Pictorial and text editors for expert system rules
Abstract
The effects of display mode on the mental models of users of expert system rule editors were studied. One group of subjects used a pictorial rule editor and a second group used a text rule editor to acquire knowledge about a simple chemical processing system. Subjects created rules and performed rule firing exercises to learn operating characteristics of the chemical system. Their mental models were tested for ability to interpret system operations, identify rules for operation, track system state changes in memory, identify and add missing pieces of incomplete rules, predict system state changes, determine system operating guidelines, design a new system operation, and evaluate rules for correctness. Results showed that subjects who used the pictorial editor were faster for tasks that required knowledge synthesis and transfer. Text editor subjects were faster for tasks that required sequential, repetitive cognitive processing. These effects were due to different mental models that resulted from the use of the two rule editors. Subjects who had used the pictorial rule editor had a knowledge-based mental model that was consistent with the operating characteristics of the chemical system because they had actually practiced with the system as they performed the knowledge acquisition. Subjects who had used the text rule editor had a rules-based model that required more cognitive processing in order to solve new problems. Evidence was found for additive effects of dual memory coding in the rule correctness task when subjects used a combination of the interfaces.
Degree
Ph.D.
Advisors
Eberts, Purdue University.
Subject Area
Industrial engineering|Computer science|Artificial intelligence
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.