The Medical Authority of AI: A Study of AI-enabled Consumer-Facing Health Technology

Title of the Paper

AI-Based Medical Authority: A Study on AI-Enabled Consumer Health Technologies

Bibliographic Information

• Subject Area: Human-Computer Interaction (HCI) and Health Informatics
• Keywords: Medical authority, Artificial intelligence, Symptom checkers, Consumer health technologies, Human-computer interaction

Research Background and Issues

• Research Background:

  • With technological advancements, consumers' perceptions of medical authority have shifted. In traditional medical practices, doctors and experts were regarded as the primary authorities. However, AI-based symptom checkers (AISCs) are now taking on the role of providing potential disease diagnoses.
  • In the current consumer-oriented industry, AISCs are gaining global popularity. In the United States alone, such technologies were used over 100 million times in 2019.
  • While the accuracy of AI technologies in clinical practice is often evaluated by the professional medical community, consumer-facing AI health technologies may not be effectively assessed due to users' limited medical knowledge.

• Core Issues:

  • Do AISCs alter the traditional definition of medical authority, and if so, how?
  • How do users evaluate the authority of AISCs, and how is this authority distributed within the healthcare ecosystem?

• Research Motivation and Related Work:

  • There is a risk that people may overly trust the diagnostic results of AISCs, potentially leading to the use of inaccurate diagnoses and associated risks.
  • Existing research primarily focuses on the accuracy and usability of AISCs, with limited studies exploring how these technologies influence users' perceptions and experiences of medical authority, making it challenging to design effective improvement strategies.

Solution

• Research Methods:

  • Conduct semi-structured interviews with 30 AISC users from China to study how users understand and evaluate the authority of AI health technologies.
  • Analyze users' interactions, perceptions, and decision-making behaviors in medical contexts.
  • Summarize the differences between AISCs and traditional medical authority and uncover the role of AISCs within the healthcare ecosystem.

• Implementation Steps and Techniques:

  1. Select research subjects: Analyze the functional design and background of popular AISC applications in China (e.g., "Doctor Health Assistant," "Chunyu Doctor").
  2. Data collection: Recruit participants through social media and snowball sampling, conduct interviews, and gather data on users' experiences, perceptions, and behaviors when using AISCs.
  3. Data analysis: Employ inductive thematic analysis to identify key patterns and themes in the interviews (e.g., sources of perceived authority, dynamic changes in authority during interactions).

• Innovative Points:

  • Investigate medical authority from the user's perspective rather than solely evaluating AISCs based on expert opinions.
  • Highlight the influence of interaction design and socio-technical networks in defining AI medical authority.
  • Extend understanding of AI's impact in the medical field, revealing how AISCs reshape and distribute medical authority.

Research Findings

• Specific Findings:

  • Users construct their judgments of AISC authority by verifying the supporting entities behind the AISCs (e.g., hospitals, companies) and through interaction design (input methods, question design, content presentation).
  • Users cross-verify information from AISCs with social network resources to adjust their medical decisions.
  • Medical authority is not solely dependent on the technology itself but is co-constructed and distributed across entities within the healthcare ecosystem (e.g., doctors, hospitals, search engines).

• Advantages:

  • Describes the distributed network of medical authority and reveals the complementary role of AISCs in patients' health management practices.
  • Emphasizes how the design of AI systems influences users' perceptions of their authority.

• Experimental and Evaluation Results:

  • Users are more inclined to choose AISC applications supported by authoritative institutions (e.g., apps developed by hospitals or trusted companies).
  • Users sometimes cross-verify AISC results with actual medical diagnoses, thereby enhancing trust in both AISCs and traditional medical authority.

• Limitations and Future Directions:

  • Current AISCs lack clear accountability mechanisms and the ability to track patients' subsequent issues.
  • Consumers' evaluations of symptom checkers are often influenced by their own knowledge, experiences, and physical sensations.
  • Further research is needed on the transparency of AI medical systems, algorithm accountability mechanisms, and patient-centered approaches.

Design and Practical Recommendations

• Enhance transparency by displaying information about developers, supporting institutions, and data sources to improve the authority and credibility of AISCs.
• Improve interaction design, such as making symptom input and question design more flexible and mimicking traditional authoritative diagnostic processes.
• Provide patient guidance and warnings to help users recognize their potential limitations in accurately interpreting diagnostic information.
• Position AISCs as tools for self-health management and clinical practice enhancement, strengthening the integration of user experience with medical outcomes.

Conclusion

• This study, through interviews, reveals how AI-enabled consumer health technologies influence users' evaluations and interaction experiences regarding medical authority. Future research in this field should focus on the ethical considerations, transparency, accountability mechanisms, and regulatory frameworks as AI medical technologies gradually integrate into medical practice.