Pith. sign in

REVIEW 4 major objections 6 minor 74 references

Can LLMs and humans be friends? Uncovering factors affecting human-AI intimacy formation

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Gradually deepening what a person and an LLM share in conversation increases felt social closeness, even when the chatbot's personality does not match the user's.

desk verdict The paper's central causal claim—that gradual self-disclosure drives intimacy in human-LLM chats—is not actually tested, because every participant got the same fixed question sequence with no control condition. read the letter →

arxiv 2505.24658 v1 pith:QJZH4OXU submitted 2025-05-30 cs.HC

Yeseon Hong , Junhyuk Choi , Minju Kim , Bugeun Kim This is my paper
classification cs.HC
keywords Human-AIinteractionIntimacySelf-disclosureSelf-criticismLargelanguagemodelChatbotGradualConversationalnaturalness
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Can a person feel friendship with a large language model? This paper argues yes, and claims that the clearest lever is how disclosure unfolds over time, not who the partner appears to be. In two experiments using a staged conversation protocol, 82 young adults answered questions that moved from light topics to personal ones, and perceived social intimacy rose steadily across the three stages. Making the LLM's responses more natural through a self-criticism step raised early intimacy and improved first impressions, while personality matching had no consistent effect, and revealing that the partner was an AI did not change results. The paper concludes that relationship-building with LLMs is driven by conversational dynamics, and offers design guidelines for gradual disclosure, colloquial language, and calibrated empathy.

What carries the argument

The carrying mechanism is a three-stage conversation protocol adapted from the classic fast-friends closeness procedure [2]: 33 questions divided into three sets of 11, with each set probing deeper self-disclosure, and intimacy measured after each stage with the Subjective Closeness Index [5] and Interpersonal Judgement Scale [10]. This protocol converts 'gradualness' from a description into a manipulated variable. Study 2 adds a second mechanism, a self-criticism loop in which the LLM critiques its own draft responses against colloquial-style and sincere-empathy criteria and rewrites them; this operationalizes 'naturalness' and is what produces the higher early-stage intimacy scores.

What would settle it

Run the same staged protocol with a control group that spends the same total time conversing but receives the 33 questions in a flat or reversed order of personal depth; if intimacy rises just as much in the control group, the light-to-personal trajectory itself is not what builds closeness.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that gradual self-disclosure—partnered dialogue that moves from surface topics to increasingly personal ones—significantly raises perceived social intimacy between a human user and an LLM, and does so independently of persona reciprocity. In Study 1, Subjective Closeness Index scores rose from a mean of 2.266 at the first stage to 3.516 at the third (Friedman test p < 0.001, all post-hoc pairwise comparisons significant). Participants in the 'unfit' persona condition actually reported higher closeness than those in the 'fit' condition, and in Study 2 persona similarity had no significant main effect at all. Whether participants believed they were talking to a person or to an AI made no significant difference, and self-criticism made responses feel more natural and boosted early-stage intimacy, though overly enthusiastic empathy sometimes broke immersion.

Load-bearing premise

The headline claim rests on the assumption that the steady rise in intimacy scores across the three stages is caused by the gradual deepening of the questions, rather than by the extra conversation time, the heavier emotional content of later questions, or simple familiarity with the partner.

Editorial extensions

If this is right

  • Chatbot designers should script escalating self-disclosure over the conversation rather than rely on a static persona.
  • Users can feel closeness to an AI even when they know it is an AI, so relationship-building does not require deception.
  • Persona-matching is not a reliable route to intimacy, and a too-similar persona combined with over-agreeable empathy can feel uncanny and reduce closeness.
  • Response naturalness shapes the first impression: colloquial style and sincere, context-limited empathy raise early-stage intimacy; excessive empathy must be calibrated.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A direct test of the paper's mechanism would compare three equally long conversations that end at the same personal depth but ramp at different speeds, separating the trajectory of disclosure from its final level.
  • The Study 1 result that a mismatched persona scored higher than a matched one suggests an upper bound on similarity: some divergence may keep the exchange feeling like a real encounter rather than a mirror; this is an empirical hypothesis the paper leaves open.
  • The results come from a single ~1-hour session with young South Korean adults on one LLM; whether gradual disclosure sustains friendship over weeks or transfers to other cultures and models is untested.
  • Because self-criticism changed both style and empathy, the early-intimacy boost cannot be attributed to either factor alone; a factorial design would parse which component carries the effect.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. This paper reports two laboratory studies in which young South Korean adults chatted with a GPT-4o-based partner through a text interface. In both studies, all participants progressed through the same 33-question sequence adapted from Aron et al., partitioned into three sets of 11 questions, with the Subjective Closeness Index (SCI) and Interpersonal Judgment Scale (IJS) administered after each set. Study 1 (N=29) varied whether the LLM persona matched, opposed, or was neutral relative to the participant's pre-questionnaire profile. Study 2 (N=53) reused the same procedure, added blind versus non-blind conditions regarding whether participants were told the partner was an LLM, and replaced the vanilla LLM with a self-criticism-based response mechanism. The authors report that SCI increased significantly across stages, that persona similarity had little or inconsistent effect, that blindness had no effect, and that self-criticism improved naturalness and early intimacy. The paper closes with design guidelines for intimacy-supporting LLM chat systems.

