Elamur Research⌕
☰
home / papers / 2026 / I Don't Need Solution. I Need Emotional Support : Empathetic LLMs based on Emotional Validation

I Don't Need Solution. I Need Emotional Support : Empathetic LLMs based on Emotional Validation

Paper · 2026 · Suhyune Son, Jungwoo Lim, Myunghoon Kang et al.

Summary

LLMs used as emotional-support chatbots tend to jump straight to advice. They repeat stock phrases and ignore how the help-seeker's feelings evolve over a dialogue. In a pilot study on ESConv, crowd workers judged most Mistral and GPT-4o replies as not comforting. To fix this, the authors propose EVA, a framework built on the theory of emotional validation developed within dialectical behavior therapy (DBT). That theory orders supportive responses into four levels that deepen as a conversation progresses: Listening and Observing, Accurate Reflection, Validating, and Radical Genuineness.

EVA trains models in two stages. First comes empathy acquisition, then alignment to validation levels using a new dataset, EVAD. EVAD is built from ESConv and ED, with responses ranked by psychology-trained annotators. The authors also introduce EVAEval, a metric that uses a RoBERTa classifier to check whether generated replies follow the expected level progression.

EVA-based models beat baseline LLMs in both automatic and expert human evaluation. They also produce fewer formulaic phrases and less of a lean toward suggesting solutions. The work matters for building conversational agents that respond to emotional needs rather than just giving advice.

Key points

  • Pilot study: 100 ESConv test dialogues were answered by Mistral-7B-Instruct and GPT-4o and rated by 15 crowd workers.
    • 87.82% of responses were judged not comforting (Fleiss' Kappa 0.69).
    • Suggesting Solution was the dominant response type, at 40.5% (Mistral) and 61.3% (GPT-4o).
    • Understanding and Validation responses were rare.
  • EVAD annotation: 12 annotators holding psychology degrees ranked ChatGPT-generated candidate responses by validation level and removed poor-quality ones.
  • Final human evaluation: carried out by a separate group of 15 psychology majors familiar with emotional validation theory.
  • EVAEval limitation: the metric depends on a RoBERTa level classifier, whose errors mostly confuse adjacent levels.
  • Code: released at github.com/sonsuhyune/EVA. ESConv is used for academic research only, and ED is licensed CC BY-NC 4.0.

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

🎓Korea UniversitySeoul · South Korea26 works · 9,490 citations🎓Korea UniversityTokyo · Japan3 works

 Fields, topics and concepts

📚 Shelves 6The 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 works

🧭 Topics 1Research subjects and languages the work is about.Sentiment & emotion · 972 works

Cite

Cite as Suhyune Son, Jungwoo Lim, Myunghoon Kang, Seongtae Hong, Yuna Hur, Evelyn H. Zi and Heuiseok Lim. 2026. I Don't Need Solution. I Need Emotional Support : Empathetic LLMs based on Emotional Validation. 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.1

@inproceedings{son2026i,
  title = {{I Don't Need Solution. I Need Emotional Support : Empathetic LLMs based on Emotional Validation}},
  author = {Suhyune Son, Jungwoo Lim, Myunghoon Kang, Seongtae Hong, Yuna Hur, Evelyn H. Zi and Heuiseok Lim},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1}
}

