Bridging the Trust Gap: Investigating the Role of Trust Transfer in the Adoption of AI Instructors for Digital Privacy Education

Research Background and Issues

• What problems or challenges did the authors identify?
  In recent years, the potential of AI instructors in the field of digital privacy education has been increasingly recognized. However, trust issues have become a major barrier to their widespread adoption, particularly as older users tend to be skeptical of AI instructors. The lack of trust leads users to reject AI recommendations, thereby limiting the advantages of personalized guidance and real-time interaction that AI can provide.

• Why is this issue important?
  Digital privacy education is crucial for improving the digital privacy literacy of the public, especially older adults. AI instructors have the potential to provide continuous and personalized guidance, which is essential for addressing widespread gaps in digital privacy knowledge. If trust issues can be resolved, the effectiveness of user learning can be enhanced, and the application scenarios of AI instructors can be expanded.

• Research Motivation and Related Work
  This study is based on trust transfer theory, exploring whether introducing AI instructors through human intermediaries can improve user trust in AI instructors and enhance the learning experience. While trust transfer theory has been applied in the business domain, its application in the educational field, particularly with AI instructors, remains a novel area of exploration.

Solution

• What methods or solutions did the authors propose?
  The authors designed an experiment to investigate whether trust transfer can help enhance user trust in AI instructors. One group was introduced to the AI instructor by a human, another group experienced a self-introduction by the AI instructor, and a third group served as a control group with teaching conducted entirely by humans.

• What is innovative about this solution?
  This approach extends trust transfer theory, traditionally applied in business contexts, to the educational domain, specifically in AI-driven privacy education. The study aims to verify whether human mediation can effectively alleviate users' "algorithm aversion" toward AI, paving the way for broader applications of AI instructors.

• What are the implementation steps and key technologies used?

  1. Develop a digital privacy education module (including privacy risks and protection strategies) and three video conditions (human introducing AI, AI self-introduction, human teaching).
  2. Recruit participants of varying ages and randomly assign them to different groups.
  3. Collect data through questionnaires on learning experiences (trust, learning perception, enjoyment) and participant characteristics (motivation, privacy concerns, etc.).
  4. Use Structural Equation Modeling (SEM) to analyze trust transfer and its impact on the learning experience.

Research Outcomes

• What specific outcomes were achieved?

  1. Trust transfer was proven effective: AI instructors introduced by humans were significantly more trusted than those introducing themselves, with trust levels comparable to the fully human-taught condition.
  2. Trust directly influenced participants' learning perception and enjoyment, and indirectly enhanced the learning experience through the "attractiveness traits" of the instructor.
  3. The effects of trust transfer were consistent across participants of different ages, genders, and regions, demonstrating strong adaptability.

• What advantages does it have compared to existing solutions?
  Compared to traditional trust-building methods that rely solely on AI performance or transparency, trust transfer offers a faster and more effective way to establish trust. It is also better suited for first-time or short-term interaction scenarios.

• What were the experimental or evaluation results?

  1. The group with human-introduced AI (HA) demonstrated significantly higher trust and perceived instructor professionalism compared to the AI self-introduction group (AA).
  2. The HA group also showed advantages in enjoyment and learning perception, with performance similar to the human teaching group (HH).
  3. Participants from urban backgrounds exhibited higher levels of technological trust and learning motivation, further supporting the feasibility of trust transfer across diverse groups.

• Limitations and Future Directions

  1. The experiment primarily involved participants from the United States, which may limit the cultural generalizability of the findings.
  2. The teaching module lacked real-time interaction; future research should explore how trust evolves with dynamic AI instructors in long-term interactions.
  3. The study did not directly measure actual knowledge improvement; future research could incorporate objective assessments of learning outcomes.
  4. Further investigation is needed into the potential impact of different human introducer characteristics (e.g., age, gender) on trust transfer effectiveness.

Through the above analysis, this study provides a novel pathway to address the issue of trust deficiency in AI instructors. It confirms the effectiveness and broad applicability of trust transfer, offering significant value for both practical implementation and theoretical research.