To Cooperate or Not to Cooperate: A Systematic Review and Meta-Analysis of Human Driving Behavior in Interactions with Autonomous Vehicles

Paper Title

To Cooperate or Not to Cooperate: A Systematic Review and Meta-Analysis of Human Driving Behavior in Interactions with Autonomous Vehicles

Publication Info

• Topic area: Human driving behavior in interactions with autonomous vehicles (AVs) and its implications for traffic safety and efficiency.
• Keywords: Human-driven vehicles, autonomous vehicles, cooperation, driving behavior, meta-analysis, traffic safety, human–AI interaction, ecological validity, AV trajectory planning, traffic regulations.

Background and Problem

• Problem / challenge: Human drivers (HVs) exhibit inconsistent levels of cooperation when interacting with autonomous vehicles (AVs), with prior studies providing mixed and inconclusive findings. Existing reviews lack systematic methodologies and fail to address key moderators, limiting their ability to provide precise estimates or generalizable insights.
• Significance: Understanding HV–AV cooperation is critical for the safe and efficient integration of AVs into traffic systems. Uncooperative behavior can lead to safety risks, inefficiencies, and hinder AV adoption.
• Motivation and related work: Previous narrative reviews and empirical studies have reported conflicting results—some showing less cooperation with AVs, others showing similar or even greater cooperation compared to HVs. However, these studies often lacked control groups, focused on limited scenarios, or used non-behavioral measures. This paper addresses these gaps by conducting a systematic review and meta-analysis.

Solution

• Proposed approach: A systematic review and meta-analysis synthesizing data from 24 articles, 27 samples, 32 effect sizes, and 5,778 participants to quantify HV–AV cooperation and identify moderating factors.
• Novelty:
  1. First meta-analysis to quantify the magnitude and direction of HV–AV cooperation.
  2. Comprehensive examination of potential moderators, including experimental settings, demographic factors, and publication year.
  3. Inclusion of diverse driving scenarios and evaluation measures to provide a holistic understanding.
  4. Discussion of theoretical, methodological, and practical implications for AV development and traffic regulations.
• Procedure and key techniques:
  • Systematic literature search across three databases with explicit inclusion criteria.
  • Calculation of effect sizes using Hedges’ g to compare HV–AV and HV–HV cooperation.
  • Meta-regression to examine moderators such as evaluation measures, data collection methods, study designs, AV driving styles, and demographic factors.
  • Sensitivity analysis, heterogeneity assessment, and publication bias evaluation to ensure robustness.

Results

• Concrete findings:
  • HVs are significantly less cooperative with AVs than with other HVs (Hedges’ g = −0.19, 95% CI [−0.31, −0.07], p = .003).
  • Cooperation with AVs has increased in more recent studies (b = 0.07, p = .013).
  • No significant moderation effects were found for evaluation measures, data collection methods, study designs, or demographic factors.
• Advantage over baselines: Quantifies the previously inconsistent findings, providing a standardized effect size for HV–AV cooperation and identifying trends over time.
• Experiments / evaluation:
  • Included 27 samples from 24 studies, covering 5,778 participants across Europe, Asia, and the Americas.
  • Driving scenarios included lane-changing (40.5%), car-following (23.4%), turning (17.1%), head-on interactions (9.9%), and side-by-side interactions (3.6%).
  • Evaluation measures included interaction decisions (k = 19), time-based measures (k = 10), speed (k = 2), and driving volatility (k = 1).
• Limitations and future work:
  • Small number of included studies limits statistical power, especially for moderator analyses.
  • Ecological validity concerns due to reliance on surveys and driving simulators.
  • Limited exploration of high-risk scenarios (e.g., adverse weather, dawn/dusk).
  • Lack of integrative theoretical frameworks and open scientific practices.
  • Future research should focus on larger, more diverse samples, real-world data, and advanced AV trajectory planning algorithms.

Summary

This study provides the first systematic review and meta-analysis of HV–AV cooperation, revealing that human drivers are less cooperative with AVs than with other HVs (Hedges’ g = −0.19). Cooperation has increased in recent years, likely due to growing familiarity with AVs. The findings highlight the need for AV trajectory planning to account for uncooperative human behavior and suggest strategies such as signaling human presence and adopting assertive driving styles to foster cooperation. Future research should address ecological validity, expand theoretical frameworks, and improve replicability to support the safe integration of AVs into traffic systems.