Significance. If the central causal claim were supported, this would be a valuable extension of Aron et al.'s fast-friends procedure to human-LLM interaction, with practical implications for companion chatbots. The paper has notable strengths: it uses established intimacy scales, reports non-parametric tests appropriate for small repeated-measures samples, provides the full question and prompt materials in the appendices, and includes qualitative post-interview data that generate concrete hypotheses about style and empathy. These materials enhance reproducibility and will be useful to later work. However, the headline result is not established by the reported design, because the gradualness of self-disclosure is never varied. As a causal contribution, the paper's significance is therefore currently unrealized; the observed SCI increase could equally be produced by time on task, the emotional content of later questions, or growing familiarity.

major comments (4)
  1. [Section 3.3; Tables 1, 7, 9] The central claim that gradual self-disclosure "significantly enhances perceived social intimacy" (Abstract) is not supported by the design, because the disclosure trajectory is never varied. All participants receive the same fixed sequence of 33 questions, partitioned into three increasingly intimate sets (Section 3.3), and no condition uses static, shallow, or reverse-ordered disclosure. The increase in SCI across stages seen in Tables 1, 7, and 9 is therefore confounded with cumulative conversation time, the intrinsic emotional content of later questions, familiarity, and repeated measurement. The Introduction states that the authors "systematically vary the disclosure trajectory across interactions" (Section 1), but no such variation appears in either study. A control condition that presents the same material in a non-gradual order or with constant disclosure depth is the load-bearing missing cell for the causal attribution.
  2. [Section 4.5.3] The claim that self-criticism fosters higher intimacy in early stages is based on comparing Study 1 and Study 2 Stage 1 SCI means (Study 1: M = 2.26; Study 2: M = 2.86) across two different participant pools recruited in different months, with no inferential test reported. Because Study 2 has no within-experiment vanilla-LLM control condition, any difference between the studies could reflect sampling, timing, or procedural drift rather than the self-criticism mechanism. At minimum, the comparison should be accompanied by a statistical test and an explicit acknowledgment of its quasi-experimental status; ideally, response style should be varied within a single experiment.
  3. [Sections 3.1 and 3.3; RQ1.2] The factor described as "reciprocity" is not manipulated. The fit/unfit/neutral conditions vary the similarity between the participant's pre-questionnaire profile and the LLM persona (Section 3.1), while mutual self-disclosure is held constant: both the participant and the LLM answer the same Aron questions (Section 3.3). Consequently, the abstract's "regardless of persona reciprocity" conflates persona similarity with self-disclosure reciprocity, and RQ1.2/RQ2.2 cannot support any conclusion about the effect of reciprocal disclosure.
  4. [Tables 5 and 11] The Study 2 mean SCI and IJS values reported in Table 11 are numerically identical to the Study 1 values in Table 5 (e.g., Unfit SCI at Stages 1-3 are 3.055, 3.722, 4.361 in both tables), despite the different sample sizes (29 vs. 53) and the addition of blindness conditions. Since identical means across independent samples are extremely unlikely, either the table contains a copy error or the descriptive statistics were not recomputed for Study 2. This undermines the textual claims that "similar to Study 1, we observed..." and the descriptive support for RQ2.2; the authors should correct or explain this.
minor comments (6)
  1. [Table 10, Section 4.5.2] The stage main effect for IJS in Study 2 is reported as p = 0.0535 with an asterisk, and the text calls it a "significant main effect"; under the conventional 0.05 threshold used elsewhere, this should be reported as non-significant.
  2. [Section 4.4] The sentence "The detailed implementation is provided in These pre-questionnaire items are provided in the Appendix C" is a fragment and appears to be a merge error; it should read "The detailed implementation is provided in Appendix C."
  3. [Section 4.4] The Shapiro-Wilk results are described as "statistically insignificant" despite p-values below 0.05 (SCI: 0.0212, 0.0006, 0.0062; IJS: 0.0160, 0.0338); these are significant departures from normality and the wording is reversed.
  4. [References] Reference [6] contains a typo: "Social dialongue" should be "Social dialogue."
  5. [Section 3.3] The procedure allowed participants to review chat history while completing the intimacy measures; this could introduce a reflection or recency effect, and the decision to allow it should be justified or tested.
  6. [Section 4.1] The random assignment imbalances in Study 2 (more female participants in the unfit and non-blind conditions) are reported but not included as covariates; given the small cell sizes, the authors should discuss or adjust for this imbalance.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the paper is an empirical study using external scales and fresh samples; its main inferential threat is a missing control condition, not a derivation that reduces to its own inputs.

