A Collaborative Crowdsourcing Method for Designing External Interfaces for Autonomous Vehicles

Paper Title

A Collaborative Crowdsourcing Method for Designing External Interfaces for Autonomous Vehicles

Publication Info

• Topic area: Participatory design and crowdsourcing for autonomous vehicle interfaces.
• Keywords: Crowdsourcing, participatory design, autonomous vehicles, human-machine interfaces, eHMIs, pedestrian safety, iterative design, user-centered design, creativity, scalability.

Background and Problem

• Problem / challenge: Traditional participatory design methods are resource-intensive, limited to small groups, and lack scalability. Crowdsourcing platforms offer scalability but often restrict creativity and depth of engagement.
• Significance: Effective external Human–Machine Interfaces (eHMIs) are crucial for pedestrian safety and trust in autonomous vehicles. Current eHMI designs lack standardization and fail to address diverse user needs comprehensively.
• Motivation and related work: Prior participatory design studies have generated valuable insights but are constrained by small sample sizes and limited generalizability. Crowdsourcing has shown promise in enhancing creativity and scalability but has not been applied extensively to technical domains like eHMI design.

Solution

• Proposed approach: Collaborative crowdsourcing method combining iterative idea-building, structured visualization, and expert feedback to design eHMIs for autonomous vehicles.
• Novelty:
  1. Introduces a scalable participatory method for technology design, demonstrated through eHMIs.
  2. Provides insights into user expectations, emphasizing familiar and standardized signals.
  3. Validates crowdsourced designs against benchmarks, showing superior interpretability and user experience.
• Procedure and key techniques:
  • Iterative idea-building: Participants refine and expand on prior submissions, guided by expert feedback.
  • Structured visualization: Tree-based visualization organizes concepts across iterations.
  • Expert analysis: Independent experts evaluate concepts on effectiveness and feasibility using a 7-point Likert scale.
  • Crowdsourced collaboration: Participants engage in creative cycles, responding to low- and high-risk traffic scenarios.
  • Final evaluation: Video-based simulations compare crowdsourced designs with baseline eHMIs using interpretability (response times) and user experience metrics.

Results

• Concrete findings:
  • Popular-design outperformed baseline and innovative-design in interpretability (faster response times) and user experience (higher UEQ-S scores).
  • Innovative-design ranked second, showing potential for improvement with slight modifications.
• Advantage over baselines:
  • Popular-design reduced response times significantly compared to baseline (t = −2.77, p = .020).
  • User experience scores for popular-design were significantly higher than both baseline and innovative-design.
• Experiments / evaluation:
  • 131 participants evaluated three eHMI designs (popular, innovative, baseline) in video simulations of high-risk scenarios.
  • Metrics included response times, UEQ-S scores, and qualitative feedback on clarity and safety.
• Limitations and future work:
  • Expert feedback may introduce biases; future work should involve multiple experts for validation.
  • Crowdsourcing platform (Prolific) may limit diversity; testing across platforms is needed.
  • Evaluation focused on visual elements only; future studies should assess multimodal designs and real-world conditions.

Summary

This study presents a scalable collaborative crowdsourcing method for designing eHMIs for autonomous vehicles, combining iterative idea-building, structured visualization, and expert feedback. Participants generated creative and practical designs, emphasizing familiar and standardized signals for clarity and safety. Evaluation showed that the popular-design outperformed baseline designs in interpretability and user experience, while innovative-design highlighted potential for further improvement. The findings demonstrate the value of integrating crowdsourced creativity with expert guidance and suggest scalability as a promising avenue for participatory design in emerging technologies.