Prosocial AI Apologies on the Road: Emotional Compensation for Other Drivers' Misbehavior

Paper Title

Prosocial AI Apologies on the Road: Emotional Compensation for Other Drivers' Misbehavior

Publication Info

• Topic area: AI-mediated emotional regulation in traffic interactions
• Keywords: AI apologies, prosocial lies, emotional regulation, traffic safety, AR-HUD, driving behavior, social proxy, forgiveness, anger mitigation, intelligent transportation systems

Background and Problem

• Problem / challenge: Traditional communication mechanisms in traffic (e.g., gestures) are ineffective due to physical barriers and high-speed dynamics, leaving victims of driving violations without emotional repair. Current systems lack automated, effective methods to mitigate negative emotions and repair social relations in traffic.
• Significance: Emotional responses like anger and retaliatory behaviors in traffic increase safety risks. Addressing these emotions through AI-mediated apologies could enhance traffic safety and social harmony.
• Motivation and related work: Prior research has shown the benefits of apologies and AR-HUD systems in reducing road rage and fostering empathy. However, these systems often rely on the offender’s intent, which limits their effectiveness when the offender lacks responsibility or empathy. This study introduces an AI-driven apology mechanism to bridge this gap.

Solution

• Proposed approach: An AI-based "Social Proxy" system that delivers apologies on behalf of offending drivers using an AR-HUD interface, employing prosocial lies to mitigate negative emotions and promote forgiveness.
• Novelty:
  1. Introduction of an AI-driven "Social Proxy" mechanism for traffic conflict resolution.
  2. Empirical evaluation of prosocial lies as a cognitive intervention tool in traffic scenarios.
  3. Systematic analysis of apology depth and its impact on emotional relief and forgiveness.
• Procedure and key techniques:
  • Developed a multimodal HMI prototype integrating visual (AR-HUD icons) and auditory (voice messages) apology cues.
  • Designed five levels of apology depth: No Apology (NA), Regret Only (RO), Regret + Responsibility (RR), Regret + Responsibility + Explanation (RRE), and Full Apology (FA).
  • Conducted a 2x5 mixed-design driving simulator experiment with 40 participants to evaluate the system in low- and high-risk traffic scenarios.

Results

• Concrete findings:
  • AI apologies significantly reduced anger and increased forgiveness across both low- and high-risk scenarios.
  • Apologies with richer content (RR, RRE, FA) were more effective than simple apologies (RO) or no apology (NA).
  • Participants derived psychological benefits and exhibited high acceptance of the AI-mediated prosocial lies.
• Advantage over baselines:
  • RR, RRE, and FA significantly outperformed NA in mitigating anger and enhancing forgiveness (e.g., PANAS: FA = 3.00 vs. NA = -1.90, p <.001).
  • RRE and FA demonstrated the highest acceptance and perceived psychological benefits compared to simpler strategies.
• Experiments / evaluation:
  • Participants (N = 40) experienced two risk scenarios (low-risk: slow vehicle; high-risk: forced cut-in) and five apology depths in a driving simulator.
  • Emotional states, anger, forgiveness, acceptance of communication strategies, and psychological benefits were measured using validated scales (e.g., PANAS, ARS, HFS).
• Limitations and future work:
  • Limited to two risk scenarios; broader ecological validity is needed.
  • Conducted in a simulator rather than real-world settings.
  • Did not address offender behavior or long-term effects; future work should explore accountability mechanisms and longitudinal studies.

Summary

This study introduces an AI-driven apology system that uses prosocial lies to mitigate negative emotions in traffic conflicts. Through a driving simulator experiment, the system was shown to reduce anger, increase forgiveness, and provide psychological benefits, particularly when using richer apology content (e.g., explanations and commitments). While participants accepted the system’s emotional regulation benefits, concerns about offender accountability were raised. The findings highlight the potential of AI as a social proxy in intelligent transportation systems, offering a pathway for safer and more empathetic traffic interactions.