Elamur Research⌕
☰
home / papers / 2026 / Psychological Counseling Cannot Be Achieved Overnight: Automated Psychological Counseling Through Multi-Session Conversations

Psychological Counseling Cannot Be Achieved Overnight: Automated Psychological Counseling Through Multi-Session Conversations

Paper · 2026 · Bichen Wang, Junzhe Wang, Yixin Sun et al.

Summary

Most LLM-based counseling systems treat therapy as a single conversation. Real therapy such as Cognitive Behavioral Therapy (CBT) unfolds over many visits, with goals revised as the client changes. This paper frames automated counseling as a longitudinal task made of three parts: memory extraction, goal planning and counseling generation. Because real transcripts cannot be shared for privacy reasons, the authors built MusPsy-Dataset, a synthetic CBT-grounded corpus. It is derived from published psychological case reports that record each client's evolving condition and per-session goals.

Generation is top-down. Short seed dialogues first sketch the whole multi-session arc, and each seed is then expanded into a full session, using earlier seeds, CBT technique descriptions and real counseling fragments as guidance. Session-level memories summarize the client's progress. On this data they trained MusPsy-Model, a memory-augmented counselor. It was evaluated with basic indicators, WAI and PANAS against a simulated LLM client. Baselines tend to stall and repeat themselves, while MusPsy-Model keeps improving client affect across sessions. This shows why continuity matters for counseling agents.

Key points

  • Data sources: client profiles combine static traits (pseudonym, gender, age, occupation, initial issues) with dynamic, session-by-session states (life events, emotion, cognition). Counseling goals are stored separately as the counselor's plan. All of this is extracted from case reports by GPT-4o and checked manually.
  • Seed conversations: each seed has only 3–4 turns (about 100–200 tokens) and is generated in one pass for all sessions together, so the cross-session flow stays coherent before expansion.
  • Evaluation design: the model was tested over 6 sessions with an LLM-simulated client, using WAI (working alliance) and PANAS (affect). Human expert evaluation was added to offset bias from GPT-4o-based judging.
  • Main result: after roughly two sessions, baselines produce near-duplicate dialogues. MusPsy-Model reduces negative and increases positive emotion over the long run, and its alliance scores trend upward.
  • Availability and limitations: the dataset is released at github.com/Rener2005/MusPsy. The authors note several limitations: it covers CBT only, gives little attention to resistant clients, relies on synthetic data whose realism is limited, and raises privacy concerns around memory.

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

🎓Harbin Institute of TechnologyHarbin · China158 works · 28,841 citations

 Fields, topics and concepts

📚 Shelves 7The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksPsychology · 1,552 worksExperimental and Cognitive Psychology · 870 worksApplied Psychology · 21 works

Cite

Cite as Bichen Wang, Junzhe Wang, Yixin Sun, Xing Fu, Yanyan Zhao and Bing Qin. 2026. Psychological Counseling Cannot Be Achieved Overnight: Automated Psychological Counseling Through Multi-Session Conversations. 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.818

@inproceedings{wang2026psychological,
  title = {{Psychological Counseling Cannot Be Achieved Overnight: Automated Psychological Counseling Through Multi-Session Conversations}},
  author = {Bichen Wang, Junzhe Wang, Yixin Sun, Xing Fu, Yanyan Zhao and Bing Qin},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.818},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.818}
}

