Statistical analysis of the effect of input errors on expert systemadvice

摘要

A randomized, statistical test method is presented in which systemresponses to specific input sequences are compared to those obtainedwhen some system queries are answered incorrectly. Over a representativesample of input sequences this method leads to an overall errorsensitivity rate as well as query-specific error rates that reflect keyquestions in the input sequence. The test method developed here can beused to generate overall and specific analyses of the consequences ofuser input errors in the application of a medical expert system. Asimple example is provided by a nonbranching, three-rule system forwhich all inputs are yeso, and three advice-oriented outputs areprovided