Actual Achieved Gain and Optimal Perceived Gain: Modeling Human Take-over Decisions Towards Automated Vehicles' Suggestions

TL;DR

The paper introduces Actual Achieved Gain (AAG) and Optimal Perceived Gain (OPG) to quantify driver decision quality during take-overs in semi-automated driving, framing decisions as weighted gains/losses influenced by ADS accuracy. Through Study 1, it establishes a Thurstone-scale-based measure of perceived gains/losses across route selection, overtaking, and collision avoidance, then validates AAG/OPG in Studies 2 and 3 under varying decision times and ADS accuracies. The findings show that with sufficient decision time, AAG converges toward OPG, indicating rational decision-making, while time pressure induces conservative or following behaviors and irrational patterns. Study 4 demonstrates AAG-based interventions (voice and multimodal alerts) can improve alignment between AAG and OPG and enhance decision accuracy, suggesting practical paths to improve human-ADS collaboration in real-world take-over scenarios.

Abstract

Driver decision quality in take-overs is critical for effective human-Autonomous Driving System (ADS) collaboration. However, current research lacks detailed analysis of its variations. This paper introduces two metrics--Actual Achieved Gain (AAG) and Optimal Perceived Gain (OPG)--to assess decision quality, with OPG representing optimal decisions and AAG reflecting actual outcomes. Both are calculated as weighted averages of perceived gains and losses, influenced by ADS accuracy. Study 1 (N=315) used a 21-point Thurstone scale to measure perceived gains and losses-key components of AAG and OPG-across typical tasks: route selection, overtaking, and collision avoidance. Studies 2 (N=54) and 3 (N=54) modeled decision quality under varying ADS accuracy and decision time. Results show with sufficient time (>3.5s), AAG converges towards OPG, indicating rational decision-making, while limited time leads to intuitive and deterministic choices. Study 3 also linked AAG-OPG deviations to irrational behaviors. An intervention study (N=8) and a pilot (N=4) employing voice alarms and multi-modal alarms based on these deviations demonstrated AAG's potential to improve decision quality.

Figures (7)

• Figure 1: The experiment platform for Study 2 and 3. The participants took a first-perspective view. (1) The system told participants of ADS's accuracy before the start of automated driving using voice. (3) The system told participants the suggestions of ADS when driving near the take-over intersection using voice. (4) The alarm indicating the time ran out in Study 3. When driving near the take-over intersection, the alarm rang. The "0 D" on the screen illustrated 0 miles per hour while "D" indicated gear level (forward gear).
• Figure 2: Illustration of the different take-over tasks on the experiment platform. The arrows and text were added manually for clarification and were not part of the original system.
• Figure 3: (a) AAG and OPG in different tasks with different accuracy. Errorbar indicated one standard deviation. The lighter bar indicated OPG while the darker bar indicated AAG. (b) The correlation between AAG and ground truth accuracy, where red share indicates 95% confidence interval.
• Figure 4: Follow rate and conservative rate for different tasks and ADS accuracy. Error bar indicated one standard deviation.
• Figure 5: (a) AAG and OPG with different tasks with different time. Errorbar indicated one standard deviation. (b) The regression plot of AAG and real world decision accuracy. The red shade indicated 95% confidence interval.