References

30 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Anaïs Baur, Anne Trösken, and Babette Renneberg. 2024. Content and attainment of individual treatment goals in cbt. Psychotherapy Research, 34(1):111– 123. find
  2. Judith S Beck. 2020. Cognitive behavior therapy: Basics and beyond. Guilford Publications. find
  3. Judith S Beck, AT Beck, and JS Beck. 2011. Cognitive behavior therapy: basics and beyond. ed. New York. find
  4. Adam O Horvath. 2001. The alliance. Psychotherapy: Theory, research, practice, training, 38(4):365. find
  5. Adam O Horvath, AC Del Re, Christoph Flückiger, and Dianne Symonds. 2011. Alliance in individual psychotherapy. Psychotherapy, 48(1):9. find
  6. Adam O Horvath and B Dianne Symonds. 1991. Relation between working alliance and outcome in psychotherapy: A meta-analysis. Journal of counseling psychology, 38(2):139. find
  7. Linette Lawlor-Savage and Jennifer L Prentice. 2014. Digital cognitive behaviour therapy (cbt) in canada: Ethical considerations. Canadian Psychology/Psychologie canadienne, 55(4):231. find
  8. Sueon Lee, Sunghwan Mac Kim, Minju Kim, Dongjin Kang, Dongil Yang, Harim Kim, Minseok Kang, Dayi Jung, Min Kim, Seungbeen Lee, and 1 others. 2024. Cactus: Towards psychological counseling conversations using cognitive behavioral theory. In Findings of the Association for Computational Linguistics: E… find
  9. Taryn B Bemister and Keith S Dobson. 2011. An updated account of the ethical and legal considerations of record keeping. Canadian Psychology/Psychologie canadienne, 52(4):296. find
  10. Richard J Longmore and Michael Worrell. 2007. Do we need to challenge thoughts in cognitive behavior therapy? Clinical psychology review, 27(2):173– 187. find
  11. Michelle G Craske. 2010. Cognitive–behavioral therapy. American Psychological Association. find
  12. Thomas Munder, Fabian Wilmers, Rainer Leonhart, Hans Wolfgang Linster, and Jürgen Barth. 2010. Working alliance inventory-short revised (wai-sr): psychometric properties in outpatients and inpatients. Clinical Psychology & Psychotherapy: An International Journal of Theory & Practice, 17(3):231–239. find
  13. Keith S Dobson and David JA Dozois. 2021. Handbook of cognitive-behavioral therapies. Guilford Publications. find
  14. 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.
  15. Kristina Fenn and Majella Byrne. 2013. The key principles of cognitive behavioural therapy. InnovAiT, 6(9):579–585. find
  16. Xing Fu, Haozhen Li, Bichen Wang, Hao Yang, Yanyan Zhao, and Bing Qin. 2025. Look beyond feeling: Unveiling latent needs from implicit expressions for proactive emotional support. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 21571–21598, Suzhou, Ch… find
  17. Kiran L Grant, Magenta Bender Simmons, and Christopher G Davey. 2018. Three nontraditional approaches to improving the capacity, accessibility, and quality of mental health services: An overview. Psychiatric Services, 69(5):508–516. find
  18. Steven C Hayes and Stefan G Hofmann. 2018. Process-based CBT: The science and core clinical competencies of cognitive behavioral therapy. New Harbinger Publications. find
  19. Hongbin Na. 2024. Cbt-llm: A chinese large language model for cognitive behavioral therapy-based mental health question answering. In Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024), pages 2930–2940. find
  20. World Health Organization. 2024. Global tuberculosis report 2024. World Health Organization. find
  21. Huachuan Qiu, Hongliang He, Shuai Zhang, Anqi Li, and Zhenzhong Lan. 2024. Smile: Single-turn to multi-turn inclusive language expansion via chatgpt for mental health support. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 615–636, Miami, Florida, USA. Association fo…
  22. Huachuan Qiu and Zhenzhong Lan. 2024. Interactive agents: Simulating counselor-client psychological counseling via role-playing llm-to-llm interactions. arXiv preprint arXiv:2408.15787.
  23. Carl R Rogers. 1952. " client-centered" psychotherapy. Scientific American, 187(5):66–75. find
  24. Hasina Samji, Judy Wu, Amilya Ladak, Caralyn Vossen, Evelyn Stewart, Naomi Dove, David Long, and Gaelen Snell. 2022. Mental health impacts of the covid-19 pandemic on children and youth–a systematic review. Child and adolescent mental health, 27(2):173– 189. find
  25. Hao Sun, Zhenru Lin, Chujie Zheng, Siyang Liu, and Minlie Huang. 2021. PsyQA: A Chinese dataset for generating long counseling text for mental health support. In Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021, pages 1489–1503, Online. Association for Computational Linguis… find
  26. Joseph Weizenbaum. 1966. Eliza—a computer program for the study of natural language communication between man and machine. Communications of the ACM, 9(1):36–45. find
  27. Mengxi Xiao, Qianqian Xie, Ziyan Kuang, Zhicheng Liu, Kailai Yang, Min Peng, Weiguang Han, and Jimin Huang. 2024. HealMe: Harnessing cognitive reframing in large language models for psychotherapy. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: L… find
  28. Chenhao Zhang, Renhao Li, Minghuan Tan, Min Yang, Jingwei Zhu, Di Yang, Jiahao Zhao, Guancheng Ye, Chengming Li, and Xiping Hu. 2024. CPsyCoun: A report-based multi-turn dialogue reconstruction and evaluation framework for Chinese psychological counseling. In Findings of the Association for Computat… find
  29. Weixiang Zhao, Yanyan Zhao, and Xin Lu. 2022. Cauain: Causal aware interaction network for emotion recognition in conversations. In IJCAI, pages 4524–4530. find
  30. Weixiang Zhao, Yanyan Zhao, Shilong Wang, and Bing Qin. 2023. Transesc: Smoothing emotional support conversation via turn-level state transition. In Findings of the Association for Computational Linguistics: ACL 2023, pages 6725–6739. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2026-psychological-counseling-cannot-be-achieved/