Elamur Research⌕
☰
home / papers / 2026 / MIThinker: A Plug-and-Play Policy-Optimized Thinker For Motivational Interviewing Counseling

MIThinker: A Plug-and-Play Policy-Optimized Thinker For Motivational Interviewing Counseling

Paper · 2026 · Yizhe Yang, Palakorn Achananuparp, Heyan Huang et al.

Summary

LLM-based Motivational Interviewing (MI) counselors usually respond without explicit reasoning tied to counseling technique. Prompt-based systems that do reason, such as CAMI, rely on several costly modules. The authors propose MIThinker, a small plug-and-play model. It produces a structured "therapeutic thought" before each turn: an assessment of the client's mental state and a justified choice of MI strategy. Any backbone LLM can then use this thought to write the reply.

Real sessions never record what counselors were thinking. To fill this gap, the AugR1-MI pipeline infers "oracle thoughts" from AnnoMI transcripts. An LLM generates candidate thoughts, and only those that make regenerated responses closer to the real counselor reply are kept. MIThinker is trained first with SFT on this data and then with GRPO using composite rewards.

The resulting agent, MindfulMI, matches CAMI in expert-rated MI competency and clearly beats Chain-of-Strategy prompting, at roughly an order of magnitude lower compute. This makes interpretable, theory-grounded counseling agents more practical to deploy.

Key points

  • Data and pipeline: The work uses the 110 high-quality AnnoMI sessions, split 90/10/10 by session.
    • Twenty oracle thoughts are generated per context-response pair.
    • Each thought covers five ToM dimensions (belief, desire, intention, emotion, trust) plus a rationale for the chosen MI strategy.
    • Thoughts are drawing on MI, the Transtheoretical Model, and Theory of Mind.
    • The pipeline yields about 31k training triplets.
  • Filtering: A thought is kept if the GPT-4o response conditioned on it is semantically closer to the ground-truth reply than a no-thought response. Similarity is measured with all-MiniLM-L6-v2. This is a relative criterion rather than a fixed threshold.
  • Training: SFT is followed by GRPO, with rewards for format adherence, thought alignment, and reasoning reasonableness.
  • Expert evaluation:
    • Three experts scored 60 contexts, with moderate agreement (Fleiss' κ = 0.60).
    • They found no significant difference between MIThinker and oracle thoughts.
    • In pairwise comparisons, MIThinker won 32% of the time against oracle thoughts and generally beat CoT and R1-Dist.
  • Limitations:
    • MindfulMI elicits change talk less well than CAMI, especially on relationship and legal topics.
    • The evaluation relies on simulated clients and staged sessions.
    • The structured thoughts sometimes read as less natural than CoT or R1 reasoning.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Singapore Management UniversitySingapore · Singapore35 works · 2,457 citations🎓Beijing Institute of TechnologyBeijing · China43 works · 3,652 citations🎓Australian National UniversityCanberra · Australia24 works · 6,522 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksPsychology · 1,649 worksApplied Psychology · 25 works

Cite

Cite as Yizhe Yang, Palakorn Achananuparp, Heyan Huang, Jing Jiang and Ee-Peng Lim. 2026. MIThinker: A Plug-and-Play Policy-Optimized Thinker For Motivational Interviewing Counseling. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.163

@inproceedings{yang2026mithinker,
  title = {{MIThinker: A Plug-and-Play Policy-Optimized Thinker For Motivational Interviewing Counseling}},
  author = {Yizhe Yang, Palakorn Achananuparp, Heyan Huang, Jing Jiang and Ee-Peng Lim},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.163},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.163}
}

