Artificial Intelligence and poor prognosis

An analysis of realism and pedagogical implications

Abstract

Is AI able to imitate human communication in the medical field, with particular attention to communicating poor prognosis? We generated responses to various prompts to evaluate their accuracy and human perception. They were transformed into cards, then evaluated by doctors with experience in different fields. The results, unexpectedly, demonstrate that AI can produce surprisingly realistic responses, indistinguishable from human ones even for experienced professionals. Furthermore, if on the one hand, AI manages to embody a good and compassionate doctor, on the other it can also embody a sadistic and ruthless one.