How Doctors Weigh AI Predictions Against Clinical Trial Evidence in Cancer Care

What is it about?

As AI tools enter cancer care, doctors will need to weigh personalized predictions against evidence from clinical trials. We studied how 32 physicians made treatment decisions for simulated multiple myeloma patients. To determine between treatment regimens, doctors reviewed clinical trial results and AI-generated estimates of survival and side effects (with and without information about how the AI model was developed and tested). When the two sources agreed, doctors expectedly felt more confident. When they disagreed, many changed their decision to follow the AI, even before learning whether the model was reliable or had been trained on similar patients. These findings show why clinicians must be trained to critically evaluate predictions given by medical AI tools.

Featured Image

Photo by Vitaly Gariev on Unsplash

Photo by Vitaly Gariev on Unsplash

Why is it important?

AI tools are beginning to influence high-stakes medical decisions. However, presenting more information does not guarantee that clinicians will understand or appropriately use it. Our study found that physicians often followed AI-supported treatments before reviewing evidence about the model’s quality. Even when quality information changed their confidence, many could not explain what that information meant. These findings highlight the need for safer decision-support systems that require clinicians to assess an AI model’s reliability before viewing its predictions, alongside stronger AI training and institutional safeguards.