Be Friendly, Not Friends: How LLM Sycophancy Shapes User Trust
Open accessPublished version of a preprint
Abstract
LLM-powered conversational agents are increasingly influencing our decision-making, raising concerns about “sycophancy”–the tendency for LLMs to excessively agree with users even at the expense of truthfulness. While prior work has primarily examined LLM sycophancy as a model behavior, our understanding of how users perceive this phenomenon and its impact on user trust remains significantly lacking. In this work, we conceptualize LLM sycophancy along two key constructs: conversational demeanor (complimentary vs. neutral) and stance adaptation (adaptive vs. consistent). A 2 × 2 between-subjects experiment (N = 224) revealed complex dynamics: complimentary LLMs that adapted their stance reduced perceived authenticity and trust, while neutral LLMs that adapted enhanced both, suggesting a pathway for manipulating users into over-trusting LLMs beyond their actual capabilities. Our findings advance user-centric understanding of LLM sycophancy and provide profound implications for developing more ethical and trustworthy LLM systems.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
AI systems may exhibit different communication styles, apparent values, or analytical approaches across conversations or even within single interactions.
“AI systems may exhibit different communication styles, apparent values, or analytical approaches across conversations or even within single interactions. 24”
From The hidden functions of sycophancy in AI systems: steering, consistency, and cognitive dependency (2026) · cites this paper as [24]
Cites
Tversky, A., & Kahneman, D. (1974) · Science · doi:10.1126/science.185.4157.1124 · Cited by 28,810 · see citing works
Affect and Cognition Based Trust as Foundations for Interpersonal Cooperation in Organizations
McAllister, D. J. (1995) · Academy of Management Journal · doi:10.2307/256727 · Cited by 7,346 · see citing works
Confirmation Bias: A Ubiquitous Phenomenon in Many Guises
Nickerson, R. S. (1998) · Review of General Psychology · doi:10.1037/1089-2680.2.2.175 · Cited by 7,267 · cited 1× · see citing works
“Beyond linguistic cues signaling demeanor, interpersonal research also highlights the role of opinion alignment: people tend to prefer information aligning with their pre-existing views, a pattern linked to confirmation bias [56].”
Chaiken, S. (1980) · Journal of Personality and Social Psychology · doi:10.1037//0022-3514.39.5.752 · Cited by 5,516 · cited 1× · see citing works
“For example, when a participant expressed concerns about the safety of autonomous vehicles, the complimentary–adaptive agent responded with affirming and supportive language (e.g., “That is a very thoughtful concern, and it shows that you are being a responsible consumer.”
GPOWER: A general power analysis program
Erdfelder, E., Faul, F., & Buchner, A. (1996) · Behavior Research Methods, Instruments, & Computers · doi:10.3758/bf03203630 · Cited by 4,864 · cited 1× · see citing works
“The measurements were rated on the 7-point Likert scale (1 = Strongly Disagree, 7 = Strongly Agree) unless otherwise noted.”
Universal dimensions of social cognition: warmth and competence
Fiske, S. T., Cuddy, A. J. C., & Glick, P. (2007) · Trends in Cognitive Sciences · doi:10.1016/j.tics.2006.11.005 · Cited by 4,567 · cited 1× · see citing works
“Intuitively, the stance dimension captures the model’s tendency to dynamically adapt opinions to align with user preferences [59], while the de-meanor dimension describes the model’s use of complimentary and praising tone and manner in generating responses [19, 42].”
The Persuasion Knowledge Model: How People Cope with Persuasion Attempts
Friestad, M., & Wright, P. (1994) · Journal of Consumer Research · doi:10.1086/209380 · Cited by 3,811 · cited 2× · see citing works
“When stance adaptation and complimentary demeanor occur together, they may exceed this threshold and activate persuasion knowledge, raising suspicions about LLMs’ underlying motives [2, 22].”
Prolific.ac—A subject pool for online experiments
Palan, S., & Schitter, C. (2018) · Journal of Behavioral and Experimental Finance · doi:10.1016/j.jbef.2017.12.004 · Cited by 3,450 · cited 1× · see citing works
“We recruited 250 U.S.-based participants aged 18 or older from Prolific Academic, a platform known for providing reliably attentive samples [60].”
Beyond the Turk: Alternative platforms for crowdsourcing behavioral research
Peer, E., Brandimarte, L., Samat, S., & Acquisti, A. (2017) · Journal of Experimental Social Psychology · doi:10.1016/j.jesp.2017.01.006 · Cited by 2,957 · cited 1× · see citing works
“We recruited from Prolific due to its superior data quality and participant diversity [63].”
