Medical AI benefits vary by expertise, with nonexperts more easily misled
A one-size-fits-all approach likely isn't the best strategy when designing artificial intelligence systems that assist users in disease diagnosis.
AI assistance appears to benefit both nonexperts and medical professionals in diagnosing skin diseases, but the impact of explainable AI methods varies based on users' expertise. Researchers from MIT and other institutions conducted a study published in Nature Medicine, examining how different AI explainability techniques affected the diagnostic accuracy of nonexperts and clinicians.
Nonexperts exhibited improved diagnostic accuracy when using AI assistance, largely due to their tendency to trust the system's predictions, regardless of their correctness. They found AI explanations particularly convincing when they were vague or generic, even when the AI's assessment was incorrect. Conversely, clinicians were not influenced by erroneous AI help and performed best when presented only with the model’s prediction, without additional interpretation.
While AI systems can enhance performance in specific health settings, the reliance on AI models can introduce errors, particularly for those with limited medical knowledge. The study's authors emphasize the necessity of designing AI systems that align with users' abilities and foster critical thinking rather than fostering overreliance.
This finding is especially relevant as patients increasingly seek AI support for their healthcare needs. The researchers recommend tailoring AI explanations to encourage informed decision-making and mitigate the risk of misleading less knowledgeable users.
Written by urgent.news from Medical Xpress's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
