Human Benchmarking of Expert Systems
Abstract
This document outlines the strategy used for benchmarking expert systems to human performance. Two major alternatives for human benchmarking of expert system are proposed: computer science driven or psychological process driven. The computer science driven approach is either (1) expert system driven in which one picks an expert system which encodes an expert, derives psychological processes, and tests the processes with people, or (2)expert system shell driven in which one estimates the intelligence of the shell (parent), assumes that applied expert systems will have similar intelligence, then follows the procedures of the expert system driven approach. The psychological process driven approach involves picking a psychological process, finding an expert system that exemplifies the process, and then tailoring a test for the expert system and for people. The computer science driven approach was used to develop the expert system benchmarking methodology by relating cognitive taxonomies and expert system taxonomies.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1990
- Accession Number
- ADA233601
Entities
People
- Anat Jacoby
- Harold F. O'neil Jr.
- Kathleen M. Swigger
- Yujing Ni
Organizations
- University of California, Los Angeles