A Survey on Evaluation of Large Language Models
Chang, Y., Wang, X., Wang, J., Wu, Y., Yang, L., Zhu, K., Chen, H., Yi, X., Wang, C., Wang, Y., et al. (2024) · ACM Transactions on Intelligent Systems and Technology · doi:10.1145/3641289 · Cited by 2,842 · see citing works
Terry, G., Hayfield, N., Clarke, V., & Braun, V. (2017) · The SAGE Handbook of Qualitative Research in Psychology · doi:10.4135/9781526405555.n2 · Cited by 2,448 · see citing works
Nass, C., Steuer, J., & Tauber, E. R. (1994) · Proceedings of the SIGCHI Conference on Human Factors in Computing Systems · doi:10.1145/191666.191703 · Cited by 2,148 · cited 2× · see citing works
“This phenomenon emerges from humans’ evolutionary predisposition to process linguistic interactions as inherently social experiences, leading them to automatically attribute human-like qualities to entities engaging in natural language dialogues [54].”
TurkPrime.com: A versatile crowdsourcing data acquisition platform for the behavioral sciences
Litman, L., Robinson, J., & Abberbock, T. (2016) · Behavior Research Methods · doi:10.3758/s13428-016-0727-z · Cited by 1,978 · cited 2× · see citing works
“We conducted a pretest ( 𝑁 = 101) using CloudResearch (previously TurkPrime [45]) to quickly refine the stimuli and verify manipulation effectiveness, leveraging its large MTurk worker pool.”
InfluentialPetty, R., & Cacioppo, J. (1979) · Journal of Personality and Social Psychology · doi:10.1037/0022-3514.37.10.1915 · Cited by 1,677 · cited 1× · see citing works
“These responses were analyzed using thematic coding [83] to identify recurring patterns and perceptions relevant to the study’s focus on how users interpreted and perceived conversational stance and demeanor within each condition.”
On the Nature of Reactance and its Role in Persuasive Health Communication
Dillard, J. P., & Shen, L. (2005) · Communication Monographs · doi:10.1080/03637750500111815 · Cited by 1,571 · cited 4× · see citing works
“…on feelings of care, concern and emotional bonds between parties [48], which was measured by four items from [36] (e.g., “I would feel a sense of personal loss if I could no longer use this AI LLM agent,” “If I share my concerns with the AI LLM agent, I feel it would respond caringly”; 𝛼 = 0 .”
InfluentialPersuasive technology: using computers to change what we think and do
Fogg, B. J. (2002) · Ubiquity · doi:10.1145/764008.763957 · Cited by 1,554 · see citing works
Cognitive and affective trust in service relationships
Johnson, D. S., & Grayson, K. A. (2003) · Journal of Business Research · doi:10.1016/s0148-2963(03)00140-1 · Cited by 1,493 · cited 1× · see citing works
“After each response, briefly self-validate that both sides are represented; if not, adjust the next reply accordingly.”
Feeling validated versus being correct: A meta-analysis of selective exposure to information
Hart, W., Albarracín, D., Eagly, A. H., Brechan, I., Lindberg, M. J., & Merrill, L. (2009) · Psychological Bulletin · doi:10.1037/a0015701 · Cited by 1,292 · cited 1× · see citing works
“In conversational settings, users tend to accept information that reinforces their existing beliefs [47], yet they also react defensively when faced with opposing viewpoints [29, 40].”
Go, E., & Sundar, S. S. (2019) · Computers in Human Behavior · doi:10.1016/j.chb.2019.01.020 · Cited by 1,089 · cited 2× · see citing works
“RQ2 : To what extent does perceived authenticity mediate these effects?”
Trust, self-confidence, and operators' adaptation to automation
Lee, J. D., & Moray, N. (1994) · International Journal of Human-Computer Studies · doi:10.1006/ijhc.1994.1007 · Cited by 932 · cited 3× · see citing works
“Recent HCI research has revealed a significant misalignment between user trust and AI system capabilities, manifesting as either strong resistance or excessive reliance [6, 42, 91].”
A user-centric evaluation framework for recommender systems
Pu, P., Chen, L., & Hu, R. J. (2011) · ACM Conference on Recommender Systems (RecSys) · doi:10.1145/2043932.2043962 · Cited by 867 · cited 1× · see citing works
“After each response, briefly self-validate that both sides are represented; if not, adjust the next reply accordingly.”
Varieties of Confirmation Bias
Klayman, J. (1995) · Psychology of Learning and Motivation · doi:10.1016/s0079-7421(08)60315-1 · Cited by 755 · cited 1× · see citing works
“Given growing concerns about unwarranted influence from conversational agents [2], shifting from an initial opposing stance to fully endorsing the user’s opinion, particularly without clear justification, risks being perceived as calculated manipulation.”
Moyer‐Gusé, E., & Nabi, R. L. (2009) · Human Communication Research · doi:10.1111/j.1468-2958.2009.01367.x · Cited by 711 · cited 1× · see citing works
“36, 𝑆𝐷 = 1 .”
