absence of justification explanations was pervasive, spreading to measures that were not affected by this feature, such as ease of use. Most subject perceptual measures were strongly linked to this one characteristic, indicating that user acceptance of an expert system may hinge on the presence of justification explanations. Differences were also found between optional and automatic invocation of explanations, indicating that even a single keystroke for invocation may affect a user's decision to view an explanation, and thereby indirectly affect their perceptions of the system's adequacy.
