Enhancing AI-Driven Psychological Consultation: Layered Prompts with Large Language Models

TL;DR

The paper addresses the accessibility and scalability challenges of psychological consultation by leveraging large language models (LLMs) with a layered prompting framework. It combines initial information gathering prompts, context-sensitive follow-ups, empathy-driven prompts, and scenario-based prompts to enhance information collection, empathy, and contextual understanding in AI-driven counseling. A diverse, anonymized dataset of psychological dialogues and a GPT-4-based evaluation framework evaluate performance across relevance, empathy, context, and user satisfaction, demonstrating significant improvements over baselines and Chain-of-Thought prompting. The work provides a scalable, data-driven approach to augment mental health support with AI, offering practical implications for online counseling services and domain-specific prompt engineering.

Abstract

Psychological consultation is essential for improving mental health and well-being, yet challenges such as the shortage of qualified professionals and scalability issues limit its accessibility. To address these challenges, we explore the use of large language models (LLMs) like GPT-4 to augment psychological consultation services. Our approach introduces a novel layered prompting system that dynamically adapts to user input, enabling comprehensive and relevant information gathering. We also develop empathy-driven and scenario-based prompts to enhance the LLM's emotional intelligence and contextual understanding in therapeutic settings. We validated our approach through experiments using a newly collected dataset of psychological consultation dialogues, demonstrating significant improvements in response quality. The results highlight the potential of our prompt engineering techniques to enhance AI-driven psychological consultation, offering a scalable and accessible solution to meet the growing demand for mental health support.