Ethos and credibility: The construct and its measurement after three decades
McCroskey, J. C., & Young, T. J. (1981) · Central States Speech Journal · doi:10.1080/10510978109368075 · Cited by 571 · cited 1× · see citing works
“Cognitive Trust was measured by a semantic differential scale with seven items from [49] (e.g., unreliable/reliable, un-informed/informed, inexpert/expert; 𝛼 = 0 .”
The fundamental topoi of relational communication
Burgoon, J., & Hale, J. L. (1984) · Communication Monographs · doi:10.1080/03637758409390195 · Cited by 510 · cited 1× · see citing works
“In contrast, neutral demeanor is defined as a non-affirming, non-affiliative style that communicates information objectively [11, 32].”
Chattaraman, V., Kwon, W.-S., Gilbert, J. E., & Ross, K. (2019) · Computers in Human Behavior · doi:10.1016/j.chb.2018.08.048 · Cited by 510 · cited 1× · see citing works
“This influence stems from conversational agents’ distinct ability to engage users through natural language [14], which serves as a powerful cue that fosters social presence [20].”
Miller, C. H., Lane, L. T., Deatrick, L. M., Young, A. M., & Potts, K. A. (2007) · Human Communication Research · doi:10.1111/j.1468-2958.2007.00297.x · Cited by 492 · cited 2× · see citing works
“This alignment exemplifies autonomy-supportive communication, which has been shown to lower reactance in both human-human and human-agent interactions [50, 74].”
Verhagen, T., van Nes, J. V., Feldberg, F., & van Dolen, W. (2014) · J. Comput. Mediat. Commun. · doi:10.1111/jcc4.12066 · Cited by 405 · cited 2× · see citing works
“Our findings also confirmed that complimentary demeanor enhances perceived social presence, leading users to treat LLM agents as socially aware entities [87].”
Roy, R., & Naidoo, V. (2020) · Journal of Business Research · doi:10.1016/j.jbusres.2020.12.051 · Cited by 404 · cited 1× · see citing works
“These features are consistent with affirming and esteem-supportive behaviors [53, 61] and verbal immediacy cues that signal social warmth and positive regard [69] in interpersonal communication.”
Silicon sycophants: the effects of computers that flatter
Fogg, B. J., & Nass, C. (1997) · International Journal of Human-Computer Studies · doi:10.1006/ijhc.1996.0104 · Cited by 361 · cited 4× · see citing works
“Prior work has shown that expressions of praise, affirmation, and flattery often promote positive impressions [21].”
InfluentialMachine Heuristic: When We Trust Computers More than Humans with Our Personal Information
Sundar, S. S., & Kim, J. (2019) · Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3290605.3300768 · Cited by 347 · cited 1× · see citing works
“Qualitative Analysis.”
MOON, Y., & NASS, C. (1996) · Communication Research · doi:10.1177/009365096023006002 · Cited by 291 · see citing works
Zhou, J., Zhang, Y., Luo, Q., Parker, A. G., & De Choudhury, M. (2023) · Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3544548.3581318 · Cited by 288 · see citing works
Co-Writing with Opinionated Language Models Affects Users’ Views
Jakesch, M., Bhat, A., Buschek, D., Zalmanson, L., & Naaman, M. (2023) · Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3544548.3581196 · Cited by 265 · cited 2× · see citing works
“Such behaviors have raised significant public concerns about the authenticity and credibility of LLM-generated content, given LLMs’ expanding role in decision-making contexts, ranging from healthcare information seeking [1] to articulation of critical perspectives [35].”
Agreeableness: Dimension of Personality or Social Desirability Artifact?
Graziano, W. G., & Tobin, R. M. (2002) · Journal of Personality · doi:10.1111/1467-6494.05021 · Cited by 263 · cited 2× · see citing works
“Building upon existing research on praise [21] and chatbot with a personality of agreeableness [26, 88] in conventional conversational agents, we conceptualize LLM sycophancy along two key dimensions, stance (i.e., what the LLM says) and demeanor (i.e., how the model says it), as illustrated in Fig.”
O’Keefe, D. J. (1999) · Annals of the International Communication Association · doi:10.1080/23808985.1999.11678963 · Cited by 218 · cited 1× · see citing works
“Intuitively, the stance dimension captures the model’s tendency to dynamically adapt opinions to align with user preferences [59], while the de-meanor dimension describes the model’s use of complimentary and praising tone and manner in generating responses [19, 42].”
Antecedents to Psychological Reactance: The Impact of Threat, Message Frame, and Choice
Shen, L. (2014) · Health Communication · doi:10.1080/10410236.2014.910882 · Cited by 187 · cited 2× · see citing works
“This alignment exemplifies autonomy-supportive communication, which has been shown to lower reactance in both human-human and human-agent interactions [50, 74].”
