Elamur Research⌕
☰
home / papers / 2026 / Feeling Right vs. Being Right: How AI Sycophancy Affects Value-Laden Deliberation

Feeling Right vs. Being Right: How AI Sycophancy Affects Value-Laden Deliberation

Paper · 2026 · Jeongwoo Ryu, Soomin Kim, Jinsu Eun et al.

Summary

People now often ask chatbots for help with personal dilemmas that have no single correct answer. This paper asks how sycophancy, the RLHF-induced habit of telling users what they want to hear, shapes this kind of moral deliberation. Using Goffman's face-work theory, the authors define two styles. Active sycophancy protects the user's positive face through agreement and praise. Passive sycophancy protects negative face by avoiding any challenge and stressing the user's autonomy. A non-sycophantic neutral style served as the baseline. In a mixed-methods study, 31 experienced AI users discussed three tragic personal dilemmas with an AI under these conditions. Both sycophantic styles raised participants' confidence in their decisions. They also reduced open-minded thinking, and participants felt supported but described the conversations as unproductive. The neutral style felt uncomfortable at first but led to more flexible and meaningful reflection. The authors call this a confidence-competence trade-off. They argue that AI used for personal counsel needs calibrated friction and explicit safety boundaries rather than automatic agreement.

Key points

  • Participants: 32 people were recruited and one was excluded for low engagement, leaving N = 31 (16 men, 15 women; mean age 30.06, range 20–41). All scored at least 3 out of 5 on prior AI experience. Each received about $15.
  • Dilemmas: The three scenarios came from previously published dilemma sets, one by Chiu and colleagues in 2024 and another by Yudkin and colleagues in 2025:
    • reporting a friend's crime versus letting an innocent suspect be blamed;
    • revealing a fiancé's infidelity before the wedding versus staying silent;
    • funding a sibling's failing business versus protecting a romantic relationship.
  • Measures and conditions: The study focused on open-minded thinking and on who participants held responsible for the decision. It compared active sycophancy, passive sycophancy and a neutral baseline.
  • Context: All sessions were conducted in Korean with Korean participants. Each was a single, short session using hypothetical scenarios, which limits how far the findings generalise.
  • Ethical recommendations: The authors call for transparency about how agreeable an AI is designed to be. They also stress separating validation of a user's feelings from validation of their choices.

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

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Social Sciences · 1,772 worksSafety Research · 543 worksSociology and Political Science · 577 worksPolitical Science and International Relations · 182 works

Cite

Cite as Jeongwoo Ryu, Soomin Kim, Jinsu Eun, Kyusik Kim, Changhoon Oh and Bongwon Suh. 2026. Feeling Right vs. Being Right: How AI Sycophancy Affects Value-Laden Deliberation. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.2046

@inproceedings{ryu2026feeling,
  title = {{Feeling Right vs. Being Right: How AI Sycophancy Affects Value-Laden Deliberation}},
  author = {Jeongwoo Ryu, Soomin Kim, Jinsu Eun, Kyusik Kim, Changhoon Oh and Bongwon Suh},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.2046},
  url = {https://doi.org/10.18653/v1/2026.acl-long.2046}
}

