Transparency is vital for AI usage in health care, patient-provider relationship, OHIO researchers find

Source Domain: www.ohio.edu

To examine their hypothesis, the three investigators enlisted respondents to participate in a scenario-based survey experiment on Mechanical Turk (MTurk), a crowdsourcing marketplace. Data from 655 MTurk respondents was used to gather a sample. Bansal, Matta and Diaz-Ordonez used attention checks and asked follow-up questions to ensure their data was reliable and valid.

As anticipated, transparency was found to be vital to the patient-provider relationship in relation to AI usage, and thus their key hypothesis was supported. Transparency in AI usage was important, and it led to higher trust in the health care provider who used it, as well as higher trust in the AI as used by their doctors. Essentially, the researchers found it to be true that if you trust the provider, then you trust the tools they use.

What was more surprising was the reaction to the role of accuracy in AI diagnoses and the impact on the patient-provider relationship. When transparency increased, trust increased—as expected, but when AI was more accurate, trust actually went down or stagnated. Bansal has a few potential explanations for why this is the case.

(Image courtesy of Adobe Stock)

“People are afraid that if AI becomes too accurate, doctors might not use their own critical judgment and it will be outsourced to AI,” he said. “Especially primary care. I think that fear is being captured here.”

Matta said these findings have the potential to change the whole game in terms of how we think about the relationship between artificial intelligence and trust.

“The reason that this is not just a big deal, but a huge deal is because it’s contrary to beliefs that accuracy improves trust,” said Matta. “The implications are massive. Without this discovery, we would all be afraid that doctors are going to be replaced by AI, but because of this study we can say ‘not so fast.’”

These findings were presented at the May 2025 Midwest Association for Information…

Source