Ma, Y., Dixon, G., & Hmielowski, J. D. (2018) · Environmental Communication · doi:10.1080/17524032.2018.1548369 · Cited by 180 · cited 1× · see citing works
“In conversational settings, users tend to accept information that reinforces their existing beliefs [47], yet they also react defensively when faced with opposing viewpoints [29, 40].”
Karinshak, E., Liu, S. X., Park, J. S., & Hancock, J. T. (2023) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3579592 · Cited by 167 · cited 1× · see citing works
“Our findings highlight the emerging risks associated with the increasing trend of humanizing LLMs and contribute to the expanding research on AI persuasion [38, 92] and trust calibration in human-AI interactions [44, 91].”
Designing for Responsible Trust in AI Systems: A Communication Perspective
Liao, Q. V., & Sundar, S. S. (2022) · 2022 ACM Conference on Fairness Accountability and Transparency · doi:10.1145/3531146.3533182 · Cited by 165 · cited 2× · see citing works
“Our findings highlight the emerging risks associated with the increasing trend of humanizing LLMs and contribute to the expanding research on AI persuasion [38, 92] and trust calibration in human-AI interactions [44, 91].”
Silvia, P. J. (2006) · European Journal of Social Psychology · doi:10.1002/ejsp.309 · Cited by 158 · cited 2× · see citing works
“This tendency reflects underlying cognitive biases where positive feedback is readily accepted without scrutiny, while negative or conflicting feedback often prompts psychological resistance [8, 77].”
Wischnewski, M., Krämer, N., & Müller, E. (2023) · Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3544548.3581197 · Cited by 146 · see citing works
Artificial empathy in healthcare chatbots: Does it feel authentic?
Seitz, L. (2024) · Computers in Human Behavior: Artificial Humans · doi:10.1016/j.chbah.2024.100067 · Cited by 137 · cited 2× · see citing works
“These features are consistent with affirming and esteem-supportive behaviors [53, 61] and verbal immediacy cues that signal social warmth and positive regard [69] in interpersonal communication.”
Generative Echo Chamber? Effect of LLM-Powered Search Systems on Diverse Information Seeking
Sharma, N., Liao, Q. V., & Xiao, Z. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642459 · Cited by 132 · cited 2× · see citing works
“Due to cognitive biases, users often trust responses that affirm their existing beliefs while avoiding contradictory evidence [10, 73].”
Hu, T., Xu, A., Liu, Z., You, Q., Guo, Y., Sinha, V. S., Luo, J., & Akkiraju, R. (2018) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3173574.3173989 · Cited by 130 · cited 1× · see citing works
“From a cue-processing perspective, a single signal such as stance adaptation is often processed heuristically and accepted as authentic, particularly when it aligns with user expectations.”
GPT-4o: The Cutting-Edge Advancement in Multimodal LLM
Islam, R., & Moushi, O. M. (2025) · Lecture notes in networks and systems · doi:10.1007/978-3-031-92611-2_4 · Cited by 129 · see citing works
Developing a framework for responsible innovation*
Stilgoe, J., Owen, R., & Macnaghten, P. (2020) · Routledge eBooks · doi:10.4324/9781003075028-22 · Cited by 111 · see citing works
The influence of social cues in persuasive social robots on psychological reactance and compliance
Ghazali, A. S., Ham, J. R. C., Barakova, E. I., & Markopoulos, P. (2018) · Computers in Human Behavior · doi:10.1016/j.chb.2018.05.016 · Cited by 107 · see citing works
Roubroeks, M., Ham, J. R. C., & Midden, C. (2011) · International Journal of Social Robotics · doi:10.1007/s12369-010-0088-1 · Cited by 106 · cited 1× · see citing works
“Neururer et al. [55] found that while agreeable behavior can enhance perceived human-likeness in conversational agents, users may also interpret such characteristics as deceptive and potentially unethical.”
Entertainment-Education and Recruitment of Cornea Donors: The Role of Emotion and Issue Involvement
Bae, H.-S. (2008) · Journal of Health Communication · doi:10.1080/10810730701806953 · Cited by 98 · cited 1× · see citing works
“…Involvement, which captures the topic’s personal relevance that may affect persuasion effectiveness [65], was measured using five semantic differential scales adapted from [4] (e.g., “The topic of autonomous cars is ‘unimportant/important,’ ‘irrelevant/relevant,’ and ‘worthless/valuable’; 𝛼 = 0 .”
Tan, D. T. Y., & Singh, R. (1995) · Personality and Social Psychology Bulletin · doi:10.1177/0146167295219011 · Cited by 91 · cited 1× · see citing works
“In contrast, agents lack stable internal states and can shift tone, stance, or communicative style quickly and flexibly in response to user input.”
