Empathetic Cascading Networks: A Multi-Stage Prompting Technique for Reducing Social Biases in Large Language Models

TL;DR

This paper introduces Empathetic Cascading Networks, a multi-stage prompting framework that significantly improves empathy and inclusivity in large language models, with high EQ scores and maintained response quality.

Contribution

The paper proposes a novel multi-stage prompting technique, ECN, to enhance empathetic and inclusive responses in large language models, advancing beyond existing methods.

Findings

ECN achieves the highest Empathy Quotient scores on GPT-3.5-turbo and GPT-4.

ECN maintains competitive Regard and Perplexity metrics.

ECN demonstrates potential for socially responsible conversational AI.

Abstract

This report presents the Empathetic Cascading Networks (ECN) framework, a multi-stage prompting method designed to enhance the empathetic and inclusive capabilities of large language models. ECN employs four stages: Perspective Adoption, Emotional Resonance, Reflective Understanding, and Integrative Synthesis, to guide models toward generating emotionally resonant and contextually aware responses. Experimental results demonstrate that ECN achieves the highest Empathy Quotient (EQ) scores across GPT-3.5-turbo and GPT-4, while maintaining competitive Regard and Perplexity metrics. These findings emphasize ECN's potential for applications requiring empathy and inclusivity in conversational AI.