Retrieval of Knowledge through Algorithmic Decomposition
Abstract
The report summarizes the results of a series of studies to evaluate the effectiveness of algorithms for estimating unknown quantities. The results of the first experiment show that, as the structure of the aid increases, the subjects' performance improves in terms of both accuracy and consistency across subjects. The second experiment, however, suggests that algorithms used without understanding are of limited help. Further experiments suggest substantial problems in designing decision aids based on algorithmic decomposition, not because the principles of creating algorithms are hard to learn but because the users may be misled by their own misinformation and lack of arithmetic skills. These findings lead us to believe that such decision aids should be used in situations where algorithms can be carefully and deliberately designed and computational aids are also available. Possible approaches are suggested for the design of an expert system using algorithmic decomposition. (Author) (KR)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1990
- Accession Number
- ADA225667
Entities
People
- Sarah Lichtenstein