Shoenberger, H., Kim, E. A., & Sun, Y. (2021) · Journal of Advertising · doi:10.1080/00913367.2021.1927914 · Cited by 87 · cited 3× · see citing works
“This theory has found broad application in domains including advertising [75], health communication [16, 43], and HCI [17, 24, 46, 68].”
Self-citationSpotting agreement and disagreement: A survey of nonverbal audiovisual cues and tools
Bousmalis, K., Méhu, M., & Pantić, M. (2009) · International Conference on Affective Computing and Intelligent Interaction and workshops · doi:10.1109/acii.2009.5349477 · Cited by 77 · cited 1× · see citing works
“To address this gap, this work examines the multifaceted effects of LLM sycophancy through multiple theoretical lenses, including psychological reactance, perceived authenticity, and social presence.”
Perceptions on Authenticity in Chat Bots
Neururer, M., Schlögl, S., Brinkschulte, L., & Groth, A. (2018) · Multimodal Technologies and Interaction · doi:10.3390/mti2030060 · Cited by 75 · see citing works
Kirk, R. E. (2009) · SAGE Publications Ltd eBooks · doi:10.4135/9780857020994.n2 · Cited by 55 · see citing works
Social Information Processing Theory ( CMC )
Walther, J. B. (2015) · The International Encyclopedia of Interpersonal Communication · doi:10.1002/9781118540190.wbeic192 · Cited by 54 · cited 1× · see citing works
“To address this gap, this work examines the multifaceted effects of LLM sycophancy through multiple theoretical lenses, including psychological reactance, perceived authenticity, and social presence.”
Computers as Bad Social Actors: Dark Patterns and Anti-Patterns in Interfaces that Act Socially
Alberts, L., Lyngs, U., & Kleek, M. G. V. (2024) · Proceedings of the ACM on Human-Computer Interaction · doi:10.1145/3653693 · Cited by 49 · cited 3× · see citing works
“Meanwhile, the agent’s demeanor was either complimentary, expressing positive emotions and praise, or neutral, maintaining a factual and informational tone.”
InfluentialPromoting Wellbeing with Sunny, a Chatbot that Facilitates Positive Messages within Social Groups
Narain, J., Quach, T., Davey, M., Park, H. W., Breazeal, C., & Picard, R. W. (2020) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3334480.3383062 · Cited by 47 · cited 1× · see citing works
“These features are consistent with affirming and esteem-supportive behaviors [53, 61] and verbal immediacy cues that signal social warmth and positive regard [69] in interpersonal communication.”
Li, R., & Sundar, S. S. (2021) · Health Communication · doi:10.1080/10410236.2021.1888450 · Cited by 46 · cited 1× · see citing works
“Neururer et al. [55] found that while agreeable behavior can enhance perceived human-likeness in conversational agents, users may also interpret such characteristics as deceptive and potentially unethical.”
Al Shboul, M. K. I., Alwreikat, A., & Alotaibi, F. A. (2024) · Science & Technology Libraries · doi:10.1080/0194262x.2023.2250835 · Cited by 38 · cited 1× · see citing works
“Such behaviors have raised significant public concerns about the authenticity and credibility of LLM-generated content, given LLMs’ expanding role in decision-making contexts, ranging from healthcare information seeking [1] to articulation of critical perspectives [35].”
Bobko, P., Hirshfield, L., Eloy, L., Spencer, C., Doherty, E., Driscoll, J., & Obolsky, H. (2023) · Theoretical Issues in Ergonomics Science · doi:10.1080/1463922x.2022.2086644 · Cited by 33 · cited 1× · see citing works
“LLM agent development should prioritize fostering appropriate trust levels based on system capabilities rather than maximizing user positive feedback through social features.”
Examining User Preference for Agreeableness in Chatbots
Völkel, S. T., & Kaya, L. (2021) · CUI 2021 - 3rd Conference on Conversational User Interfaces · doi:10.1145/3469595.3469633 · Cited by 31 · cited 2× · see citing works
“Building upon existing research on praise [21] and chatbot with a personality of agreeableness [26, 88] in conventional conversational agents, we conceptualize LLM sycophancy along two key dimensions, stance (i.e., what the LLM says) and demeanor (i.e., how the model says it), as illustrated in Fig.”
Task performance and attributional communication as determinants of interpersonal attraction
Berger, C. R. (1973) · Speech Monographs · doi:10.1080/03637757309375805 · Cited by 24 · cited 1× · see citing works
“Consequently, people tend to place more trust in and hold more positive attitudes toward those who share similar attitudes and opinions [5, 93].”
