An Empirical Validation of Task Analysis Predictions Using Multiple Autoflight Interfaces

Luis Prada

Advisor: Deborah Boehm- Davis

David J. King Hall, 2005
January 24, 2008, 07:00 PM to 07:00 PM

Abstract:

Limits on time, money, and materials often prevent design teams from evaluating the usability of new systems during early formative stages of development. This paper describes several approaches intended to supplement user testing by predicting operator performance analytically. Predictions from KLM, CPM-GOMS, NGOMSL, CogTool and a label assessment technique are then compared to empirical results obtained during the testing of existing and prototype autopilot interfaces at a major aircraft manufacturer. Results showed that the prototype outperformed the reference system on almost all measures. When compared to the human data, all of the analysis techniques were found to be predictive, with the exception of NGOMSL's time predictions.