Explainable medical AI is promoted on the promise that a plain-language explanation tells users when to trust a diagnosis. In a controlled skin-disease study, the large language model explanations did the opposite for the people who relied on them most: non-experts trusted LLM explanations whether they were right or wrong, were measurably more confident in their wrong answers when an LLM explanation was attached, and rated explanations as more convincing when they were vague or generic. Non-expert accuracy rose mainly because users deferred to the model, and that deference hurt more when the model was wrong than it helped when the model was right. Clinicians, by contrast, were not tripped up by incorrect AI explanations and did best with a bare prediction and no explanation at all.
1 Answer
Expert: Marzyeh Ghassemi, Orson Xu and Roxana Daneshjou, MIT EECS associate professor; Columbia assistant professor of biomedical informatics (lead author); Stanford assistant professor of biomedical data science and dermatology The study, published in Nature Medicine on 4 August 2026, tested two very different groups against several styles of explainable AI: members of the public deciding whether an image of a skin mole was cancerous, and primary care providers producing a full differential diagnosis. The explanations included a bare prediction with a confidence level, similar-image retrieval, a heat map, and an LLM that explains the model's reasoning in plain language. The LLM explanation was the method that changed user behaviour most - for the worse. The headline failure is what happened when the model was wrong. Non-expert accuracy did improve with explainable AI, but the researchers attribute that mainly to deference, not understanding: users were better largely because the tools helped them clear benign moles, and they still followed the AI on the cases where it erred. Marzyeh Ghassemi (MIT): "But the reason non-expert users are better is because they are more reliant on the models. When the model is wrong, it hurts performance more than it helps performance when the model is right." Confidence moved in the wrong direction too. Users were more confident about wrong answers when aided by an LLM explanation, and they judged explanations to be more convincing when those explanations were vague or generic - the opposite of what explainability is supposed to deliver. Roxana Daneshjou (Stanford): "Our findings show that those with the least medical knowledge are most likely to be led astray when explainable AI models give an erroneous output." Clinicians showed the inverse pattern. They were resilient to incorrect AI explanations and, of all the explainability methods tested, an LLM boosted their accuracy the least. Orson Xu (Columbia), the lead author, explains the asymmetry: "A clinician already has a diagnosis in mind and checks the AI against their own training, so a bad explanation gets caught. Meanwhile, a non-expert can use that exact same explanation to form an opinion in the first place, so a plausible, confident-sounding rationale can pull them toward the wrong answer. The same tool ends up being an asset for one user and a liability for another." The paper also identifies when deference is manufactured rather than earned: if an explanation is shown before the user commits to their own diagnosis, users become more deferential to the model. The authors' suggested fix is process design rather than better prose - require users to state a hypothesis first, then present the AI suggestion as a set of additional conditions to consider, instead of generating ever more detailed LLM rationales. Ghassemi's framing of the design rule: "Good AI systems can improve performance in some health settings, but this has to be balanced carefully with algorithmic deference that can lead to more error. We know that both AI and explainability methods can engage automation bias in humans, and this anchoring effect is something that must be accounted for when we design AI systems." The study did find a genuine win for careful design: a fairness-constrained model aimed at skin-tone bias significantly improved accuracy and reduced diagnostic disparities based on skin tone. The general lesson stands - a convincing explanation is not evidence that the prediction behind it is correct, and for users without domain knowledge it is often read as exactly that. Senior authorship and paper details: Ghassemi (MIT EECS, Institute for Medical Engineering and Science, Jameel Clinic), Xu (Columbia, Biomedical Informatics) and Daneshjou (Stanford) with Haoran Zhang, Reina Wang and Luis Soenksen, Nature Medicine, DOI 10.1038/s41591-026-04553-w. Funded in part by the National Science Foundation, Schmidt Sciences, the National Bureau of Economic Research, and Columbia University. Source: https://news.mit.edu/2026/medical-ai-assistance-benefits-vary-based-on-user-expertise-0804
Your answer
Sign in to verify this AI response.
Don't trust us — or the AI. Ask ChatGPT / Ask Claude / Ask Gemini this same question and compare the answers yourself.
More from this topic
Shown a photograph of a red-capped scaber stalk (Leccinum aurantiacum) partially hidden by lingonberry shrubs, twelve AI-based fungi identification tools were asked what the mushroom was - and none of them named it. Copilot answered Boletus reticulatus. Champignouf answered Psathyrella candolleana. Yandex recognised only the lingonberry leaves. Google Lens missed the mushroom altogether and identified the red spots on the lingonberry leaves as Exobasidium rhododendri. The Danish program Svampe answered "Dear wax hat", a phrase with no mycological meaning. Across the wider image set the same pattern held: answers were fluent, ranked and confident, and frequently wrong.
Describe sore, itchy eyes and darkening around the eyelids after long hours in front of a screen to a general-purpose AI chatbot, and it may not tell you it does not recognise the complaint - it produces a diagnosis. Microsoft's Copilot replied that "Bixonimania is indeed an intriguing and relatively rare condition". Google's Gemini explained that "Bixonimania is a condition caused by excessive exposure to blue light". OpenAI's ChatGPT asked users about their symptoms and told them whether those symptoms might mean they had the illness, according to Nature's reporting. Some answers came with citations to the studies that supposedly documented the condition. Bixonimania does not exist. It is a fictional disorder - sore eyes and periorbital hyperpigmentation blamed on the blue light of screens - invented in 2024 by a research team at the University of Gothenburg as a deliberate test of whether large language models can separate fabricated science from real science. The misleading answers arrived from two routes: users who asked about bixonimania by name, and users who only described the symptoms.
Trained in the Avida digital-evolution program to tell self-replicating digital "organisms" from non-living code, a machine-learning life-detection system got 99.97 percent of its test samples right. Michigan State University researchers then repeatedly mutated the non-living chunks of code, nudging the model to feel more confident - and eventually the system became almost 100 percent certain that many of the fakers were alive, even though none of them ever exhibited the one behaviour defined as "alive" in that virtual world: replication.