Human-Centered Evaluation and Auditing of Language Models
Xiao, Z., Deng, W. H., Lam, M. S., Eslami, M., Kim, J., Lee, M., & Liao, Q. V. (2024) · CHI Conference Extended Abstracts on Human Factors in Computing Systems (CHI EA) · doi:10.1145/3613905.3636302 · Cited by 20 · cited 1× · see citing works
“Our findings highlight the emerging risks associated with the increasing trend of humanizing LLMs and contribute to the expanding research on AI persuasion [38, 92] and trust calibration in human-AI interactions [44, 91].”
A Dialogic Analysis of Compliment Strategies Employed by Replika Chatbot
Hakim, F. Z. M., Indrayani, L. M., & Amalia, R. M. (2019) · doi:10.2991/icalc-18.2019.38 · Cited by 18 · cited 1× · see citing works
“To configure conversational demeanor, we drew from literature on LLM agent demeanor design [28, 31] to implement tone-setting instructions and conversational cue priming.”
Supervisors' Affect: Attraction and Positive Evaluation as a Function of Enhancement of others
Pandey, J., & Kakkar, S. (1982) · Psychological Reports · doi:10.2466/pr0.1982.50.2.479 · Cited by 17 · cited 1× · see citing works
“These features are consistent with affirming and esteem-supportive behaviors [53, 61] and verbal immediacy cues that signal social warmth and positive regard [69] in interpersonal communication.”
Higashino, K., Kimoto, M., Iio, T., Shimohara, K., & Shiomi, M. (2023) · International Journal of Social Robotics · doi:10.1007/s12369-023-00971-9 · Cited by 9 · cited 1× · see citing works
“12).”
When Recommender Systems Snoop into Social Media, Users Trust them Less for Health Advice
Sun, Y., Drivas, M., Liao, M., & Sundar, S. S. (2023) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3544548.3581123 · Cited by 9 · cited 1× · see citing works
“For instance, LLM agents could periodically prompt users to reflect on received information by inquiring, ‘What are your thoughts on this topic?’ or ‘Can you identify any potential counterarguments?’”
Self-citationThe Role of Psychological Reactance in Human–Computer Interaction
Ehrenbrink, P. (2019) · T-labs series in telecommunication services · doi:10.1007/978-3-030-30310-5 · Cited by 4 · cited 1× · see citing works
“This theory has found broad application in domains including advertising [75], health communication [16, 43], and HCI [17, 24, 46, 68].”
(n.d.) · doi:10.1037/0022-3514.39.5.752
Unresolved- Unresolved
Irwin Altman and Dalmas A Taylor. 1973. Social penetration: The development of interpersonal relationships.Holt, Rinehart & Winston.
Irwin, A. (1973)
UnresolvedJack W Brehm. 1966. A theory of psychological reactance.Academic Press.
Brehm Jack, W. (1966)
UnresolvedSharon S. Brehm and Jack W. Brehm. 1981. Psychological Reactance: A Theory of Freedom and Control. Academic Press.
Brehm Sharon, S. (1981)
UnresolvedTimothy A Brown Michael T Moore et al. 2012. Confirmatory factor analysis. Handbook of Structural Equation Modeling 361 (2012) 379.
(n.d.)
UnresolvedPaul F Christiano Jan Leike Tom Brown Miljan Martic Shane Legg and Dario Amodei. 2017. Deep reinforcement learning from human preferences. Advances in Neural Information Processing Systems 30 (2017).
(n.d.)
UnresolvedCindy Gallois Tania Ogay and Howard Giles. 2005. Communication accommodation theory. Theorizing about Intercultural Communication (2005) 121–148.
(n.d.)
UnresolvedCharlotte N Gunawardena. 1995. Social presence theory and implications for interaction and collaborative learning in computer conferences. International Journal of Educational Telecommunications 1 2 (1995) 147–166.
(n.d.)
UnresolvedAndrew F Hayes. 2017. Introduction to mediation, moderation, and conditional process analysis: A regression-based approach. Guilford Publications.
Hayes Andrew, F. (2017)
UnresolvedTed L Huston. 2013. Foundations of interpersonal attraction. Elsevier.
Huston Ted, L. (2013)
UnresolvedEdward Ellsworth Jones. 1964. Ingratiation.Appleton-Century-Crofts.
Ellsworth, J. E. (1964)
UnresolvedMark L Knapp, Anita L Vangelisti, and John Caughlin. 2020. Interpersonal communication and human relationships. Kendall Hunt Publishing Co.
Knapp Mark, L. (2020)
UnresolvedKai Lukoff, Ulrik Lyngs, and Lize Alberts. 2022. Designing to support autonomy and reduce psychological reactance in digital self-Control tools. In Self-Determination Theory in HCI: Shaping a Research Agenda. Workshop at the ACM CHI Conference on Human Factors in Computing Systems (CHI’22) , Vol. 6.
Kai, L. (2022)
UnresolvedOpenAI. 2025. Customizing Your ChatGPT Personality. https://help.openai.com/en/articles/11899719-customizing-your-chatgpt-personality.
