Work with AI and Work for AI: Autonomous Vehicle Safety Drivers’ Lived Experiences

Title of the Paper

Work with AI and Work for AI: Autonomous Vehicle Safety Drivers’ Lived Experiences

Paper Information

• Subject Area: Human-Computer Interaction (HCI), Autonomous Driving Safety, AI Labor
• Keywords: Autonomous Driving, Artificial Intelligence, Human-Computer Interaction, Safety Drivers, AI Labor, Responsibility Ethics, AI Perception, AI Education

Research Background and Issues

• Identified Problems or Challenges:

  • The development of autonomous driving technology has given rise to a new profession—safety drivers. These drivers are tasked with supervising and operating autonomous vehicles in high-risk real-world traffic environments, yet their work practices and experiences remain underexplored within the HCI community.
  • Most HCI studies on autonomous driving are based on laboratory experiments or short-term observations, disconnected from real-world, long-term practices.
  • Autonomous driving technology cannot perfectly handle all situations on public roads, transferring the risks of driving tasks to the professional group of safety drivers.

• Significance:

  • Understanding the practices, experiences, and challenges of safety drivers can inspire future HCI research, particularly regarding long-term human-AI interaction experiences.
  • This study sheds light on how grassroots labor interacts with technology in an evolving AI-driven society.

• Research Motivation and Related Work:

  • Current research primarily focuses on trust calibration, situational awareness, and handover control in autonomous driving systems, with little attention paid to the long-term interactions between safety drivers and autonomous systems.
  • The study integrates discussions on AI user experience, the impact of AI on labor, and related topics.

Solution

• Methods and Solutions:

  • The authors conducted semi-structured interviews with 26 Chinese safety drivers to investigate their real-world experiences and work practices.
  • Through an in-depth analysis of the human-computer interaction and professional challenges faced by safety drivers, the study provides the first empirical research on long-term, real-world scenarios.

• Innovations:

  • The study is the first to comprehensively document the lived experiences of the first passengers in autonomous driving development—safety drivers.
  • It explores how non-technical users learn AI knowledge and the mechanisms of trust and interaction between autonomous vehicles and humans.
  • It identifies key professional challenges related to responsibility allocation and ethical dilemmas, reflecting potential societal issues posed by AI in real-world settings.

• Implementation Steps and Key Techniques:

  • Semi-structured interviews: Conducted in phases to collect data on safety drivers’ work practices, perception development, experiences of collaborating with AI, and well-being challenges.
  • Data coding and thematic analysis: Classified and analyzed interview data based on key themes (e.g., work practices, perceptions of AVs, work well-being).
  • Comparison with existing research: Results were compared with laboratory studies and short-term observations to identify potential gaps.

Research Outcomes

• Specific Findings:

  • Analyzed how safety drivers adapt their perceptions of autonomous driving technology and make takeover decisions in specific scenarios.
  • Clarified cognitive preferences, learning characteristics, and collaboration patterns exhibited by safety drivers during long-term interactions with autonomous vehicles.
  • Highlighted that safety drivers bear accumulated risks from the autonomous driving development chain while facing limited career growth and marginalization.

• Advantages:

  • Provides first-hand, real-world feedback for the autonomous driving field, revealing gaps between laboratory and real-world environments.
  • Offers insights into improving human-vehicle collaboration in autonomous driving and enhancing worker well-being.

• Experimental or Evaluation Results:

  • Interview analysis revealed that safety drivers develop more accurate mental models through practical experience rather than theoretical training alone.
  • In real-world driving, safety drivers tend to trust their intuition over decisions made by the autonomous driving system in emergency situations.
  • Safety drivers face restricted knowledge and skill development, limiting their opportunities for career advancement or further education.

• Limitations and Future Directions:

  • Limitations: The sample size is limited, and participants may avoid critical comments due to the sensitivity of their work, leading to subjective and potentially unrepresentative results.
  • Future Directions: Expand the study sample to include other stakeholders in autonomous driving technology; conduct quantitative research for validation; explore more effective training and knowledge transfer strategies for drivers.

References

The paper includes a detailed list of references covering key academic works in autonomous driving technology, AI labor, user experience research, and related fields.