When used by humans, large language models often lack sufficient information to make a correct diagnosis, a new study in ...