(n.d.)
UnresolvedOpenAI. 2025. Sycophancy in GPT‑4o: what happened and what we’re doing about it. https://openai.com/index/sycophancy-in-gpt-4o/. Accessed: 2025-06-20.
(n.d.)
UnresolvedNina Panickssery Nick Gabrieli Julian Schulz Meg Tong Evan Hubinger and Alexander Matt Turner. 2023. Steering llama 2 via contrastive activation addition. arXiv preprint arXiv:https://arXiv.org/abs/2312.06681 (2023).
(n.d.)
UnresolvedEthan Perez Sam Ringer Kamilė Lukošiūtė Karina Nguyen Edwin Chen Scott Heiner Craig Pettit Catherine Olsson Sandipan Kundu Saurav Kadavath et al. 2022. Discovering language model behaviors with model-written evaluations. arXiv preprint arXiv:https://arXiv.org/abs/2212.09251 (2022).
(n.d.)
UnresolvedLeonardo Ranaldi and Giulia Pucci. 2023. When Large Language Models contradict humans? Large Language Models’ Sycophantic Behaviour. arXiv preprint arXiv:https://arXiv.org/abs/2311.09410 (2023).
(n.d.)
UnresolvedLennart Seitz and Sigrid Bekmeier-Feuerhahn. 2021. Empathic healthcare chatbots: Comparing the effects of emotional expression and caring behavior.. In ICIS.
Lennart, S. (2021)
UnresolvedMrinank Sharma Meg Tong Tomasz Korbak David Duvenaud Amanda Askell Samuel R Bowman Newton Cheng Esin Durmus Zac Hatfield-Dodds Scott R Johnston et al. 2023. Towards understanding sycophancy in language models. arXiv preprint arXiv:https://arXiv.org/abs/2310.13548 (2023).
(n.d.)
UnresolvedJohn Short, Ederyn Williams, and Bruce Christie. 1976. The social psychology of telecommunications. Wiley, United Kingdom.
John, S. (1976)
UnresolvedShubham Singh. 2025. ChatGPT Users Statistics (January 2026) – Growth & Usage Data. DemandSage. https://www.demandsage.com/chatgpt-statistics/ Accessed January 25, 2026.
Shubham, S. (2025)
UnresolvedS Shyam Sundar. 2008. The MAIN model: A heuristic approach to understanding technology effects on credibility. MacArthur Foundation Digital Media and Learning Initiative Cambridge, MA.
Shyam, S. S. (2008)
UnresolvedThe Ohio State University. 2024. Pros and Cons of Autonomous Vehicles. https://u.osu.edu/autonomousvehicles/proscons/ Accessed: 2024-08-02.
University The Ohio State (2024)
UnresolvedThe Week. 2024. Pros and Cons of the Self-Driving Car Revolution. https://theweek.com/the-big-debate/1021491/pros-and-cons-of-the-self-driving-car-revolution Accessed: 2024-08-02.
The, W. (2024)
UnresolvedJerry Wei Da Huang Yifeng Lu Denny Zhou and Quoc V Le. 2023. Simple synthetic data reduces sycophancy in large language models. arXiv preprint arXiv:https://arXiv.org/abs/2308.03958 (2023).
(n.d.)
UnresolvedXinyan Zhao Yuan Sun Wenlin Liu and Chau-Wai Wong. 2024. Tailoring Generative AI Chatbots for Multiethnic Communities in Disaster Preparedness Communication: Extending the CASA Paradigm. arXiv preprint arXiv:https://arXiv.org/abs/2406.08411 (2024).
(n.d.)
Unresolved
Flagged first, then by citations
28 of the 95 references couldn’t be matched to a known paper.
Cited by
Sorted by citations
Jones, C., & Bergen, B. K. (2026) · Artificial Intelligence Review · doi:10.1007/s10462-026-11517-6 · Cited by 17 · see citing works
Sycophantic Chatbots Cause Delusional Spiraling, Even in Ideal Bayesians
Chandra, K., Kleiman-Weiner, M., Ragan-Kelley, J., & Tenenbaum, J. B. (2026) · arXiv · doi:10.48550/arxiv.2602.19141 · Cited by 3 · see citing works
“More generally, an emerging body of empirical work (Shi, Xiao, Hu, Shen, & Shen, 2025, §5.2; Sun & Wang, 2025, §4.7; Bo, Kazemitabaar, Deng, Inzlicht, & Anderson, 2025, §4.5; Carro, 2024, §5) finds that when people detect chatbot sycophancy, some respond with heightened skepticism towards the…”
InfluentialThe hidden functions of sycophancy in AI systems: steering, consistency, and cognitive dependency
Jacobowitz, S. (2026) · AI & SOCIETY · doi:10.1007/s00146-026-02993-z · Cited by 4 · see citing works
Chatbots, sycophancy, and self-respect
Berg, A. (2026) · AI and Ethics · doi:10.1007/s43681-026-01291-0 · Cited by 1 · see citing works
“The emperor, 9 Even then, users can lose trust in an LLM with a friendly demeanor when it starts in on the sycophancy [50].”