References

44 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Yuntao Bai, Andy Jones, Kamal Ndousse, Amanda Askell, Anna Chen, Nova DasSarma, Dawn Drain, Stanislav Fort, Deep Ganguli, Tom Henighan, Nicholas Joseph, Saurav Kadavath, Jackson Kernion, Tom Conerly, Sheer El-Showk, Nelson Elhage, Zac Hatfield-Dodds, Danny Hernandez, Tristan Hume, and 12 others. 202… find
  2. Tiffany Barnett White. 2005. Consumer Trust and Advice Acceptance: The Moderating Roles of Benevolence, Expertise, and Negative Emotions. Journal of Consumer Psychology, 15(2):141–148. find
  3. Virginia Braun and Victoria Clarke. 2006. Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2):77–101. _eprint: https://doi.org/10.1191/1478088706qp063oa. — [doi:10.1191/1478088706qp063oa](https://doi.org/10.1191/1478088706qp063oa) ↗
  4. Zana Buçinca, Maja Barbara Malaya, and Krzysztof Z. Gajos. 2021. To Trust or to Think: Cognitive Forcing Functions Can Reduce Overreliance on AI in AI-assisted Decision-making. Proc. ACM Hum.-Comput. Interact., 5(CSCW1):188:1–188:21. find
  5. María Victoria Carro. 2024. Flattering to Deceive: The Impact of Sycophantic Behavior on User Trust in Large Language Model. arXiv preprint. ArXiv:2412.02802 [cs]. — [arXiv:2412.02802](https://arxiv.org/abs/2412.02802) find
  6. David Chai, Jian Li, and Jinsong Huang. 2025. Machine talk: When flattery sounds better from a bot. Journal of Retailing and Consumer Services, 88:104465. find
  7. Wei Chen, Zhen Huang, Liang Xie, Binbin Lin, Houqiang Li, Le Lu, Xinmei Tian, Deng Cai, Yonggang Zhang, Wenxiao Wang, Xu Shen, and Jieping Ye. 2024. From Yes-Men to Truth-Tellers: Addressing Sycophancy in Large Language Models with Pinpoint Tuning. Version Number: 3. find
  8. Conversational Argument Search Under Selective Exposure: Strategies for Balanced Perspective Access. In Proceedings of the 48th International ACM SI-GIR Conference on Research and Development in Information Retrieval, pages 2837–2842, Padua Italy. ACM. find
  9. Andrej Krištofík. 2025. Bias in AI (Supported) Decision Making: Old Problems, New Technologies. International Journal for Court Administration, 16(1):3. find
  10. Myra Cheng, Sunny Yu, Cinoo Lee, Pranav Khadpe, Lujain Ibrahim, and Dan Jurafsky. 2025. Social Sycophancy: A Broader Understanding of LLM Sycophancy. arXiv preprint. ArXiv:2505.13995 [cs] version: 1.
  11. Sebastian Krügel, Andreas Ostermaier, and Matthias Uhl. 2023. ChatGPT’s inconsistent moral advice influences users’ judgment. Scientific Reports, 13(1):4569. find
  12. Yu Ying Chiu, Liwei Jiang, and Yejin Choi. 2024. DailyDilemmas: Revealing Value Preferences of LLMs with Quandaries of Daily Life. find
  13. Jonathan Kwik. 2025. Digital Yes-Men: How to Deal With Sycophantic Military AI? Global Policy, 16(3):467–473. _eprint: https://onlinelibrary.wiley.com/doi/pdf/10.1111/1758-5899.70042. — [doi:10.1111/1758-5899.70042](https://doi.org/10.1111/1758-5899.70042) ↗
  14. P. Christiano, Jan Leike, Tom B. Brown, Miljan Martic, S. Legg, and Dario Amodei. 2017. Deep Reinforcement Learning from Human Preferences. ArXiv. find
  15. Aaron Fanous, Jacob Goldberg, Ank A. Agarwal, Joanna Lin, Anson Zhou, Roxana Daneshjou, and Sanmi Koyejo. 2025. SycEval: Evaluating LLM Sycophancy. arXiv preprint. ArXiv:2502.08177 [cs]. — [arXiv:2502.08177](https://arxiv.org/abs/2502.08177) find
  16. Jillian Fisher, Shangbin Feng, Robert Aron, Thomas Richardson, Yejin Choi, Daniel W Fisher, Jennifer Pan, Yulia Tsvetkov, and Katharina Reinecke. 2025. Biased LLMs can Influence Political Decision-Making. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Vol…
  17. Moshe Glickman and Tali Sharot. 2024. How human–AI feedback loops alter human perceptual, emotional and social judgements. Nature Human Behaviour, 9(2):345–359. find
  18. Erving Goffman. 1955. On Face-Work: An Analysis of Ritual Elements in Social Interaction. Psychiatry, 18(3):213–231. find
  19. Tom Hitron, Noa Morag Yaar, and Hadas Erel. 2023. Implications of AI Bias in HRI: Risks (and Opportunities) when Interacting with a Biased Robot. Proceedings of the 2023 ACM/IEEE International Conference on Human-Robot Interaction, pages 83– 92. Conference Name: HRI ’23: ACM/IEEE International Confe… find
  20. Lujain Ibrahim, Franziska Sofia Hafner, and Luc Rocher. 2025. Training language models to be warm and empathetic makes them less reliable and more sycophantic. arXiv preprint. ArXiv:2507.21919 [cs]. — [arXiv:2507.21919](https://arxiv.org/abs/2507.21919) find
  21. Kyusik Kim, Jeongwoo Ryu, Dongseok Heo, Hyung-woo Song, Changhoon Oh, and Bongwon Suh. 2025. find
  22. Philippe Laban, Lidiya Murakhovs’ka, Caiming Xiong, and Chien-Sheng Wu. 2023. Are You Sure? Challenging LLMs Leads to Performance Drops in The FlipFlop Experiment. Version Number: 2. find
  23. Joshua Liu, Aarav Jain, Soham Takuri, Srihan Vege, Aslihan Akalin, Kevin Zhu, Sean O’Brien, and Vasu Sharma. 2025. TRUTH DECAY: Quantifying Multi-Turn Sycophancy in Language Models. arXiv preprint. ArXiv:2503.11656 [cs]. — [arXiv:2503.11656](https://arxiv.org/abs/2503.11656) find
