A multiattribute utility approach to inference understandability and explanation
From Semantic Portal Wiki
{{#vardefine:category|Publication}}{{#vardefine:templatename|i.publication}}{{#vardefine:package|smwbp_instance_templates}}
| Edit |
Reference: {{#vardefine:pagename|a multiattribute utility approach to inference understandability and explanation }}
- [[]]
bibtex
{{#vardefine:pagename|A multiattribute utility approach to inference understandability and explanation }}{{#vardefine:key| }}
abstract: We describe reaearch on the use of multiattribute utility models for tailoring expert-system inference and explanation to classes of users within alternative decision-making contexts. In particular, we introduce a formalization of automated explanation. We describe how useful explanations can be synthesezed through the optimization of utility models that capture components of value in explanation. We discuss techniques for balancing the benefits of increased detail or completeness of analyses and explanations with the costs associated with accompanying increases in computational and cognitive complexity. We describe how the nature and optimal solution of such tradeoffs depend on the user and on the context. After introducing early work on user-specific multiattribute utilty models in the PATHFINDER expert system for tissue-pathology diagnoses, we discuss current research on making complex reasoning more valuable through the explicit representation and manipulation of knowledge about multiple attributes of value in computational behavior.
download:
- paper:
- slides:
| Abstract | We describe reaearch on the use of multiat … We describe reaearch on the use of multiattribute utility models for tailoring expert-system inference and explanation to classes of users within alternative decision-making contexts. In particular, we introduce a formalization of automated explanation. We describe how useful explanations can be synthesezed through the optimization of utility models that capture components of value in explanation. We discuss techniques for balancing the benefits of increased detail or completeness of analyses and explanations with the costs associated with accompanying increases in computational and cognitive complexity. We describe how the nature and optimal solution of such tradeoffs depend on the user and on the context. After introducing early work on user-specific multiattribute utilty models in the PATHFINDER expert system for tissue-pathology diagnoses, we discuss current research on making complex reasoning more valuable through the explicit representation and manipulation of knowledge about multiple attributes of value in computational behavior. ibutes of value in computational behavior. |
| Author | Eric Horvitz + |
| Bibtype | techreport + |
| Institution | Knowledge Systems, AI Laboratory + |
| Key | KSL-87-28 + |
| Note | 34 pages. + |
| Number | KSL-87-28 + |
| Tag | Computer science + |
| Title | A Multiattribute Utility Approach to Inference Understandability and Explanation + |
| Tr id | KSL-87-28 + |
| Year | 1987 + |