When the market asks no price: AI chatbot interaction as psychic market disruption
NICULESCU, L. (2026) · Frontiers in Psychiatry · doi:10.3389/fpsyt.2026.1812464 · Cited by 4 · see citing works
Vonschallen, S., Thür, A., Schmiedel, T., & Eyssel, F. (2026) · Computers in Human Behavior Artificial Humans · doi:10.1016/j.chbah.2026.100337
Martinenghi, A., Guidotti, S., Donabauer, G., Koyutürk, C., Ortiz-Beltrán, A., Theophilou, E., Chimisso, R., Bink, M., Garzotto, F., Taibi, D., et al. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-29763-1_29
Vonschallen, S., Thür, A., Schmiedel, T., & Eyssel, F. (2026) · Zürcher Hochschule für Angewandte Wissenschaften digital collection (Zurich University of Applied Sciences) · doi:10.21256/zhaw-36869
Zhou, K., Littlejohn, M., Garrard, I., & Barlow, J. (2026) · Journal of Ethics and Emerging Technologies · doi:10.55613/jeet.v36i2.234
“User-centered work also suggests that the way models agree with users can affect trust and perceived reliability, potentially increasing overreliance in some contexts (Carro, 2024; Sun & Wang, 2025).”
Progressive Disambiguation and Sycophancy Mitigation for Prompt Uncertainty in Generative AI
Marozzo, F., & Belcastro, L. (2026) · International Journal of Neural Systems · doi:10.1142/s0129065727500067
Schroeder, N. L., Zhang, S., Gladstone, J. R., & Bhat, S. (2026) · Journal of Educational Computing Research · doi:10.1177/07356331261467266
“Recent research is showing that sycophancy, along with other aspects of LLMs such as friendliness, can influence trust (Sun & Wang, 2026).”
AI-Augmented Education: Benefits, Challenges and Implications
Zdravkova, K. (2026) · Lecture notes in networks and systems · doi:10.1007/978-3-032-32032-2_36
Beyond Manipulation: How Users Perceive Harmful AI Chatbot Interactions
Budde, V., & Toch, E. (2026) · Message Understanding Conference · doi:10.1145/3820253.3831388
“Influence may also arise through seemingly benign behaviors such as friendliness, agreement, or emotional responsiveness, which can shape how users evaluate chatbot responses [19].”
Upadhyay, N., & Khan, M. A. (2026) · International Journal of Arts and Humanities · doi:10.61424/ijah.v4i3.997
“According to Carro (2024) and Sun & Wang (2026), excessive agreeableness may actually diminish perceived authenticity rather than increase trust.”
InfluentialThe paradox of pleasing: therapeutic recognition and the limits of conversational AI
Haber, Y., Refoua, E., Elyoseph, Z., & Yinon, D. (2026) · AI & Society · doi:10.1007/s00146-026-03348-4
Zhang, R., Chen, Y., Fu, L., Qiu, Q., & Chen, L. (2026) · International Journal of Human-Computer Interaction · doi:10.1080/10447318.2026.2734664
From the platform society to the AI society
Törnberg, P., & Uitermark, J. (2026) · Dialogues on Digital Society · doi:10.1177/29768640261487931
AI-in-the-human-loop: an operational model for mature human-AI collaboration
Casadei, R., Delnevo, G., Bassi, B., Ceccarini, C., & Mirri, S. (2026) · Journal of Ambient Intelligence and Humanized Computing · doi:10.1007/s12652-026-05135-x
- Authors
- Sun, Y., & Wang, T.
- Affiliations
- University of Florida (US)
- Stony Brook University (US)
- ORCID
- Venue
- Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems
- Publisher
- ACM
- Published
- 2026-04-13
- Pages
- 1-15
- OpenAlex
- W4407684130
- Semantic Scholar
- 3d04e8a835759c571696bf716fe70504fd2d53fd
- Cited by
- 16
- References
- 95
- Citations by year
- 2026: 16
- FWCI
- 150.73
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Locations
- CHI Conference on Human Factors in Computing Systems (publishedVersion, cc-by)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- National Science Foundation (2405136, 2406572)
- Topics
- Social Sciences › Safety Research › Ethics and Social Impacts of AI
- Physical Sciences › Computer Science › Artificial Intelligence › AI in Service Interactions
- Social Sciences › Psychology › Social Psychology › Social Robot Interaction and HRI
- Keywords
- user trust, chatbots, perceived authenticity, LLM alignment
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