References

56 references, 15 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, et al. 2023. Gpt-4 technical report. arXiv preprint arXiv:2303.08774.
  2. William A Anthony. 1971. A methodological investigation of the" minimally facilitative level of interpersonal functioning". Journal of Clinical Psychology, 27(1). find
  3. Godfrey T Barrett-Lennard. 1993. The phases and focus of empathy. British journal of medical psychology, 66(1):3–14. find
  4. Rishi Bommasani, Drew A Hudson, Ehsan Adeli, Russ Altman, Simran Arora, Sydney von Arx, Michael S Bernstein, Jeannette Bohg, Antoine Bosselut, Emma Brunskill, et al. 2021. On the opportunities and risks of foundation models. arXiv preprint arXiv:2108.07258.
  5. Clara E Hill. 2020. Helping skills: Facilitating exploration, insight, and action. American Psychological Association. find
  6. Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2021. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685.
  7. Brant R Burleson. 2003. Emotional support skills. In Handbook of communication and social interaction skills, pages 569–612. Routledge. find
  8. Gillian Cameron, David Cameron, Gavin Megaw, Raymond Bond, Maurice Mulvenna, Siobhan O’Neill, Cherie Armour, and Michael McTear. 2017. Towards a chatbot for digital counselling. In Proceedings of the 31st international BCS human computer interaction conference (HCI 2017). BCS Learning & Development. find
  9. Albert Q Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, et al. 2023. Mistral 7b. arXiv preprint arXiv:2310.06825. find
  10. 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
  11. Yaru Cao, Hongzhi Yu, and Fucheng Wan. 2024. Enhancing emotional support conversation system via integrating mental state-strategy reasoning. In 2024 6th International Conference on Machine Learning, Big Data and Business Intelligence (MLBDBI), pages 259–263. IEEE. find
  12. William Kidder, Jason D’Cruz, and Kush R Varshney. 2024. Empathy and the right to be an exception: What llms can and cannot do. arXiv preprint arXiv:2401.14523. — [arXiv:2401.14523](https://arxiv.org/abs/2401.14523) find
  13. Yirong Chen, Xiaofen Xing, Jingkai Lin, Huimin Zheng, Zhenyu Wang, Qi Liu, and Xiangmin Xu. 2023. Soulchat: Improving llms’ empathy, listening, and comfort abilities through fine-tuning with multi-turn empathy conversations. In Findings of the Association for Computational Linguistics: EMNLP 2023, p… find
  14. Young-Jun Lee, Chae-Gyun Lim, and Ho-Jin Choi. 2022. Does gpt-3 generate empathetic dialogues? a novel in-context example selection method and automatic evaluation metric for empathetic dialogue generation. In Proceedings of the 29th International Conference on Computational Linguistics, pages 669–… find
  15. Shauna Concannon, Ian Roberts, and Marcus Tomalin. 2023a. An interactional account of empathy in human-machine communication. Human-Machine Communication, 6. find
  16. Jiwei Li, Michel Galley, Chris Brockett, Jianfeng Gao, and William B Dolan. 2016. A diversity-promoting objective function for neural conversation models. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies…
  17. Shauna Concannon, Ian Roberts, and Marcus Tomalin. 2023b. An interactional account of empathy in human-machine communication. Human-Machine Communication. find
  18. Jean Decety. 2010. The neurodevelopment of empathy in humans. Developmental neuroscience, 32(4):257– 267. find
  19. Qintong Li, Hongshen Chen, Zhaochun Ren, Pengjie Ren, Zhaopeng Tu, and Zhumin Chen. 2019. Empdg: Multiresolution interactive empathetic dialogue generation. arXiv preprint arXiv:1911.08698. — [arXiv:1911.08698](https://arxiv.org/abs/1911.08698) find
  20. Nancy Eisenberg and Natalie D Eggum. 2009. Empathic responding: Sympathy and personal distress. The social neuroscience of empathy, 6(2009):71–830. find
  21. Chin-Yew Lin. 2004. ROUGE: A package for automatic evaluation of summaries. In Text Summarization Branches Out, pages 74–81, Barcelona, Spain. Association for Computational Linguistics.
  22. Vojtech Formanek and Ondrej Sotolar. 2024. Quantitative assessment of intersectional empathetic bias and understanding. arXiv preprint arXiv:2411.05777. — [arXiv:2411.05777](https://arxiv.org/abs/2411.05777) find
  23. Fengyi Fu, Lei Zhang, Quan Wang, and Zhendong Mao. 2023. E-core: Emotion correlation enhanced empathetic dialogue generation. arXiv preprint arXiv:2311.15016. find
  24. Zhaojiang Lin, Andrea Madotto, Jamin Shin, Peng Xu, and Pascale Fung. 2019. Moel: Mixture of empathetic listeners. arXiv preprint arXiv:1908.07687.
  25. George M Gazda et al. 1977. Human relations development: A manual for educators. find
  26. Marsha M Linehan. 1997. Validation and psychotherapy. find
  27. Xu Han, Zhengyan Zhang, Ning Ding, Yuxian Gu, Xiao Liu, Yuqi Huo, Jiezhong Qiu, Yuan Yao, Ao Zhang, Liang Zhang, et al. 2021. Pre-trained models: Past, present and future. AI Open, 2:225–250. find
  28. June M Liu, Donghao Li, He Cao, Tianhe Ren, Zeyi Liao, and Jiamin Wu. 2023. Chatcounselor: A large language models for mental health support. arXiv preprint arXiv:2309.15461. — [arXiv:2309.15461](https://arxiv.org/abs/2309.15461) find
  29. Siyang Liu, Chujie Zheng, Orianna Demasi, Sahand Sabour, Yu Li, Zhou Yu, Yong Jiang, and Minlie Huang. 2021. Towards emotional support dialog systems. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Lan…
