![]() |
|||
![]()
|
![]() |
![]() |
![]() |
Chapter 7
|
1. | Introduction | |
2. | Stages in the Development of Explanation Facilities | |
3. | Types of Expert System Explanations and Strategies for Providing Explanations | |
4. | Recent Studies of the Use of Expert System Explanations | |
5. | Factors Influencing the Design of Explanation Facilities | |
5.1. | Task Characteristics | |
5.2. | Characteristics of the Explanations | |
5.3. | Interface Design and Provision Strategies | |
5.4. | User Characteristics | |
6. | Research Issues and Future Trends | |
7. | Summary | |
References |
The ability to provide explanations that clarify its functioning and recommendations has long been recognized as being an integral component of an expert system (ES). For example, the MYCIN system incorporated a simple explanation facility providing justification and reasoning-trace explanations. There has also been evidence to suggest that ES explanations are highly utilized by users and that this has an impact on their performance (Ye and Johnson, 1995; Dhaliwal and Benbasat, 1996). The ability to provide explanations has also been rated as being the primary user requirement for an expert system (Teach and Shortliffe, 1981). Largely based on Clancey's (1983) characterization of the epistemological roles that knowledge can play in expert system explanations, it has generally come to be accepted that explanation facilities must provide three types of knowledge corresponding to the following explanation labels: Why, How, and Strategic explanations. Why explanations provide justification knowledge clarifying the underlying reasons and contextual underpinnings for an action or state based on causal models. How explanations provide reasoning trace knowledge that clarifies inference structure and describes contents. Strategic explanations provide meta-knowledge that clarifies problem-solving strategy and representational structure.
In the 1980s, early research into the design of explanation facilities focused on how the knowledge encased in the knowledge base of an ES could be represented in a manner that would facilitate both optimal ES operation as well as the provision of explanations to users. The focus was on machine learning algorithms and knowledge representational formalisms that could simultaneously achieve both objectives. In the 1990s, however, research into the design of ES explanation facilities has taken on an empirical interface design orientation that recognizes that not all the knowledge required for ES explanation purposes can be obtained from the knowledge base of an ES. It is now viewed as a separate design problem that must be considered in isolation and that poses its own unique set of challenges.
This chapter focuses on development and design issues pertaining to the building of expert system explanation facilities. The next section discusses the relationship between the development life cycle of an expert system and that for building explanation facilities. It argues that special consideration must be given to the unique challenges pertaining to the development of explanation facilities. Section 3 focuses on the historical progression in the types of explanations that are incorporated in ES explanation facilities. These types of explanations are also linked to several generic strategies for providing explanations so as to foster user learning during the use of an ES and its explanations. Section 4 describes recent studies that have evaluated the provision of explanations to users as an interface design issue and summarizes their findings. Section 5 discusses various design considerations that have to be considered by designers of ES explanation facilities. Section 6 highlights various research areas that are currently being investigated by the ES explanations research community, and Section 7 presents the conclusion.
Previous | Table of Contents | Next |
![]() |
|
Use of this site is subject certain Terms & Conditions. Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details. |