Abstract
The current study investigated the effects of secondary task relevance on measures of vigilance decrement in a Level 2 semi-autonomous driving simulation. Past research has demonstrated that over the course of a vigilance task, vigilance (or sustained attention) decreases which leads to worse performance on measures like reaction time and accuracy. This phenomenon is known as the vigilance decrement. However, the underlying cause of the vigilance decrement is under debate. Secondary tasks can not only shed light on this debate but they might also potentially help mitigate this vigilance decrement if designed and implemented properly. Therefore the current research used secondary tasks to manipulate task demand and task engagement to further investigate the cause of the vigilance decrement.
Keywords
Despite more advanced autonomous abilities in vehicles, drivers must stay vigilant during a monotonous task for an event that rarely occurs. Driving, whether manual or automated, is known as a vigilance task that requires the human to sustain their attention to detect rare and infrequent events (Greenlee et al., 2018; Körber et al., 2015; Mishler & Chen, 2023). As a vigilance task continues, human performance typically declines. This phenomenon is known as the vigilance decrement (Parasuraman, 1986). The overload and underload theories are two contrasting explanations for this phenomenon. The resource theory, or the overload explanation, posits that the vigilance decrement results from limited cognitive resources being depleted as time goes on (Parasuraman, 1986; Parasuraman & Davies, 1977; Warm et al., 2008). The mindlessness theory, or the underload explanation, proposes that long, monotonous tasks result in disengagement and the vigilance decrement (Manly et al., 1999). The present study examined the effects of vigilance in partially automated driving and methods of mitigation via secondary tasks to address the overload versus underload debate.
The study included data from 116 participants via a partially automated driving simulator. Following a practice drive, participants drove on a two-lane road with a stream of oncoming cars in the opposite lane. Participants notified the automation via button press whenever an oncoming vehicle briefly crossed into the driver’s lane. Out of the 270 vehicles that passed by, 68 committed this hazardous move over the drive.
Participants verbally answered audible yes/no questions as a secondary task. The type of questions was the between-subjects independent variable. In addition to a control condition with no secondary tasks, there were three secondary-task conditions: non-driving related (NDR), non-visual driving related (NVDR), and visual driving related (VDR). The questions used in the NDR condition consisted of general knowledge questions unrelated to driving. The questions in the NVDR condition were related to the act of driving but required no visual input to correctly respond. The questions in the VDR condition cued participants to look at a specific region of the visual field to be able to answer the questions. Each condition was designed to either increase task engagement via task-relevant questions (VDR), decrease the demand on limited cognitive resources via questions that did not require visual input (NDR), or both (NVDR). While participants experienced a seamless 45-minute session, the drive was split into four equal periods of watch (POW 1-4) for measuring the vigilance decrement over time, which created the within-subjects independent variable. The dependent variables included participants’ response time (RT) to dangerous events and their hit rate (HR) in identifying them. In addition, subjective reports of engagement, distress, and worry were measured via the Short Stress State Questionnaire (SSSQ; Helton, 2004) and global workload was measured using the NASA-TLX (Hart & Staveland, 1988) at the end of the experiment.
The results of this study showed significant effects of POW on both RT and HR, indicating that participants took longer to respond and noticed fewer hazardous events as time progressed. This result further supports research demonstrating that drivers in partially automated vehicles are subject to the vigilance decrement (Greenlee et al., 2018, 2019; Mishler & Chen, 2023). However, no significant main effects of secondary-task condition, nor any significant interactions were found. Overall, the results do not seem to support either the overload or underload explanation alone. Support for one or the other would require a significant interaction between the two IVs. The overload theory would be strongly supported by an interaction showing that vigilance performance decreased as task demands increased (control being the lowest followed by NDR, NVDR, and VDR sequentially) and time on task increased. On the other hand, an interaction indicating that vigilance performance increased with task engagement (control being the lowest followed by NDR, NVDR, and VDR being the highest) as POW progressed from POW 1 to 4, would be used in favor of the underload theory.
While the current results do not appear to support either school of thought alone, they do seem to be in line with a theory like the resource-control account of sustained attention. This theory proposes that all thoughts, both task related and irrelevant, draw attentional resources from the same limited supply and are directed by a depleting executive control (Thomson et al., 2015). Therefore, the vigilance decrement would be influenced by both resource depletion and disengagement/mind-wandering (see Mishler & Chen, 2023). So, all thoughts (driving-relevant or not) would cause comparable declines in resources, which could explain the similar results found in this study. However, considering the lack of statistical significance, future studies are needed to further confirm the current results.
Footnotes
Acknowledgements
This work was supported by the National Science Foundation award # 2245055.
We would also like to thank April Tong, Brandon Huang, Francesca Bliss, and Meera Borle for their help in data collection.
