Tailored Conversations beyond LLMs: A RL-Based Dialog Manager

ACL ARR 2025 February Submission2524 Authors

Abstract: In this work, we propose a novel framework that integrates large language models (LLMs) with an RL-based dialogue manager for Motivational Interviews (MI). MI is a therapeutic approach that emphasizes collaboration and supports behavioral change by guiding patients to explore the reason and motivation behind their unhealthy behaviors. By leveraging hierarchical reinforcement learning to model the structured phases of MI and employ meta-learning to enhance adaptability across diverse user types, our approach enhances adaptability and efficiency, enabling the system to learn from limited data, transition fluidly between MI phases, and personalize responses to heterogeneous patient needs. Our findings demonstrate that the proposed dialogue manager outperforms an LLM baseline in terms of reward, effectively structuring MI interactions while maintaining adaptability.

Paper Type: Long

Research Area: Dialogue and Interactive Systems

Research Area Keywords: Dialogue and Interactive Systems

Contribution Types: Model analysis & interpretability

Languages Studied: English

Submission Number: 2524