CALM-IT: Generating Realistic Long-Form Motivational Interviewing Dialogues with Dual-Actor Conversational Dynamics Tracking

2026-01-15

AI Summary

This work addresses the challenge of strategic drift in large language models during mental health conversations, which often undermines the effectiveness of long-term interventions. The authors propose a dual-role state space modeling approach that, for the first time, formalizes the therapist–client interaction as a bidirectionally evolving dynamic process. This framework continuously tracks both parties’ psychological states, goal alignment, and short-term intentions in real time, and integrates a goal-driven policy selection mechanism to generate motivational interviewing dialogues. Experimental results demonstrate that the proposed method significantly improves the precision of intervention timing and dialogue stability, maintaining high goal consistency over extended interactions. Notably, it achieves a client acceptance rate of 64.3%, outperforming existing baselines.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Large Language Models (LLMs) are increasingly used in mental health-related settings, yet they struggle to sustain realistic, goal-directed dialogue over extended interactions. While LLMs generate fluent responses, they optimize locally for the next turn rather than maintaining a coherent model of therapeutic progress, leading to brittleness and long-horizon drift. We introduce CALM-IT, a framework for generating and evaluating long-form Motivational Interviewing (MI) dialogues that explicitly models dual-actor conversational dynamics. CALM-IT represents therapist-client interaction as a bidirectional state-space process, in which both agents continuously update inferred alignment, mental states, and short-term goals to guide strategy selection and utterance generation. Across large-scale evaluations, CALM-IT consistently outperforms strong baselines in Effectiveness and Goal Alignment and remains substantially more stable as conversation length increases. Although CALM-IT initiates fewer therapist redirections, it achieves the highest client acceptance rate (64.3%), indicating more precise and therapeutically aligned intervention timing. Overall, CALM-IT provides evidence for modeling evolving conversational state being essential for generating high-quality long-form synthetic conversations.

Problem

• long-form dialogue
• motivational interviewing
• conversational dynamics
• therapeutic alignment
• large language models

Innovation

• dual-actor conversational dynamics
• state-space modeling
• Motivational Interviewing
• long-form dialogue generation
• therapeutic alignment