Small language models improve realistic patient dialogue in addiction counseling
Multi-Objective Aligned Small Language Model Framework for SUD Patient Dialogue Generation
Computation and LanguageArtificial Intelligence
Summary
Addiction counseling needs conversations that show what patients really think and feel, which is hard for big AI models to do well. The authors created a smaller AI system that better understands patient thoughts and matches responses to specific patient histories and questions. They used smart training techniques to teach the smaller models to talk like bigger, more powerful ones, while keeping patient dialogue more realistic and helpful. Their tests showed the smaller models performed better than usual methods, especially when patients' thoughts were complex or open-ended.
What this means in practice
- •For clinical ai developers: Build dialogue systems that produce clinically realistic patient responses reflecting cognitive states during substance use disorder counseling.
- •For healthcare technology teams: Deploy smaller AI models for patient dialogue in resource-limited healthcare settings while preserving privacy and reducing operational costs.
- •For virtual therapy platform builders: Integrate cognitively aligned dialogue generation to improve realism and engagement in virtual addiction counseling services.$Commercial implications: Enables more realistic patient simulation modules for therapy platforms, enhancing user experience and treatment outcomes.
Authors
Thushara Manjari Naduvilakandy, Hyeju Jang, Mohammad Al Hasan
Abstract
Substance Use Disorder (SUD) counseling requires patient responses that reflect underlying cognitive states such as beliefs, coping strategies, and readiness for change. Although large language models (LLMs) can generate fluent text, they often fail to produce cognitively coherent and clinically realistic patient behavior, especially under ethical and data-scarce clinical settings. Moreover, deploying frontier-scale LLMs in healthcare applications presents practical challenges including high computational cost, latency, privacy concerns, and limited deployability in resource-constrained environments, motivating the need for cognitively aligned small language models (SLMs). We propose a cognitively grounded framework for SUD patient dialogue generation that explicitly models and aligns latent cognitive components with patient histories and counselor questions. Our pipeline consists of two stages: cognitive component detection and cognitive component-aligned dialogue generation. To enable effective learning with smaller models, we combine knowledge distillation from high-capacity teacher models, preference optimization from human-annotations, and attention-guided reward shaping. Extensive evaluations using automatic scores like BERTScore, ROUGE, METEOR and BLEU, and LLM-as-judge hit-metrics against both human and teacher-model references show that cognitively informed fine-tuning substantially improves cognitive realization and alignment over a generic instruction-tuned baselines and mental health domain specific SLMs, with particularly strong gains for open-ended cognitive components.