How aversion to AI-using doctors crosses demographic groups

No evidence that many key demographics affect the AI-use penalty for doctors 

Yet the research goes further: it also looks at the effect of the demographics of these doctors and consumers (specifically race, gender and ethnicity) on such an AI-use penalty, and the research finds surprisingly that the AI-use penalty is largely independent of such demographics.

“We wanted to know whether the AI-use penalty depends on who the doctor is and who is evaluating them,” says co-author Yeun Joon Kim. “We examined various demographic characteristics of both physicians and healthcare consumers. Interestingly, we found little evidence that these characteristics made a substantial difference in how consumers evaluated doctors who disclosed AI use.” 

“Previous research shows that people’s evaluations of others are often influenced by demographic characteristics such as age, gender and race. We wanted to see whether these characteristics would also shape reactions to doctors using AI. Surprisingly, the AI-use penalty was largely unaffected by physicians’ demographic characteristics. This suggests that AI aversion may be somewhat distinctive from other forms of social bias.”

“What particularly surprised me was that we found no clear evidence that female or Black doctors faced an additional penalty for using AI. Previous studies have shown that women and racial minorities may face disadvantages in how others evaluate them. So, we wondered whether these physicians might face a double penalty when disclosing AI use. Interestingly, our findings did not support this expectation.” 

Warmth and competence are particularly important in medical settings

The research is based on a sample of 1,030 US-based participants whose age, gender and race approximate the composition of the US adult population. They were shown hypothetical advertisements for family doctors as a supposed part of a market research study. 

“Within these evaluations, we focused on perceived competence and warmth, 2 fundamental dimensions of social perception that reflect judgments about a person’s ability and intentions,” says the study published in the journal npj Digital Medicine, part of the Nature publishing group. “These dimensions are particularly relevant in medical settings, where consumers must assess both whether a physician is capable and whether they act in the consumers’ best interest.” 

I want a second opinion

The research then turned to behavioural intentions and found that consumers of AI-using doctors had reduced willingness to book an appointment, to recommend the doctor to others and to book with this doctor again – the consequences of the perceptions about competence and warmth. Customers of the AI-using doctor were also significantly more likely to request a second opinion from another doctor, and they also indicated that they would travel less far, pay less money and wait less long to see the AI-using doctor. 

“Overall, physicians who stated that they use (versus never use) AI were perceived as significantly less warm and competent,” says the research. “This AI-use penalty manifested also in various behavioural intentions, like a reduced appointment willingness. Thereby, we extend prior research in this context, which focused on a limited set of evaluation criteria (such as trustworthiness). Moreover, our findings shed light on the moderating influence of physician and consumer demographics. Importantly, the observed penalty was largely insensitive to physician demographics (age, gender and race), suggesting that AI-use functions as a salient cue.” 

AI use signals to consumers that a doctor is outsourcing clinical judgement

The authors say that the concept of a cue-based perspective (which focuses on specific signals) helps explain the AI-use penalty findings. 

“In the absence of direct information about a physician’s ability or intentions, consumers rely on observable signals. AI-use disclosure appears to function as such a cue, potentially being interpreted as a sign of ‘outsourcing’ key aspects of clinical judgment. Specifically, AI use may signal a lack of benign intent (warmth) or expertise (competence).” 

Younger healthcare consumers averse to AI use by doctors may feel threatened by AI in their own careers 

While most demographics of consumers had little effect on the results, consumers under age 45 were more likely to rate AI-disclosing doctors harsher on warmth and competence – and such younger consumers were also less willing to pay, travel and wait for an appointment with such AI-using physicians. 

The authors suggest that such younger consumers may be more sceptical about AI use by doctors because they are more familiar with AI and thus more aware of the technology’s limitations in such a sensitive context as healthcare. 

Yeun Joon Kim says, “while we did not directly examine why younger consumers were more averse to AI-using doctors, one possible explanation is their greater familiarity with AI. Younger people may be more aware of both its impressive capabilities and its limitations. They may recognise that AI can make mistakes, particularly in complex medical situations. At the same time, knowing how powerful AI has become may make them more concerned that doctors could rely too heavily on it, rather than exercising their own clinical judgment. In healthcare, where patients expect doctors to take personal responsibility for their decisions, these concerns may help explain why greater familiarity with AI does not necessarily translate into greater acceptance of its use.”

The importance of how AI use is communicated

The implications of the broader findings, says the research, is that “adopting AI is not merely a technical or organisational challenge but also a social one which should be considered alongside performance and efficiency gains when evaluating AI adoption. Second, these results underscore the importance of how AI use is communicated and enacted in consumer interactions. Different ways of introducing and explaining the use of AI – emphasising physician responsibility, shared decision making and consumer benefit – might mitigate negative perceptions.” 

The authors also suggest that training in how to discuss AI with patients may be useful, potentially with greater attention to the concerns of younger consumers. This should not be taken as a call for patients to trust AI uncritically. As the paper notes, a degree of scepticism is appropriate given the limitations of current AI systems. 

“A key takeaway from our research is that aversion to AI-using doctors is not simply driven by the superficial demographic characteristics of physicians, such as their age, gender or race,” says Joon. “Previous research has shown that these characteristics often influence how people evaluate others. Yet we found that the AI-use penalty was largely unaffected by them. This suggests that AI aversion may be somewhat different from conventional forms of social bias, potentially reflecting more fundamental concerns about the role of AI in healthcare.” 

He continues: “As AI becomes increasingly integrated into healthcare, policymakers and healthcare organisations need to understand why patients are sceptical of doctors who use it. Some concerns may be justified, given the limitations of current AI systems, while others may arise from misunderstandings about how doctors use AI. Understanding these different sources of scepticism will be important for ensuring that AI can be adopted responsibly without undermining patients’ trust in their doctors.”