References

48 references, 10 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, and 1 others. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774.
  2. Gallus Bischof, Anja Bischof, and Hans-Jürgen Rumpf. 2021. Motivational interviewing: an evidence-based approach for use in medical practice. Deutsches Ärzteblatt International, 118(7):109–115. find
  3. Sven Buechel and Udo Hahn. 2017. Emobank: Studying the impact of annotation perspective and representation format on dimensional emotion analysis. find
  4. Keqi Chen, Zekai Sun, Huijun Lian, Yingming Gao, and Ya Li. 2025. Psy-copilot: Visual chain of thought for counseling. arXiv preprint arXiv:2503.03645. — [arXiv:2503.03645](https://arxiv.org/abs/2503.03645) find
  5. Yu Ying Chiu, Ashish Sharma, Inna Wanyin Lin, and Tim Althoff. 2024. A computational framework for behavioral assessment of llm therapists. arXiv preprint arXiv:2401.00820. — [arXiv:2401.00820](https://arxiv.org/abs/2401.00820) find
  6. Ben Cohen, Moreah Zisquit, Stav Yosef, Doron Friedman, and Kfir Bar. 2024. Motivational interviewing transcripts annotated with global scores. Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 11642–11657. find
  7. Fabio Cuzzolin, Alice Morelli, Bogdan Cirstea, and Barbara J Sahakian. 2020. Knowing me, knowing you: theory of mind in ai. Psychological medicine, 50(7):1057–1061. find
  8. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, and 1 others. 2024. The llama 3 herd of models. arXiv preprint arXiv:2407.21783.
  9. Paul Ekman. 1992. An argument for basic emotions. Cognition & emotion, 6(3-4):169–200.
  10. Chris Frith and Uta Frith. 2005. Theory of mind. Current biology, 15(17):R644–R645. find
  11. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, and 1 others. 2025. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948.
  12. Mozhdeh Hashemzadeh, Alireza Rahimi, Firoozeh Zare-Farashbandi, Amir Mansur Alavi-Naeini, and Azra Daei. 2019. Transtheoretical model of health behavioral change: A systematic review. Iranian journal of nursing and midwifery research, 24(2):83–90. find
  13. Shang-Ling Hsu, Raj Sanjay Shah, Prathik Senthil, Zahra Ashktorab, Casey Dugan, Werner Geyer, and Diyi Yang. 2023. Helping the helper: Supporting peer counselors via ai-empowered practice and feedback. arXiv preprint arXiv:2305.08982. find
  14. He Hu, Yucheng Zhou, Juzheng Si, Qianning Wang, Hengheng Zhang, Fuji Ren, Fei Ma, and Laizhong Cui. 2025a. Beyond empathy: Integrating diagnostic and therapeutic reasoning with large language models for mental health counseling. arXiv preprint arXiv:2505.15715. — [arXiv:2505.15715](https://arxiv.org/abs/2505.15715) find
  15. Minxue Niu, Yara El-Tawil, Amrit Romana, and Emily Mower Provost. 2025. Rethinking emotion annotations in the era of large language models. IEEE Transactions on Affective Computing. find
  16. He Hu, Yucheng Zhou, Lianzhong You, Hongbo Xu, Qianning Wang, Zheng Lian, Fei Richard Yu, Fei Ma, and Laizhong Cui. 2025b. Emobench-m: Benchmarking emotional intelligence for multimodal large language models. arXiv preprint arXiv:2502.04424. — [arXiv:2502.04424](https://arxiv.org/abs/2502.04424) find
  17. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting of the Association for Computational Linguistics, pages 311–318.
  18. Dongjin Kang, Sunghwan Kim, Taeyoon Kwon, Seungjun Moon, Hyunsouk Cho, Youngjae Yu, Dongha Lee, and Jinyoung Yeo. 2024. Can large language models be good emotional supporter? mitigating preference bias on emotional support conversation. arXiv preprint arXiv:2402.13211. find
  19. James O Prochaska, Susan Butterworth, Colleen A Redding, Verna Burden, Nancy Perrin, Michael Leo, Marna Flaherty-Robb, and Janice M Prochaska. 2008. Initial efficacy of mi, ttm tailoring and hri’s with multiple behaviors for employee health promotion. Preventive medicine, 46(3):226–231. find
  20. Yanyang Li, Shuo Liang, Michael Lyu, and Liwei Wang. 2024. Making long-context language models better multi-hop reasoners. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2462–2475. find
  21. James O Prochaska and Charles C DiClemente. 2005. The transtheoretical approach. Handbook of psychotherapy integration, 2:147–171. find
  22. Chin-Yew Lin. 2004. Rouge: A package for automatic evaluation of summaries. In Text summarization branches out, pages 74–81.
  23. Chris Yuhao Liu, Liang Zeng, Yuzhen Xiao, Jujie He, Jiacai Liu, Chaojie Wang, Rui Yan, Wei Shen, Fuxiang Zhang, Jiacheng Xu, Yang Liu, and Yahui Zhou. 2025. Skywork-reward-v2: Scaling preference data curation via human-ai synergy. arXiv preprint arXiv:2507.01352.
  24. Ilya Loshchilov, Frank Hutter, and 1 others. 2017. Fixing weight decay regularization in adam. arXiv preprint arXiv:1711.05101, 5(5):5. find
  25. Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, and 1 others. 2023. Self-refine: Iterative refinement with self-feedback. Advances in Neural Information Processing Systems, 36:46534–46594. find
  26. William Miller and Stephen Rollnick. 2002. Motivational Interviewing: Preparing People for Change, 2nd ed., volume 25. Journal for Healthcare Quality. find
