Trust and the Impact of AI on Health Care

Earning Trust

This standard and the topic of trust on AI in health care was featured during the “Trust and The Impact of AI on Health Care,” session at CES 2021. In the session, subject matter experts from Philips, the American Medical Association, and the Duke Margolis Center for Health Policy shared their perspectives on opportunities and barriers on the promise of AI for health care.

As addressed in the conference session, trustworthiness is a complicated concept since the nature of AI can make people suspicious of product performance, specifically in health care applications. Many factors go into earning and sustaining trust in an AI health care product or application and these factors vary depending on the type of use and end user/stakeholder.

ANSI/CTA-2090 considers three major expressions of how trust is created and maintained: Human Trust, Technical Trust and Regulatory Trust. While each of these concepts are individually explored in the standard, it is important to note that each of these expressions are interconnected and some topics (e.g., addressing bias) might be a factor of all three types of identified trust.

• Human Trust focuses on fostering humanistic factors that affect the creation and maintenance of trust between the developer and users. Specifically, human trust is built upon human interaction, explainability, user experience and levels of autonomy of the AI solution.
• Technical Trust focuses on the technical execution of the design and training of an AI system to deliver results as expected. Technical Trust can also be defined by considerations for data quality and integrity, including issues of bias, data security, privacy, source, and access.
• Regulatory Trust focuses on compliance and enforcement of the law and regulations by industries to gain trust. This trust can include information from regulatory agencies (e.g., FDA, FTC), federal and state laws and accreditation boards, and international standardization frameworks.

Establishing these pillars of trust represents a step in advancing the use of AI in health care, but it is important to note that the identification of the types of trust is only as valuable as it is communicated and understood by the diverse end users in the ecosystem. The conference session also emphasized the importance of collaboration across the health care ecosystem to help in the establishment of trust.

One mechanism to support collaboration is industry standards through a consensus driven process. Addressing these challenges can improve innovation and focus on the urgent need to understand and agree upon key concepts for AI in health care in a timely manner.