Driving and Multitasking: The Good, the Bad, and the Dangerous

So you have an idea of what it will take for you to drive post-stroke. Your doctor will have no clue other than to reflexively say you can't drive. I've been driving 8 years now, even drive in Chicago rush hour traffic, not fun but doable.

Driving and Multitasking: The Good, the Bad, and the Dangerous

Menno Nijboer¹, Jelmer P. Borst¹, Hedderik van Rijn² and Niels A. Taatgen^1*

• ¹Department of Artificial Intelligence, University of Groningen, Groningen, Netherlands
• ²Department of Psychology, University of Groningen, Groningen, Netherlands

Previous research has shown that multitasking can have a positive or a negative influence on driving performance. The aim of this study was to determine how the interaction between driving circumstances and cognitive requirements of secondary tasks affect a driver's ability to control a car. We created a driving simulator paradigm where participants had to perform one of two scenarios: one with no traffic in the driver's lane, and one with substantial traffic in both lanes, some of which had to be overtaken. Four different secondary task conditions were combined with these driving scenarios. In both driving scenarios, using a tablet resulted in the worst, most dangerous, performance, while passively listening to the radio or answering questions for a radio quiz led to the best driving performance. Interestingly, driving as a single task did not produce better performance than driving in combination with one of the radio tasks, and even tended to be slightly worse. These results suggest that drivers switch to internally focused secondary tasks when nothing else is available during monotonous or repetitive driving environments. This mind wandering potentially has a stronger interference effect with driving than non-visual secondary tasks.

Introduction

There is a general belief that driving cannot be combined with any other task without affecting driving performance. Several studies have found evidence for this (Ranney et al., 2000), ranging from phone conversations (Strayer and Johnston, 2001; Treffner and Barrett, 2004) to music listening (Brodsky, 2001). However, recent evidence has indicated that multitasking could also be beneficial for driving when the right circumstances are met (Gershon et al., 2009; Atchley and Chan, 2010; Ünal et al., 2012). In this work we develop some theoretical explanations for both decrease and increase in performance, and test these in an experiment for which we predict to see both effects.

Multitasking Interference in Driving

When two tasks require the same perceptual or cognitive resource at the same time, they are said to overlap with regards to that resource. Overlap in resource use between concurrently performed tasks leads to contention for those resources (Pashler, 1994; Wickens, 2002; Salvucci and Taatgen, 2008). In turn, this contention typically leads to reduced task performance (e.g., Just et al., 2008; Borst et al., 2010b; Strayer et al., 2013; Nijboer et al., 2014). For our purposes, we will describe overlap in terms of the resources defined in the threaded cognition theory (Salvucci and Taatgen, 2008), which offers an account that is precise in terms of timing, and that is based on resources defined in the ACT-R cognitive architecture (Anderson, 2007). The resources that are most relevant for driving (and secondary tasks next to driving) are the visual perception, auditive perception, declarative memory, working memory and motor control. Although, driving requires all of these resources to some degree (Herbert, 1963; Anstey et al., 2005), the demands on these resources vary depending on the traffic situation. For example, driving on a quiet road mainly requires visual perception and motor control. Moreover, resources are not always required full-time: it is acceptable to look away from the road for short periods of time. According to threaded cognition, resources are assigned to tasks based on two principles: greediness and politeness. The greediness principle states that a task can be used when it is not used by any other task at a particular moment, but has to wait if the resource is in use. The politeness principle states that a task should release a resource as soon as it is not needed anymore. For example, routine driving does not require the use of working memory. Therefore, a secondary task such as having a conversation that does require working memory does not interfere with driving. However, if the driving situation changes in a way that requires working memory, for example planning how to cross a complex intersection, the conversation task may interfere with driving because it does not relinquish working memory soon enough.

Of all the resources that have been studied with respect to driving interference, perceptual and motor interference have been studied most (visual and auditory perception: Chaparro et al., 2005; Gherri and Eimer, 2011; manipulation of equipment: Brookhuis et al., 1991; Briem and Hedman, 1995; Janssen et al., 2012). Cognitive requirements of secondary tasks turned out to be at least as important. Of all tasks found to interfere with driving, cell-phone use has received most attention due to the high number of traffic-accidents attributed to such devices (Redelmeier and Tibshirani, 1997). In an influential study, Strayer and Johnston (2001) showed that it is primarily what they call the attentional component of holding a conversation that disrupts driving performance, by ruling out explanations related to holding the phone, speaking, or listening. Several studies have shown that holding a complex conversation in particular affects driving performance (McKnight and McKnight, 1993; Briem and Hedman, 1995).

A resource that is pivotal in large disruptions of performance in multitasking is working memory. Note that we use a restricted concept of working memory, the part that Baddeley (2012) calls the central executive, and Oberauer (2002) the focus of attention. It is therefore closely related to Strayer and Johnston's (2001) attentional component. In the threaded cognition theory, focal working memory can hold a single chunk of information. This chunk can, in turn, point to other sources to create a larger context, but it is the only element that is available for immediate information processing. Any other element needs some retrieval or recovery process to use.

Working memory is used to build up temporary representations that are needed in the near future, for example the gist of a conversation (e.g., van Rij et al., 2010, 2013) or to represent the result of partial computations in arithmetic (Borst et al., 2010a). Secondary tasks in driving experiments that involve working memory (e.g., Alm and Nilsson, 1995) typically lead to decrements in both driving performance and performance on the secondary task. The working-memory load of driving is strongly dependent on the traffic: when the road is empty the driver only has to remember information regarding the current state of the car, which can be easily retrieved from visual and aural queues that are constantly present in the environment. When there is substantial traffic, however, the driver has to keep a detailed mental model of the surrounding vehicles (Gugerty, 1997), as these will not always be visible: they might reside in the blind spot of the car, or be obscured by other vehicles.

Whenever there is a situation in which driving suddenly requires working memory, the driver has to give up the contents of working memory for the secondary task, which can lead to severe disruptions in that task. Therefore, the driver may be reluctant to do so, leading to possible dangerous decisions.

Given that accidents in both real and simulated driving are rare, we need a different measure of driving quality. In Alm and Nilsson's (1995) study, the subjects had to follow a leading car, and had to respond to that car's breaking by breaking. The response time was a measure of driving quality. Gershon et al. (2009) used a number of measures for driving quality: lateral deviation from the middle of the lane, where more deviation is associated with less attentive driving, standard deviation in speed, where more deviation indicates the driver pays less attention, and standard deviation in steering angle, where larger angles are indicative for poorer driving. We will use these indicators, with some refinements, in our own study. We will also look at overtaking behavior, where we will use consistency in overtaking distance as a measure of quality, as well as proper turn-signal use, and, given that subjects sometimes do hit other cars, the frequency with which this occurs.

More at link.