Effects of Time Headway on Driving after Prolonged Automation Exposure: Evidence for Behavioral Adaptation

Abstract

The shift of a driver’s role from an active operator to a passive supervisor of an automated vehicle (AV) can result in adverse behavioral adaptation effects (BAEs). Prior studies showed that after drivers were exposed to an AV that adopts an extremely small time headway for a prolonged duration, drivers adopted shorter time headways than what they would normally use. However, the effect of an AV’s headway on BAEs has not been examined. The current study conducted two experiments to determine whether (a) the effect of exposure duration (15, 25 min) on a driver’s headway post-automation depends on AVs’ headways (0.3 s, 0.6 s), and (b) a driver’s trust in the AV varies with AVs’ headways (0.3 s, 0.6 s, 1.5 s). Results revealed that trust was positively correlated with AV headway and a significant number of drivers reduced their headways after regaining control of the vehicle, a result consistent with prior BAE research.

Behavioral adaptation effects (BAEs) occur when users mimic sustained environmental changes, such as an automated vehicle’s (AV) small time headway (OECD, 1990). BAEs are problematic for traffic efficiency solutions that use a convoy of AVs (i.e., platooning) that maintain small time headways because drivers reduce their time headway after prolonged exposure (Brandenburg & Skottke, 2014). However, Dudley and DeLucia (2024) failed to replicate this finding, potentially due to differences among drivers’ evaluation ratings of the AV’s performance. Drivers who gave higher evaluations reduced their headways, whereas drivers who gave lower evaluations did not. The authors proposed that drivers’ evaluations of the AV’s performance were proxy measures for trust but could not confirm because they, and prior work (Brandenburg & Skottke, 2014), did not collect trust.

We report the results of two experiments to assess whether a driver’s (a) trust in an AV depends on the AV’s headway and (b) susceptibility to BAEs depends on the AV’s headway and the duration of exposure to the AV. We recruited 18 licensed drivers with normal or corrected visual acuity to monitor three AV systems that differed in headway (0.3 s, 0.6 s, 1.5 s) for five 3-minute traffic scenes. After each scene, drivers answered a single-item trust-questionnaire (“How much do you trust automated driving systems”; 1: Not at all, 7: Extremely; Manchon et al., 2022) and a single-item evaluation questionnaire (“How well would you rate the automated driving system’s performance of following the lead vehicle” ;1: Poor; 5: Great). The evaluation question was used to validate Dudley and DeLucia’s (2024) proposal that ratings of the AV’s performance were proxy measures for trust.

Trust was analyzed with a one-way (Time Headway: 0.3 s, 0.6 s, 1.5 s) repeated-measures ANOVA and revealed a significant main effect of headway, p < .001. A significant linear contrast indicated a positive relationship between trust and headway, p < .001. This suggests that drivers reported significantly greater trust when the AV used larger headways. Additionally, drivers’ evaluation and trust ratings were significantly correlated, p < .001, r = .81. This correlation confirms that evaluation ratings of the AV’s performance were proxy measures for trust (Dudley & DeLucia, 2024) and provides evidence that trust depends on the AV’s headway.

In a separate experiment, we assessed the influence of the AV’s headway and exposure duration on BAEs. Based on a power analysis conducted in G*Power with a medium effect size (f = 0.25; prior literature did not report), 28 drivers were recruited. Drivers completed a 35- or 45-minute car-following task at 60 mph. Drivers were told to drive as they normally would, obey all rules of the road, avoid crashes, and maintain a constant distance to the lead vehicle. Drivers followed a lead vehicle manually for 10 minutes to establish their baseline headway. Then drivers activated the driving automation and monitored the AV for 15 or 25 minutes. Subsequently, they regained manual control and followed the lead vehicle for 10 minutes. The post-automation drive was segmented into continuous 1-minute intervals to assess how drivers’ headways changed over time after they regained control. Lastly, drivers answered a trust-in-automation questionnaire after completing the experiment (Manchon et al., 2022).

A 2 (Duration: 15 min, 25 min; between) × 2 (AV Headway: 0.3 s, 0.6 s; between) × 10 (Interval: 0–1 min, 1–2 min, . . ., 9–10 min; within) mixed ANCOVA with trust as a covariate revealed no significant main effects, ps > .05. There was a significant interaction between AV headway and duration, p = .044, suggesting that the effect of duration depended on the AV’s headway. Post-hoc contrasts revealed that in the 0.3 s headway condition, the 25-minute duration increased drivers’ headways (M = 0.09), whereas the 15-minute duration decreased drivers’ headways (M = −0.38). In contrast, in the 0.6s headway condition, the 25-minute duration decreased drivers’ headways (M = −0.59) whereas the 15-minute duration increased drivers’ headways (M = 0.05), p = .044. No other interactions were significant.

Our experiment failed to replicate BAEs, potentially due to insufficient power. A post-hoc power analysis with the calculated effect size (f = 0.21) and Hyund-Felt correction (HF = .49) suggested that 48 participants would be needed to detect significant differences. This larger sample size is consistent with other between-subject experimental designs (Brandenburg & Skottke, 2014). However, a binomial sign test indicated that the number of drivers in our study who reduced their headways was significantly different from chance (i.e., null hypothesis that 50% of drivers would reduce their headways and 50% would not), p = .03. Additionally, the results from our significant interaction were contrary to BAE literature (OECD, 1990); specifically, an increased exposure duration should make drivers more likely to reduce their headways. However, this was not the case for the .3 s headway condition. A more robust experiment (e.g., sufficient sample size) will be conducted to assess if there are differences in behavioral adaptation effects as a function of (a) the duration of exposure (Dudley & DeLucia, 2024) and (b) the AV’s headway.