Huei-Yen Winnie Chen

What drives technology-based distractions? A structural equation model on social-psychological factors of technology-based driver distraction engagement

BACKGROUND With the proliferation of new mobile and in-vehicle technologies, understanding the motivations behind a drivers voluntary engagement with such technologies is crucial from a safety perspective, yet is complex. Previous literature either surveyed a large number of distractions that may be diverse, or too focuses on one particular activity, such as cell phone use. Further, earlier studies about social-psychological factors underlying driver distraction tend to focus on one or two factors in-depth, and those that examine a more comprehensive set of factors are often limited in their analyses methods. OBJECTIVE The present work considers a wide array of social-psychological factors within a structural equation model to predict their influence on a focused set of technology-based distractions. A better understanding of these facilitators can enhance the design of distraction mitigation strategies. METHOD We analysed survey responses about three technology-based driver distractions: holding phone conversations, manually interacting with cell phones, and adjusting the settings of in-vehicle technology, as well as responses on five social-psychological factors: attitude, descriptive norm, injunctive norm, technology inclination, and a risk/sensation seeking personality. Using data collected from 525 drivers (ages: 18-80), a structural equation model was built to analyse these social-psychological factors as latent variables influencing self-reported engagement in these three technology-based distractions. RESULTS Self-reported engagement in technology-based distractions was found to be largely influenced by attitudes about the distractions. Personality and social norms also played a significant role, but technology inclination did not. A closer look at two age groups (18-30 and 30+) showed that the effect of social norms, especially of injunctive norm (i.e., perceived approvals), was less prominent in the 30+ age group, while personality remained a significant predictor for the 30+ age group but marginally significant for the younger group. CONCLUSION Findings from this work provide insights into the social-psychological factors behind intentional engagement in technology-based distractions and in particular suggesting that these factors may be sensitive to demographic differences.

Measuring Inhibitory Control in Driver Distraction

Driver distraction research primarily focuses on voluntary distraction. Little known research explicitly evaluates driver susceptibility to involuntary distractions. This paper investigates the relationships between glance behavior in response to irrelevant stimuli in a driving simulator and measures of inhibitory control assessed through a modified flanker task. Overall, inhibitory control appears to be a mechanism that relates to number of glances and average glance duration. Data from 16 participants show that smaller flanker compatibility effects (i.e., better inhibitory control) are significantly associated with fewer glances and shorter average glance durations to irrelevant stimuli in the simulator. No significant relation was found between time fixation on the irrelevant stimulus after its onset and the size of the flanker compatibility effect.

Determining Fixed Glance Duration for Visual Occlusion Research

Little empirical evidence can be found in the visual occlusion literature to justify decisions about how glance durations should be fixed for self-paced visual occlusion investigations. This paper presents a hypothesis about how glance duration may affect performance, based on a theory of how uncertainty develops as an operator’s vision is occluded and how it is resolved during visual glances. Data are analysed from two on-road driving experiments involving a range of fixed glance durations. The analysis is repeated with data collected from an analogous study in a low fidelity driving simulator. Both analyses support the hypothesis that increasing glance duration may prolong achievable mean occlusion times, but only up to a certain point, after which essentially no changes are expected. The paper concludes with a practical recommendation for selecting fixed glance durations for (self-paced) visual occlusion studies.

Smart Driver Monitoring: When Signal Processing Meets Human Factors: In the driver's seat

This article provides an interdisciplinary perspective on driver monitoring systems by discussing state-of-the-art signal processing solutions in the context of road safety issues identified in human factors research. Recently, the human factors community has made significant progress in understanding driver behaviors and assessed the efficacy of various interventions for unsafe driving practices. In parallel, the signal processing community has had significant advancements in developing signal acquisition and processing methods for driver monitoring systems. This article aims to bridge these efforts and help initiate new collaborations across the two fields. Toward this end, we discuss how vehicle measures, facial/body expressions, and physiological signals can assist in improving driving safety through adaptive interactions with the driver, based on the drivers state and driving environment. Moreover, by highlighting the current human factors research in road safety, we provide insights for building feedback and mitigation technologies, which can act both in real time and postdrive. We provide insights into areas with great potential to improve driver monitoring systems, which have not yet been extensively studied in the literature, such as affect recognition and data fusion. Finally, a high-level discussion is given on the challenges and possible future directions for driver monitoring systems.

Test–Retest Reliability of the Susceptibility to Driver Distraction Questionnaire

The Susceptibility to Driver Distraction Questionnaire (SDDQ) investigates voluntary and involuntary factors associated with driver distraction. The questionnaire consists of 39 items in six subscales: (a) self-reported distraction engagement, (b) attitudes toward distractions, (c) perceived control of driving while engaged in distractions, (d) injunctive social norms associated with distraction engagement, (e) descriptive social norms associated with distraction engagement, and (f) susceptibility to involuntary distractions. A sample of 43 adults, ages 25 to 39 years, was used to assess the test–retest reliability of the SDDQ. The mean time between test and retest conditions was approximately 20 days. For sub-scale averages, intraclass correlation (ICC) statistics were used to assess test–retest reliability; weighted kappa statistics were used to assess individual items. The ICC results suggest good to excellent test–retest reliability for subscales of self-reported distraction engagement, attitudes toward distractions, and descriptive social norms. Perceived control of driving while engaged in distractions had fair test–retest reliability, and the injunctive norms and susceptibility to involuntary distraction subscales had poor test–retest reliabilities. The latter two subscales may have to be redesigned; this paper provides relevant suggestions for revision in the discussion of the results. As an additional preliminary analysis, data from a sample of 10 additional participants were used to investigate the consistency of responses across longer periods of time. The mean time between test–retest conditions in this sample was approximately 8 months. In general, the findings were similar to those in the main sample. Overall, the SDDQ appears to have good test–retest reliability. A larger sample is recommended for further validation of these results, in particular across long test–retest periods.