  27. William R Miller, Theresa B Moyers, Denise Ernst, and Paul Amrhein. 2003. Manual for the motivational interviewing skill code (misc). Unpublished manuscript. Albuquerque: Center on Alcoholism, Substance Abuse and Addictions, University of New Mexico. find
  28. Theresa B Moyers, Lauren N Rowell, Jennifer K Manuel, Denise Ernst, and Jon M Houck. 2016. The motivational interviewing treatment integrity code (miti 4): rationale, preliminary reliability and validity. Journal of substance abuse treatment, 65:36–42. find
  29. JO Prochaska and WF Velicer. 1997. The transtheoretical model of health behavior change. American Journal of Health Promotion, 12(1):38–48. find
  30. Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. 2023. Direct preference optimization: Your language model is secretly a reward model. Advances in neural information processing systems, 36:53728– 53741.
  31. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Y Wu, and 1 others. 2024. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300.
  32. Ian Steenstra, Farnaz Nouraei, Mehdi Arjmand, and Timothy W Bickmore. 2024. Virtual agents for alcohol use counseling: exploring llm-powered motivational interviewing. arXiv preprint arXiv:2407.08095. — [arXiv:2407.08095](https://arxiv.org/abs/2407.08095) find
  33. Xin Sun, Xiao Tang, Abdallah El Ali, Zhuying Li, Xiaoyu Shen, Pengjie Ren, Jan de Wit, Jiahuan Pei, and Jos A Bosch. 2024. Chain-of-strategy planning with llms: Aligning the generation of psychotherapy dialogue with strategy in motivational interviewing. arXiv preprint arXiv:2408.06527. — [arXiv:2408.06527](https://arxiv.org/abs/2408.06527) find
  34. Leandro von Werra, Younes Belkada, Lewis Tunstall, Edward Beeching, Tristan Thrush, Nathan Lambert, Shengyi Huang, Kashif Rasul, and Quentin Gallouédec. 2020. Trl: Transformer reinforcement learning. https://github.com/huggingface/trl. ↗
  35. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Ed Chi, Quoc Le, and Denny Zhou. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in Neural Information Processing Systems, 35:24824–24837.
  36. Anuradha Welivita and Pearl Pu. 2022. Curating a large-scale motivational interviewing dataset using peer support forums. In Proceedings of the 29th International Conference on Computational Linguistics, pages 3315–3330. find
  37. Henry M Wellman. 2018. Theory of mind: The state of the art. European Journal of Developmental Psychology, 15(6):728–755. find
  38. Heinz Wimmer and Josef Perner. 1983. Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children’s understanding of deception. Cognition, 13(1):103–128. find
  39. Zixiu Wu, Simone Balloccu, Vivek Kumar, Rim Helaoui, Diego Reforgiato Recupero, and Daniele Riboni. 2023. Creation, analysis and evaluation of annomi, a dataset of expert-annotated counselling dialogues. Future Internet, 15(3):110. find
  40. Zixiu Wu, Simone Balloccu, Vivek Kumar, Rim Helaoui, Ehud Reiter, Diego Reforgiato Recupero, and Daniele Riboni. 2022. Anno-mi: A dataset of expert-annotated counselling dialogues. In ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 6177–6181.… find
  41. Zhouhang Xie, Bodhisattwa Prasad Majumder, Mengjie Zhao, Yoshinori Maeda, Keiichi Yamada, Hiromi Wakaki, and Julian McAuley. 2024. Few-shot dialogue strategy learning for motivational interviewing via inductive reasoning. arXiv preprint arXiv:2403.15737. find
  42. Tao Yang, Tianyuan Shi, Fanqi Wan, Xiaojun Quan, Qifan Wang, Bingzhe Wu, and Jiaxiang Wu. 2023. Psycot: psychological questionnaire as powerful chain-of-thought for personality detection. arXiv preprint arXiv:2310.20256. find
  43. Yizhe Yang, Palakorn Achananuparp, Heyan Huang, Jing Jiang, Kit Phey Leng, Nicholas Gabriel Lim, Cameron Tan Shi Ern, and Ee-peng Lim. 2025. Cami: A counselor agent supporting motivational interviewing through state inference and topic exploration. arXiv preprint arXiv:2502.02807. find
  44. Tenggan Zhang, Xinjie Zhang, Jinming Zhao, Li Zhou, and Qin Jin. 2024a. ESCoT: Towards interpretable emotional support dialogue systems. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 13395–13412, Bangkok, Thailand. Associati… find
  45. Tianyi Zhang, Varsha Kishore, Felix Wu, Kilian Q Weinberger, and Yoav Artzi. Bertscore: Evaluating text generation with bert. In International Conference on Learning Representations. find
  46. Xuan Zhang, Chao Du, Tianyu Pang, Qian Liu, Wei Gao, and Min Lin. 2024b. Chain of preference optimization: Improving chain-of-thought reasoning in llms. Advances in Neural Information Processing Systems, 37:333–356. find
  47. Yucheng Zhou, Xiang Li, Qianning Wang, and Jianbing Shen. 2024. Visual in-context learning for large vision-language models. arXiv preprint arXiv:2402.11574. find
  48. Yucheng Zhou, Lingran Song, and Jianbing Shen. 2025. Training medical large vision-language models with abnormal-aware feedback. arXiv preprint arXiv:2501.01377. — [arXiv:2501.01377](https://arxiv.org/abs/2501.01377) find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/yang-2026-mithinker-plug-play-policy-optimized/