Let's Put Speaker into Interlocutor's Shoes! Exploring the Impact of Zero-Shot Chain-of-Thought on Empathy

Anonymous

Abstract: This study investigates the effectiveness of the Zero-shot Chain-of-Thought (CoT) approach, specifically the ``\textit{Let's think step by step.}'', in enhancing both the perceived empathy and empathetic reasoning abilities of 14 Large Language Models (LLMs) in total. However, our experiments indicate that Zero-shot CoT does not significantly improve LLMs' empathetic reasoning compared to Zero-shot In-Context Learning (ICL), based on a range of performance metrics. Notably, we uncover that employing the perspective-taking prompt (\ie ``\textit{Let's put {speaker} into {interlocutor}'s shoes.}'') strengthens empathetic reasoning, leading to notable improvements in perceived empathy scores. The source code will be made available upon publication.

Paper Type: long

Research Area: Dialogue and Interactive Systems

Contribution Types: Model analysis & interpretability

Languages Studied: English

Consent To Share Submission Details: On behalf of all authors, we agree to the terms above to share our submission details.