  30. Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. 2024. Direct preference optimization: Your language model is secretly a reward model. Advances in Neural Information Processing Systems, 36.
  31. Samyam Rajbhandari, Jeff Rasley, Olatunji Ruwase, and Yuxiong He. 2020. Zero: Memory optimizations toward training trillion parameter models. In SC20: International Conference for High Performance Computing, Networking, Storage and Analysis, pages 1– 16. find
  32. Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692.
  33. Hannah Rashkin, Eric Michael Smith, Margaret Li, and Y-Lan Boureau. 2019. Towards empathetic open-domain conversation models: A new benchmark and dataset. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 5370–5381, Florence, Italy. Association for Com… find
  34. Navonil Majumder, Pengfei Hong, Shanshan Peng, Jiankun Lu, Deepanway Ghosal, Alexander Gelbukh, Rada Mihalcea, and Soujanya Poria. 2020. Mime: Mimicking emotions for empathetic response generation. arXiv preprint arXiv:2010.01454.
  35. William J Reynolds and Brain Scott. 1999. Empathy: a crucial component of the helping relationship. Journal of psychiatric and mental health nursing, 6(5):363–370. find
  36. Meta. 2024. Introducing meta llama 3: The most capable openly available llm to date. find
  37. Derek Richards. 2009. Features and benefits of online counselling: Trinity college online mental health community. British Journal of Guidance & Counselling, 37(3):231–242. find
  38. Kevin M Mitchell and Bernard G Berenson. 1970. Differential use of confrontation by high and low facilitative therapists. The Journal of Nervous and Mental Disease, 151(5):303–309. find
  39. H Rudolph Schaffer. 1996. Social development. Blackwell Publishing. find
  40. Anand Mohan, Dighreandr Singh, Aakanksha Kataria, Hemant Kumar Meena, and Reeta Singh. 2025. The impact of ai on human psychology: A user perspective. 2025 International Conference on Cognitive Computing in Engineering, Communications, Sciences and Biomedical Health Informatics (IC3ECSBHI), pages 13… find
  41. Ashish Sharma, Adam Miner, David Atkins, and Tim Althoff. 2020. A computational approach to understanding empathy expressed in text-based mental health support. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 5263–5276.
  42. OpenAI-Blog. 2022. Chatgpt: Optimizing language models for dialogue. find
  43. Ekaterina Svikhnushina, Iuliana Voinea, Anuradha Welivita, and Pearl Pu. 2022. A taxonomy of empathetic questions in social dialogs. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2952–2973. find
  44. 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.
  45. Dehua Tao, Tan Lee, Harold Chui, and Sarah Luk. 2024. Modeling intrapersonal and interpersonal influences for automatic estimation of therapist empathy in counseling conversation. In ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), pages 12692–12696… find
  46. Yushan Qian, Wei-Nan Zhang, and Ting Liu. 2023. Harnessing the power of large language models for empathetic response generation: Empirical investigations and improvements. arXiv preprint arXiv:2310.05140. find
  47. EmoLLM Team. 2024a. Emollm: Reinventing mental health support with large language models. https://github.com/SmartFlowAI/EmoLLM. ↗
  48. Huachuan Qiu, Hongliang He, Shuai Zhang, Anqi Li, and Zhenzhong Lan. 2023. Smile: Single-turn to multi-turn inclusive language expansion via chatgpt for mental health support. arXiv preprint arXiv:2305.00450.
  49. Qwen Team. 2024b. Hello qwen2. find
  50. Amy JC Trappey, Aislyn PC Lin, Kevin YK Hsu, Charles V Trappey, and Kevin LK Tu. 2022. Development of an empathy-centric counseling chatbot system capable of sentimental dialogue analysis. Processes, 10(5):930. find
  51. 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.
  52. A Lanrui Wang, Jiangnan Li, Chenxu Yang, Zheng Lin, Hongyin Tang, Huan Liu, Yanan Cao, Jingang Wang, and Weiping Wang. 2025a. Sibyl: Empowering empathetic dialogue generation in large language models via sensible and visionary commonsense inference. In Proceedings of the 31st International Conferenc…
  53. Comparison of responses A comparison of responses generated by Mistral-7B-Instruct and our proposed EVA M istral to the help-seeker’s utterances is shown in Figure 1. B Xu Wang, Bo Wang, Yihong Tang, Dongming Zhao, Jing Liu, Ruifang He, and Yuexian Hou. 2025b. Ecc: Synergizing emotion, cause and com… find
  54. Prompt Template Prompt You are a helpful and considerate supporter. Your job is to generate appropriate emotional supporting utterances based on the help-seeker’s utterance. Please generate the supporter’s utterance based on the given dialogue history. find
  55. Jing Wei, Sungdong Kim, Hyunhoon Jung, and Young-Ho Kim. 2024. Leveraging large language models to power chatbots for collecting user self-reported data. Proceedings of the ACM on Human-Computer Interaction, 8(CSCW1):1–35. {dialogue history} supporter: *Table 3: Prompt templates for LLMs including G… find
  56. Zhou Yang, Zhaochun Ren, Wang Yufeng, Haizhou Sun, Chao Chen, Xiaofei Zhu, and Xiangwen Liao. 2024b. An iterative associative memory model for empathetic response generation. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 308… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/son-2026-i-don-t-need-solution/