Sentiment Differences in Resolving Moral Dilemmas: An Analysis of Student Attitudes and Large Language Models

Abstract

This study examines how large language models (LLMs) and humans differ in their emotional responses when resolving moral dilemmas in medical education. Medical students and LLMs were presented with the classic trolley and bridge dilemmas, and their responses were compared with expert-written academic texts using sentiment analysis to assess emotional tone and patterns. Students’ answers exhibited the highest emotional intensity, marked by fear and sadness, while LLMs responded with low emotional involvement, characterized by anticipation and trust. Expert texts displayed a more neutral and balanced emotional tone, intermediate between student and LLM responses. These findings suggest that human experts adopt a reflective, emotionally regulated approach to moral reasoning, whereas medical students respond with greater emotional intensity, reflecting their developmental stage. LLMs, by contrast, analyze dilemmas through factual reasoning largely devoid of affective depth. Integrating AI tools with emotional-awareness training may therefore enhance medical education in addressing ethical complexity. Introduction.