Date of Award

Summer 1991

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Engineering Management

Program/Concentration

Engineering Management and Systems Engineering

Committee Director

Laurence D. Richards

Committee Member

Resit Unal

Committee Member

Barry Clemson

Committee Member

Frederick Steier

Committee Member

Stephen A. Zahorian

Abstract

The application of Dr. Genichi Taguchi's approach for design optimization, called Robust Design, to the design of human-computer interface software is investigated. The Taguchi Method is used to select a near optimum set of interface design alternatives to improve user acceptance of the resulting interface software product with minimum sensitivity to uncontrollable noise caused by human behavioral characteristics.

Design alternatives for interaction with personal micro-computers are identified. Several important and representative alternatives are chosen as design parameters for the Taguchi matrix experiment. A noise field with three human behavioral characteristics as noise factors were chosen as a representative noise array. Task accomplishment scenarios were developed for demonstration of the design parameters on an interactive human-computer interface. Experimentation was conducted using selected human subjects to study the effect of the various settings of the design parameters on user acceptance of the interface. Using the results of the matrix experiment, a near optimum set of design parameter values was selected.

A verification experiment was developed and performed using the predicted near optimum design parameter values. Analysis of the follow-up experiment indicated improved levels of user acceptance with the near optimum values.

This study suggests that the Taguchi Method of design optimization is applicable to human-machine engineering in general, and to the design of human-computer interface software in particular, as a means of selecting a near optimum set of design alternatives. This methodology is useful in reducing the number of total experiments required for optimization where several design alternatives exist in a richly interdependent context.

DOI

10.25777/tap9-zd87

Share

COinS