The Effect of In-Car Agent Embodiment on Different Types of Information Delivery

Research Background and Issues

• Identified Problems and Challenges:
  Against the backdrop of rapid advancements in automotive technology, in-car intelligent assistants need to manage and convey various types of information that may impact the driver's experience and safety. However, the optimal design for different types of information (e.g., warning information, recommendation information, and reference information) remains unclear, particularly how the assistant's different "embodiments" (e.g., physical characteristics, role characteristics, and dynamic behaviors) can optimize information transmission efficiency.

• Why It Matters:
  The effectiveness of information transmission is critical to driving safety and user experience. Properly designed in-car assistants can not only enhance users' acceptance of information but also increase their trust and engagement, especially in scenarios requiring quick reactions during driving.

• Research Motivation and Related Work:
  Although many prior studies have explored the information display methods of embedded assistants (e.g., voice assistants, augmented reality), there is limited research on how the "embodied" characteristics of in-car assistants (e.g., physical embedding, role characteristics, and dynamic behaviors) affect the transmission of different types of information and user experience. Researchers have also noted that embodiment factors in different scenarios may significantly influence users' attention, trust, and enjoyment.

Solution

• Proposed Methods and Solutions:
  This paper introduces a prototype system called "Drop-lit," which allows the independent or combined use of different embodied characteristics of in-car assistants (e.g., physicality, role characteristics, dynamic actions) to optimize the transmission of various types of information. The study employs three representative designs—Abstracted Agent, Digital Agent, and Mixed-Media Agent—to compare the effectiveness of transmitting six types of information.

• Innovations:

  1. Proposed a "mixed-media" embodiment design (integrating digital and physical environments) for complex information scenarios.
  2. Classified and precisely defined three types of in-car information (warning information, recommendation information, and reference information) and explored the relationship between information types and user preferences.
  3. Introduced a systematic analysis of the impact of embodiment characteristics on user experience factors (e.g., attention, urgency, friendliness, trust, and enjoyment).

• Implementation Steps and Key Technologies:

  1. Information Classification: Through focus group interviews, in-car information was categorized into warning information (urgent and non-urgent), recommendation information (personalized and context-related), and reference information.
  2. Drop-lit System Design: Integrated OLED screens, servo motors, and hardware control based on Node.js and Arduino to achieve embodied dynamic and physical interactions.
  3. Experimental Design: Conducted two experimental phases, first allowing participants to evaluate the three representative agents, then enabling them to freely design and optimize embodiment characteristics.
  4. Data Analysis: Combined questionnaire data, user-customized designs, and interview records to quantify and summarize the relationship between embodiment characteristics, information transmission effectiveness, and user experience.

Research Findings

• Specific Findings:

  1. Effectiveness of Different Agents:
     • Mixed-Media Agent performed best in transmitting "urgent warning information" due to its stronger ability to attract attention and convey urgency.
     • Digital Agent was more suitable for transmitting recommendation information, as it enhanced friendliness and user engagement.
     • Abstracted Agent was most effective in transmitting simple reference information, as it required less user attention.
  2. Key User Experience Factors for Preferences:
     • Warning information needs to emphasize attention attraction and urgency.
     • Recommendation information tends to emphasize friendliness and enjoyment.
     • Reference information tends to minimize interference and be presented in a more concise manner.

• Advantages Over Existing Solutions:

  • Provides a customizable exploration platform for embodiment characteristics, enabling in-car assistants to freely adjust their appearance and behavior based on information type and user needs.
  • The system comprehensively addresses user preferences and experience expectations for different information types, offering clear design directions for future in-car assistant development.

• Experimental and Evaluation Results:

  • User research data showed high consistency among participants regarding the association between different embodiment characteristics and information types. For example, agents with dynamic levels and rapid movements were more suitable for transmitting urgent information, while static abstract graphics were better for simple information.

• Limitations and Future Directions:

  1. Limitations:
     • The experimental environment was a simulated driving scenario rather than a real-world road setting, which may affect the applicability of the results.
     • The study did not deeply explore the impact of sound or other auditory factors on embodiment characteristics and user experience.
     • Lacked validation through long-term studies of user interaction behavior.
  2. Future Directions:
     • Transition experiments to real-world driving environments to validate the authenticity of laboratory findings.
     • Consider auditory design factors such as tone and voice personalization to understand their impact on user preferences.
     • Introduce physiological data collection (e.g., reaction time and focus) to objectively evaluate the effectiveness of in-car assistants.

Conclusion

By constructing the Drop-lit platform and conducting systematic experiments, this study validates the optimized combinations of in-car assistant embodiment characteristics and different information types. The research provides clear recommendations for future in-car assistant designs, such as prioritizing mixed-media agents for transmitting urgent information and focusing on user-expected experience factors (e.g., urgency and trust). Despite existing limitations, this study offers a significant theoretical foundation and detailed design guidance for optimizing in-car intelligent assistant design and user experience.