Young and Angell 2003
From GrokCodex
[edit] The dimensions of driver performance during secondary manual tasks
This study used PCA method to identify the underlining dimensions of driver’s performance, providing a complete and valid assessment of the performance when drivers conduct a secondary task and the primary driving task simultaneously. Fifteen measures of driver performance were collected when subjects conducted the secondary visual-manual tasks chosen from 79 tasks. The measures included task completion time, eyes-off-road time, number of glances to the in-vehicle system, lane deviation, subjective workload, subjective situation unawareness, speed deviations, percent unsuccessful task completion, percent of total visual events missed, percent of forward visual event missed, percent of side visual events missed, mean single glance time to system, time to respond to total visual events, time to respond to side visual events, and time to respond to forward visual events. The measures were averaged across subject and were analyzed using PCA. The authors found that the first three PCs accounting for 83% variance among 15 measures were very meaningful and explained them as “overall driver demand”, “low workload but high inattentiveness”, and “peripheral insensitivity”.
[edit] Commentary
The contribution of this paper is to provide an example of fundamental dimensions of drivers’ performance that may estimate the effect of a secondary task on driving in general cases and to exemplify the promising usage of multivariate analysis methods in this area. They brought out “low workload but high inattentiveness”, and “peripheral insensitivity” corresponding to PC2 and PC3 that can be very promising dimensions in studying the secondary task. However, according to the authors, the findings are limited by the level of training for the secondary tasks, which was near the maximum of the short term learning curve.
At the same time, there are also some points showing that this study is incomplete. First, after finding and explaining the three PCs, the paper did not test its explanation by other data, especially the data from driving-alone situation. For example, they argued that the 2nd PC presented that drivers pay more attention to the secondary tasks, but less attention to the outside road than expected when they were doing the secondary tasks. But they did not provide the comparisons with the driving-alone situation regard to inattention to the outside world and too much attention to the secondary tasks. Without this kind of evidence, it is hard to say that the three PCs can capture the underlining dimensions of driver performance and can be used to study the effects of secondary tasks on driving. The three PCs may be only a special case in this specific study. Second, the authors did not provide the reason of why PCA was chosen in this study and the comparison to other similar methods, such as factor analysis (FA). On the contrary, FA more fits in this application because FA is to find the fundamental factors of variables while PCA is to try to explain the variance among variables. Third, the authors did not give the reasons why these 15 measures of driver performance were particularly interested since there are many other performance measures, such as average speed, glances to the mirrors and so on. The authors also argued that the first eight measures were driver workload variables. I do not see the reasons of this classification. Forth, the three PCs are also far from seen as the complete assessment of driver performance because the 15 measures may not be complete assessment of driver performance and the three PCs leave 17% variance of current measures out.
Richard A. Young & Linda S. Angell (2003) The dimensions of driver performance during secondary manual tasks. Proceedings of the second international driving symposium on human factors in driver assessment, training and vehicle design. 98-112