A FRAMEWORK FOR MODELLING AND ANALYSING VARIABILITY IN VISUAL OCCLUSION EXPERIMENTS

Research using self-paced visual occlusion has traditionally analysed mean occlusion times, thereby neglecting potential insights to be gained from variability across individual visual sampling decisions. This paper proposes a framework for analysing visual occlusion data based on a hierarchy of sampling strategies. The framework describes each sampling decision as being dependent on both system characteristics (mean performance) and information available during sampling (variability). To illustrate the framework, data from an on-road study were analysed. Self-paced occlusion times were shown to fit a descriptive function both for lane deviations observed at the end of previous visual samples and for predicted lane deviations at the end of occlusion intervals. The fact that the latter fit was better suggests that participants, especially the more experienced ones, were indeed able to use predictions in their sampling decisions.

Simulator Study of Involuntary Driver Distraction Under Different Perceptual Loads

Involuntary distraction, which occurs when driver attention is diverted unintentionally by irrelevant stimuli or events, is often overlooked in experimental studies. The present work explores how involuntary distraction affects individual drivers and whether varying perceptual load in the driving environment modulates involuntary distraction engagement. In a simulator experiment, variability in glance behavior toward irrelevant stimuli was observed among participants, and higher self-reported everyday distractibility scores using the Cognitive Failures Questionnaire were found to be associated with longer glances, but not the number of glances, toward the irrelevant stimuli. These relationships suggest that the Cognitive Failures Questionnaire scale may correlate better with the ability to disengage from a distraction than with the ability to suppress automatic attentional capture. The study also found delayed accelerator release times to lead vehicle braking events in the presence of irrelevant stimuli. The perceptual responses associated with the accelerator release times show that the delay occurred after participants glanced at the brake light, possibly indicating slower processing of information under distraction. Contrary to expectation, perceptual load, manipulated by the simulated road’s visual complexity, did not affect involuntary distraction engagement but directly affected driving performance. Overall, findings reveal potential safety concerns for involuntary driver distraction, but further work is necessary to understand how individuals with different attentional limitations are affected by this distraction type.

Gaming to Safety Exploring Feedback Gamification for Mitigating Driver Distraction

The risk of crash or near-crash significantly increases when drivers make long off-road glances to engage in distracting tasks. Providing drivers with feedback that integrates elements of game design could increase driver motivation for adopting safer behaviors. In an ongoing between-subjects simulator study with young drivers (n=29 reported in this paper), we compare four conditions for off-road glance behaviors: no feedback, real-time feedback system, post-drive feedback system (real-time feedback + post-drive feedback), and gamification feedback system (real-time feedback + post-drive feedback + game design elements). Shorter average glance duration and less frequent risky (≥2 s) glances to an in-vehicle display were observed for the post-drive system, compared to no feedback and real-time feedback, and for the gamification system, compared to no feedback. Although no added benefit of gamification over the post-drive feedback system was observed for these eye glance metrics, longer-term exposure and assessment could show improvements to be more stable with the inclusion of game design elements.

A Model of Expert Decision Making in Post-Flop Betting in Poker

Most past research in modelling poker arises from the field of artificial intelligence, based on the many rules and probabilities involved in the game. The present study examines poker from a new perspective: applying knowledge elicitation techniques to model some of the decision making processes of expert poker players in post-flop betting. Based on data obtained through observations and interviews, using expert players as study participants, a decision making model is proposed, together with an abstraction hierarchy model. It was found that poker players create a set of mental models of their opponents, the active game situation, and of themselves as perceived by their opponents, in order to achieve the purposes of always making better decisions than their opponent and, whenever possible, maximizing the consequences of the opponents mistakes. A set of strategies that are independent of specific situations and individual players were also discovered in this study.

Prevalence of Engagement in Single versus Multiple Types of Secondary Tasks: Results from the Naturalistic Engagement in Secondary Task (NEST) Dataset

We investigated engagement in single vs. multiple types of secondary tasks in distraction-affected, safety-critical events (SCEs), i.e., crashes/near-crashes, and baselines reported in the Naturalistic Engagement in Secondary Tasks (NEST) dataset. NEST was created from Second Strategic Highway Research Program (SHRP2) data for studying distractions in detail. Early descriptive analysis on NEST found that most distraction-affected SCE and baseline epochs (10 s long) include more than one type of secondary task, suggesting that a considerable number of drivers may be engaging in multiple secondary activities within a relatively short time frame, potentially being exposed to increased demands brought on by multi-tasking and task-switching. We conducted inferential statistics on NEST focusing on engagement in single vs. multiple types of tasks across SCEs and baselines. A logit model was built to compare the odds of engaging in single vs. multiple types of tasks with the following predictors: event type (SCE, baseline), environmental demand, GPS speed, and driver age. The last three predictors were included to capture the driving demands experienced, which may have impacted drivers’ task engagement behavior. Odds of engagement in multiple types of secondary tasks was higher in SCEs than baselines. Furthermore, with marginal statistical significance, drivers 65 years and over were less likely to engage in multiple types of secondary tasks than younger drivers. Overall, engagement in multiple secondary task types is more prevalent in SCEs. Most crash risk studies to date have reported the effects associated with one type of secondary task. However, it appears that these effects may be confounded by the presence of other secondary tasks.