Subjective Workload Ratings
| Full Text: |
 Pdf
(0.46 MB) |
|
Document Number: |
DOT/FAA/CT-05/32 |
|
Publication Date: |
01-2005 |
| Authors: |
Ulf Ahlstrom Ferne Friedman-Berg, PhD. |
Ahlstrom, U., & Friedman-Berg, F. (2005). Subjective workload ratings and eye movement activity measures (DOT/FAA/CT-05/32). Atlantic City International Airport, NJ: Federal Aviation Administration William J. Hughes Technical Center.
Abstract
In the present study, we evaluated the possibility of using eye movement activity measures as a correlate of cognitive workload. Using data from a high-fidelity human-in-the-loop weather simulation, we explored eye activity measures like pupil diameter, blink duration, and saccade distance, and assessed their relationship to subjective workload ratings. In our initial analysis, we established that although there was no significant effect of weather tool use on subjective workload ratings, there was a significant relationship between subjective workload ratings and our task load variable aircraft density. We found a linear increase in workload ratings with an increasing number of aircraft in the sector. In a subsequent analysis, we assessed the relationship between eye movement activity measures and aircraft density. We found that the mean blink duration and the mean saccade distance decreased as aircraft density increased, while the mean pupil diameter increased with an increasing number of aircraft in the sector. After establishing the relationship between these eye activity metrics and subjective workload, we evaluated whether we could use changes in eye movement activity along with other system state variables, like distance to weather from the outer marker, to measure ongoing controller workload. We developed both individual controller models and a general model (across controllers) to assess whether it was possible to predict the minute-by-minute number of aircraft in the sector. Using both multiple regression modeling and neural network models, we were able to produce individual controller models and general models with good prediction performances. We discuss possible applications for these findings in future air traffic control (ATC) research, in adaptive automation, and in ATC interface design.
Updated: May 04, 2012 11:21 AM
