![]() |
|||
![]()
|
![]() |
![]() |
![]() |
3. TYPES OF EXPERT SYSTEM EXPLANATIONS AND STRATEGIES FOR PROVIDING EXPLANATIONSThe Why and How explanations, which were first introduced in MYCIN (Shortliffe, 1976), remain the foundation of most explanation facilities found in current ES applications and development shells. Attempts have been made to incorporate other forms of explanations. These include the Strategic, What, and What-if explanations. The Strategic class of explanations provide insight into meta-knowledge, especially the control objectives and overall problem-solving strategies used by a system. For example, the NEOMYCIN system explicitly outlines problem-solving strategies in its own knowledge base and makes them available for explanation (Hasling, Clancey and Rennels, 1984). The What explanations are designed to give insight into object definitions or decision variables used by a system (Rubinoff, 1985). They serve as responses to queries such as: "What do you mean by object or variable name?" The What explanation is significantly different from the What-if query facilities commonly found in decision support systems. These refer to the ability to rerun a consultation with changed model parameters. While such What-if facilities can be provided as part of the ES interface, they are not viewed as being explanations per se, but rather as tools for sensitivity analysis. To be considered a distinct category of explanations, What-if has to be implemented as the direct and explicit provision of information about the sensitivity of decision variables to ES users, instead of being a facility for performing sensitivity analysis. Various classifications of the many types of explanations that should be provided by ES have been suggested. These classifications can be condensed as subscribing to one of two possible criteria for distinguishing between the types of explanations. The first criteria is the nature of the explanation queries. For example, Wick and Slagle (1989) discuss explanations whose queries begin with What, Why, How, When, Where, etc. As well, Swartout (1983) considers the How, Why, When, and What range of queries, as part of XPLAIN's explanation facility. The second criteria is the nature of the explanation responses. Swartout and Smoliar (1987) distinguish between explanations that provide terminological knowledge, domain descriptive knowledge, and problem-solving knowledge. ES users require information about procedures, reasoning traces, action goals, control, and self-knowledge. There are two ways of distinguishing explanation responses. First, responses can provide case-specific knowledge, domain knowledge, or meta-knowledge. Second, they can provide taxonomic knowledge, formal knowledge, contingent knowledge, or control knowledge. Irrespective of whether they are based on explanation queries or explanation responses, there is a major problem with all these classifications. Lacking a sound theoretical basis, the various types of explanations that comprise each of these classifications are neither consistently defined, nor is each classification comprehensive. However, largely based on Clancey's (1983) characterization of the epistemological roles that knowledge can play in ES explanation, a consensus has emerged on the three primary types of explanations that ES ought to provide. Corresponding to the three epistemological roles of structure, support, and strategy, these three types of explanations are: (1) trace explanations that describe contents and reasoning (structure), (2) deep explanations that justify underlying reasons for a state or an action based on causal models (support), and (3) strategic explanations that clarify problem-solving strategy and meta-knowledge (strategy). This taxonomy of the three primary types of explanations has also led to a convergence of opinion on the matching of explanation queries with explanation responses. This is as follows: the How explanation queries are used to provide trace explanations; the Why explanation queries are used to provide causal justifications; and the Strategic explanation queries are used to provide clarifications of control strategies and meta-knowledge. Recent work has focused on the manner in which explanations should optimally be provided as part of the user-ES interaction. It has proposed that ES explanations can be presented to users in two distinct ways to facilitate user learning during the use of an expert system (Dhaliwal and Benbasat, 1996). These are termed feedforward and feedback explanation provision strategies and are relevant to all three of the Why, How, and Strategic explanations; that is, each of these types of explanations can be presented both as feedforward and feedback. Table 2 presents definitions for the three types of explanations when they are presented either as feedforward or feedback. The feedforward explanations differ from feedback explanations as follows: (1) the feedforward version is not case-specific, while the feedback version explains aparticular case-specific outcome; (2) the feedforward version is presented prior to an assessment or diagnosis being performed, while the feedback version is presented subsequent to the assessment and after the presentation of the outcome of that assessment; and (3) the feedforward version focuses on the input cues while the feedback version focuses on the outcomes.
The idea that particular explanation provision strategies may be matched with the various types of explanations for optimizing value to ES users is an important one. It suggests that designers of ES explanation facilities must pay careful attention to the close relationship between the types of explanations and the feedforward and feedback explanation provision strategies.
|
![]() |
|
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. |