ExplAIn Yourself! Transparency for Positive UX in Autonomous Driving

Document Title

ExplAIn Yourself! Transparency for Positive UX in Autonomous Driving

Document Information

• Subject Area: Autonomous driving user experience design, explainable artificial intelligence
• Keywords: User experience, explainable artificial intelligence, autonomous driving, controllability, user acceptance, transparency, human-computer interaction, driving simulation, user study, design guidelines

Research Background and Problem

• Challenges identified by the authors:
  • In autonomous vehicles, the AI system replaces human driving behavior, which may lead to user distrust of the system and result in negative user experience (UX).
  • Passive driving experiences (e.g., passengers feeling confused or dissatisfied with the vehicle's decisions or driving style) can reduce user acceptance and sense of control.
• Importance:
  • In the early stages of autonomous driving technology adoption, user acceptance is a critical factor in determining the success of the technology.
  • Enhancing users' perceived safety and transparency can effectively reduce anxiety and promote the application of autonomous driving technology.
• Research Motivation:
  • To explore how explainable artificial intelligence (XAI) can be integrated with user experience design (UXD) to improve transparency in autonomous driving systems and enhance user trust and acceptance.

Solution

• Proposed Approach:
  • The authors designed an experimental study using a driving simulator and a mobile application to investigate the impact of system transparency on user experience.
  • Comparison of two information delivery modes: live explanations during the driving process and retrospective explanations provided via a mobile app after the drive.
• Innovative Aspects:
  • Combining XAI methods with UXD research, focusing on the impact of information transparency on the user experience of first-time autonomous vehicle users.
  • Offering a preliminary set of design guidelines for autonomous driving user experience.
• Implementation Steps:
  1. Simulate fully autonomous driving scenarios (SAE Level 5) using a fixed driving simulator and a mobile application.
  2. Divide participants into two groups: one receiving live explanation information during the ride, and the other accessing explanations retrospectively via the app after the ride.
  3. Evaluate participants' user experience and perceived safety under different test conditions.
• Key Technologies:
  • AR (augmented reality)-based visualization of information in dynamic driving scenarios.
  • Standard questionnaires (AVAM and UEQ-S) to measure user acceptance, sense of control, and user experience.

Research Outcomes

• Main Findings:
  • Providing explanatory information (whether live or retrospective) helps transform negative user experiences into neutral ones.
  • Live explanations are more effective in improving users' understanding and acceptance of autonomous driving systems.
  • A higher sense of control is significantly correlated with more positive user experiences.
• Experimental Results:
  • The group receiving live explanations demonstrated significantly better user experience and understanding of system behavior compared to the group without live explanations.
  • In the absence of live explanations, retrospective feedback via the mobile app significantly improved users' perceived sense of control and understanding.
  • However, the combined effect of live explanations and retrospective feedback was not significant.
• Limitations:
  • The use of a static driving simulator may have affected participants' realistic perception, particularly in safety evaluations.
  • Participants were predominantly young (average age 24.65), which may have skewed results toward higher technology acceptance.
  • The study focused solely on short-term experiences of first-time users, leaving long-term usage effects unexplored.
• Future Directions:
  • Investigate personalized information presentation to accommodate different user preferences and changes in transparency during long-term interactions.
  • Conduct real-time experiments in dynamic driving scenarios to further validate the impact of information transparency on user trust and experience.
  • Integrate multimodal interactions (e.g., audio, seat vibrations) to enhance the effectiveness of sensory information delivery.

Summary and Contributions

• Preliminary Design Guidelines:
  • Live explanations effectively alleviate users' negative emotions toward autonomous driving and improve their experience.
  • Retrospective feedback can also enhance user experience in the absence of live explanations, but live explanations are more effective in providing overall psychological safety.
  • Enhancing the sense of control is a key principle for designing user-friendly autonomous driving systems.
• Contributions to the Research Community:
  • The study is the first to integrate UX and XAI fields, offering practical insights and guidelines for designing more positive user experiences.
  • Provides a reference direction for future research on autonomous driving and human-computer interaction.