It’s 2026. Do you know where healthcare AI’s ‘social license’ is?

Contactless credit-card readers and police radar guns enjoy broad social license to operate just about anywhere. Self-driving cars and flock safety cameras don’t. AI in healthcare? 

It’s getting there. 

Defining social license as “a dynamic, informal and tacit form of public acceptance” that “extends beyond mere legal permission or regulatory clearance,” the researchers who offer the above finding expound on the concept in medical settings:  

“Together, trust and perceptions of benefits are central determinants of the social license for AI in healthcare.”

So what will it take for healthcare AI to progress from cautionary flashing yellow to solid, go-go green? 

That’s one question the investigators set out to answer. Led by academics at the University of Queensland in Australia, they had their findings published Aug. 4 in JAMA Network Open. 

3 keys to the kingdom 

For the project, senior author Clair Sullivan, MBBS (Hons), MD, and colleagues conducted workshops with 34 participants ranging in age from 31 to 60.

Among the eligibility criteria Sullivan and team applied were age restrictions (18 or older), language requirements (fluent in English) patient experiences (all had to have received care in a Queensland healthcare facility within the past 24 months). 

Analyzing the group’s inputs, the researchers identified three distinct elements. 

In helping to shape social license for AI in healthcare across the cohort, each of the three tended to present as a standalone attribute while complementing the other two. 

1. Relational engagement. 

Participants’ trust in clinicians and their ongoing relationships with clinicians influenced their trust in AI, Sullivan and co-authors report. “Most preferred that clinicians inform them when AI is used and their data are used,” they note. 

“However, recognizing clinicians’ time constraints, some participants reported they would not mind if someone else introduced this information, allowing clinicians to focus more on their care.”

2. Structural support. 

Participants raised concerns regarding both the use of health data by AI systems and the broader management of AI, the researchers found. Further, most made it clear they expected AI governance that was clear, dedicated and established. 

“Rather than positioning governance as a means to constrain AI use, participants saw governance as a necessary condition to enable and enhance its social license,” Sullivan and co-authors write. “This was present in all scenarios but more pronounced in long-term care scenarios. In short-term care scenarios, periods participants described as highly stressful and frightening when their lives felt at risk, these structural factors were not their priorities.”

3. Performance reliability.

Participants voiced a strong preference for AI-supported systems that had proven dependable at supplying timely and effective healthcare guidance.

“Participants expected AI-generated decisions to be accurate and reliable,” the researchers write. “Although many expressed high trust in the accuracy of AI, others raised concerns based on personal experiences” in which AI had flubbed. 

Social license for AI in healthcare, they conclude, is “grounded in public trust and the expectation that [its contributions] will align with and promote public benefits.” 

In their discussion section, Sullivan and team suggest their study is the first of its kind to reveal how healthcare consumers’ perceptions advance or hold back social license for AI in healthcare.

“In the settings of increasing need for refining and coordinating governance and regulatory frameworks at local, state and national levels, our findings provide actionable guidance for the sustainable integration of AI in healthcare while upholding principles of patient-centered care,” the authors write. “At the international level, these insights also align with broader efforts to strengthen AI governance globally.”

The study is posted in full for free.