Effects of cognitive tasks on car drivers’ behaviors and physiological responses

Doctoral thesis, 2022

The effects of drivers’ engagement in cognitive tasks (i.e., non-visual, cognitively loading activities unrelated to the task of driving) are debated and unclear. Numerous experiments show impaired driver behaviors, yet naturalistic studies typically do not find an increased crash risk. In the future, autonomous driving (AD) is expected to improve traffic safety while allowing safe engagement in cognitive (and other) tasks. Having the opportunity to perform non-driving related tasks while traveling may then motivate drivers to use AD, provided they can actually engage in the tasks. Unfortunately, research on drivers’ engagement in cognitive tasks suffers severe methodological limitations since reliable and unintrusive measures of cognitive load are lacking.

The aim of this thesis is therefore to advance the understanding of task-induced cognitive load in the context of traffic safety. This aim is split into two objectives: A) to better understand how drivers’ involvement in cognitive tasks can affect safety-relevant driver behaviors and decisions and B) to provide methodological guidance about assessing cognitive load in drivers using physiological measures.

To accomplish Objective A, effects of cognitive tasks on driver behaviors were studied during routine driving and in a safety-critical event in a driving simulator. Also, drivers’ ability to engage in a non-driving related task while using AD in real traffic was explored. In line with the cognitive control hypothesis (Engström et al., 2017), it was found that cognitive tasks negatively affected driver behaviors in situations where cognitive control was needed, for example in intersections—but not in a lead vehicle braking scenario where responses were triggered automatically by visual looming. It was also found that although the number of off-path glances decreased during cognitive load, the timing of the remaining glances was unaffected. Clearly, cognitive load has different effects on different mechanisms. When using AD, drivers were indeed capable of engaging in a non-driving related task—suggesting that AD will be able to fulfill drivers’ desire to perform such tasks while traveling, which may motivate AD usage and thus improve traffic safety (given that AD is truly safer than manual driving). Finally, a simulator study addressing Objective B showed that the measurability of cognitive load was greatly improved by recognizing that multiple coexisting mental responses give rise to different physiological responses. This approach can provide less context-dependent measurements and allows for a better, more detailed understanding of the effects of cognitive tasks.

These findings can help improve traffic safety—both by being used in system development, and as part of the systems themselves.