Perspectives in Explanation
Daniel D. Suthers · 1989
THIS REPORT REVIEWS ARTIFICIAL INTELLIGENCE RESEARCH RELEVANT TO COMPU- TATIONAL THEORIES OF EXPLANATION AS A COMMUNICATIVE ACTIVITY. THE LITERA- TURE IS ORGANIZED INTO THREE BROAD AREAS: EPISTEMOLOGICAL ADEQUACY, I.E., PROVIDING THE DIVERSITY OF KNOWLEDGE NEEDED IN EXPLANATIONS; EPISTEMOLOGI- CAL AND RHETORICAL ANALYSES OF EXPLANATORY STRUCTURE; AND ALTERNATE MECHAN- ISMS FOR GENERATING EXPLANATIONS. SEVERAL RESEARCH PROBLEMS AND DESIGN ISSUES ARE EMPHASIZED. TWO PROBLEMS OCCUR AT AN EPISTEMOLOGICAL LEVEL OF ANALYSIS. THERE IS SCATTERED RESEARCH ACCOUNTING FOR HOW EXPLANATIONS ARE CONSTRAINED BY THE INTERNAL STRUCTURE OF A DOMAIN''S BODY OF KNOWLEDGE, BY THE ROLE OF AN INDIVIDUAL''S KNOWLEDGE IN UNDERSTANDING NEW CONCEPTS AND SITUATIONS, AND BY THE WAYS IN WHICH INDIVIDUALS ARE WILLING OR ABLE TO TRANSFORM THEIR KNOWLEDGE. THESE EPISTEMOLOGICAL CONSTRAINTS ON EXPLANATION REQUIRE MORE EXPLICIT TREATMENT WITHIN A UNIFORM FRAMEWORK. FURTHER WORK IS REQUIRED TO IDENTIFY AND REPRESENT ALTERNATE CONCEPTUAL FRAMEWORKS FOR `PERSPECTIVES'' TAKEN IN AN EXPLANATION, AND TO ACCOUNT FOR AN EXPLAINER''S CHOICE OF A GOOD PERSPECTIVE FOR A GIVEN EXPLANATORY GOAL. A THIRD PROBLEM IS THE DESIGN OF A FRAMEWORK WITHIN WHICH TO OPERATIONALIZE THE VARIOUS CONSTRAINTS AND HEURISTICS DEFINING A `GOOD'' EXPLANATION AND COORDINATE AND INTEGRATE THEM WITH EACH OTHER. A SUGGESTION IS MADE, IN TERMS OF THE SOURCES OF THESE CONSTRAINTS AND THEIR IMPACT ON A PROCESS OF TRANSLATION