AI models differ from doctors in organ allocation decisions

A study by Penn State University in the United States has revealed significant discrepancies between the decision-making processes of large language models (LLMs) and medical professionals regarding hypothetical kidney transplant allocations.

Researchers found that while AI models appeared more confident in their suggestions, they were measurably less nuanced than human doctors. Unlike medical personnel who weigh multiple complex criteria simultaneously, the AI models tended to fixate on single factors, such as alcohol consumption, with disproportionate weight. Additionally, while human doctors demonstrated hesitation and struggled with ambiguity in ethically or clinically borderline cases, the AI systems showed almost no signs of doubt or uncertainty.