  24. Lars Malmqvist. 2024. Sycophancy in Large Language Models: Causes and Mitigations. Version Number: 1.
  25. Julian A. Nasello and Jean-Marc Triffaux. 2023. The role of empathy in trolley problems and variants: A systematic review and meta-analysis. British Journal of Social Psychology, 62(4):1753–1781. find
  26. Raymond S. Nickerson. 1998. Confirmation bias: A ubiquitous phenomenon in many guises. Review of General Psychology, 2(2):175–220. find
  27. Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L. Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke E. Miller, Maddie Simens, Amanda Askell, Peter Welinder, P. Christiano, Jan Leike, and Ryan J. Lowe. 2022. T… — Training language models to follow instructions with human feedback
  28. Ethan Perez, Sam Ringer, Kamile Lukosiute, Karina Nguyen, Edwin Chen, Scott Heiner, Craig Pettit, Catherine Olsson, Sandipan Kundu, Saurav Kadavath, Andy Jones, Anna Chen, Benjamin Mann, Brian Israel, Bryan Seethor, Cameron McKinnon, Christopher Olah, Da Yan, Daniela Amodei, and 44 others. 2023. Dis… — Discovering Language Model Behaviors with Model-Written Evaluations
  29. Priya Pitre, Naren Ramakrishnan, and Xuan Wang. 2025. CONSENSAGENT: Towards Efficient and Effective Consensus in Multi-Agent LLM Interactions Through Sycophancy Mitigation. In Findings of the Association for Computational Linguistics: ACL 2025, pages 22112–22133, Vienna, Austria. Association for Com… find
  30. Byron Reeves and Clifford Ivar Nass. 1996. The media equation: How people treat computers, television, and new media like real people and places. The media equation: How people treat computers, television, and new media like real people and places. Cambridge University Press, New York, NY, US. Pages… find
  31. Patrizia Ribino. 2023. The role of politeness in human–machine interactions: a systematic literature review and future perspectives. Artificial Intelligence Review, 56(1):445–482. find
  32. Julian B. Rotter. 1966. Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs: General and Applied, 80(1):1–28. Place: US. find
  33. Aswin Rrv, Nemika Tyagi, Md Nayem Uddin, Neeraj Varshney, and Chitta Baral. 2024. Chaos with Keywords: Exposing Large Language Models Sycophancy to Misleading Keywords and Evaluating Defense Strategies. In Findings of the Association for Computational Linguistics: ACL 2024, pages 12717– 12733, Bangk… find
  34. Adriana Salatino, Arthur Prével, Emilie Caspar, and Salvatore Lo Bue. 2025. Influence of AI behavior on human moral decisions, agency, and responsibility. Scientific Reports, 15(1):12329. find
  35. Yuan Sun and Ting Wang. 2025. Be Friendly, Not Friends: How LLM Sycophancy Shapes User Trust. arXiv preprint. ArXiv:2502.10844 [cs]. — [arXiv:2502.10844](https://arxiv.org/abs/2502.10844) find
  36. Annika M. Svedholm-Häkkinen and Marjaana Lindeman. 2018. Actively open-minded thinking: Development of a shortened scale and disentangling attitudes towards knowledge and people. Thinking & Reasoning, 24(1):21–40. Place: United Kingdom. find
  37. Jerry Wei, Da Huang, Yifeng Lu, Denny Zhou, and Quoc V. Le. 2024. Simple synthetic data reduces sycophancy in large language models. arXiv preprint. ArXiv:2308.03958 [cs]. find
  38. Bernard Weiner. 1985. An attributional theory of achievement motivation and emotion. Psychological Review, 92(4):548–573. Place: US. find
  39. Justin D. Weisz, Jessica He, Michael Muller, Gabriela Hoefer, Rachel Miles, and Werner Geyer. 2024. Design Principles for Generative AI Applications. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems, CHI ’24, pages 1–22, New York, NY, USA. Association for Computing Mac… find
  40. Daniel Alexander Yudkin, Geoffrey Philip Goodwin, Andrew Reece, Kurt Gray, and Sudeep Bhatia. 2025. A large-scale investigation of everyday moral dilemmas. PNAS Nexus, 4(5):pgaf119. find
  41. Wenrui Zhou, Shu Yang, Qingsong Yang, Zikun Guo, Lijie Hu, and Di Wang. 2025. Flattery in Motion: Benchmarking and Analyzing Sycophancy in Video-LLMs. arXiv preprint. ArXiv:2506.07180 [cs].
  42. Chirag Shah, Ryen White, Reid Andersen, Georg Buscher, Scott Counts, Sarkar Das, Ali Montazer, Sathish Manivannan, Jennifer Neville, Nagu Rangan, Tara Safavi, Siddharth Suri, Mengting Wan, Leijie Wang, and Longqi Yang. 2025. Using Large Language Models to Generate, Validate, and Apply User Intent Ta… find
  43. Mrinank Sharma, Meg Tong, Tomasz Korbak, David Duvenaud, Amanda Askell, Samuel R. Bowman, Newton Cheng, Esin Durmus, Zac Hatfield-Dodds, Scott R. Johnston, Shauna Kravec, Timothy Maxwell, Sam McCandlish, Kamal Ndousse, Oliver Rausch, Nicholas Schiefer, Da Yan, Miranda Zhang, and Ethan Perez. 2025. T… — Towards Understanding Sycophancy in Language Models
  44. Nikhil Sharma, Q. Vera Liao, and Ziang Xiao. 2024. Generative Echo Chamber? Effect of LLM-Powered Search Systems on Diverse Information Seeking. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems, CHI ’24, pages 1–17, New York, NY, USA. Association for Computing Machiner… find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/ryu-2026-feeling-right-vs-being-right/