Large language models (LLMs) were more likely to suggest inferior treatment options when reviewing hypothetical cases when ...