full rationale

This paper makes no formal derivation and fits no parameters. The central claim that gradual self-disclosure increases perceived intimacy is supported by repeated-measures non-parametric tests (Friedman, ART ANOVA) on the SCI and IJS scales, which are external instruments from prior human-relationship research (Aron et al.; Berscheid et al.; Byrne). There is no equation in which the outcome is defined by the input, and no fitted quantity is later relabeled as a prediction. The strongest design concern is that all participants received the same progressively deepening question sequence with no static or reverse-order control, so 'stage' is confounded with elapsed time, question content, and familiarity. That is an internal-validity limitation and a valid target for future experiments, but it is not circularity: the observed rise in SCI is an empirical result that could have been null, and it does not reduce to the experimental setup by construction. Similarly, Study 2's self-criticism criteria were informed by Study 1's qualitative complaints, but Study 2 used a fresh participant sample and the perceived-naturalness and intimacy outcomes were measured rather than enforced by the manipulation; this is iterative design, not self-definition. No load-bearing self-citation, imported uniqueness theorem, or ansatz-smuggling-via-citation appears in the text. Accordingly, the honest finding is no significant circularity, with score 0.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The central claim relies on the Aron task, the persona construction formulas, and the hand-written self-criticism prompt. None of these are fitted to outcome data, so the ledger is small, but the self-criticism prompt bundles several style changes and the persona inversion formula is an arbitrary transformation.

free parameters (3)
  • Persona inversion formula (unfit condition) = Unfit score = |8 - user score| on a 7-point Likert scale
    Hand-chosen transformation to define the opposite persona; all 'unfit' personas depend on it. Located in Appendix B.2.
  • Neutral persona setting = 4 on all 17 Likert items, age 2000
    Arbitrary fixed persona; affects the neutral condition only. Located in Appendix B.2.
  • Self-criticism style constraints = Four 100% rules (omit subject, drop periods, no questions, empathy only on persona-aligned concerns) plus Gen-Z…
    Bundle of hand-chosen style changes in Fig. 3; the effect of 'naturalness' cannot be separated from these specific rules, so the self-criticism manipulation is a compound treatment.
assumptions (3)
  • domain assumption Social Penetration Theory: intimacy develops through mutual and gradual self-disclosure
    The paper's central manipulation and interpretation rely on this theory (Section 2.1, citing Altman 1973).
  • domain assumption Aron et al.'s 36-question fast-friends procedure creates closeness in human-human pairs
    The paper adapts this procedure to human-LLM chats and assumes graduated questions induce comparable dynamics (Section 3.3).
  • ad hoc to paper The 33 used questions form three monotonically deepening self-disclosure sets
    The division into three sets of 11 is the paper's own construction and is not validated; the depth ordering is assumed from the original Aron list (Section 3.3).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Can LLMs and humans be friends? Uncovering factors affecting human-AI intimacy formation." pith.science (2026). https://pith.science/paper/QJZH4OXU

@misc{pith2026250524658,
  author       = {Pith},
  title        = {Pith review of: Can LLMs and humans be friends? Uncovering factors affecting human-AI intimacy formation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QJZH4OXU}},
  note         = {Machine review of arXiv:2505.24658}
}
read the original abstract

Large language models (LLMs) are increasingly being used in conversational roles, yet little is known about how intimacy emerges in human-LLM interactions. Although previous work emphasized the importance of self-disclosure in human-chatbot interaction, it is questionable whether gradual and reciprocal self-disclosure is also helpful in human-LLM interaction. Thus, this study examined three possible aspects contributing to intimacy formation: gradual self-disclosure, reciprocity, and naturalness. Study 1 explored the impact of mutual, gradual self-disclosure with 29 users and a vanilla LLM. Study 2 adopted self-criticism methods for more natural responses and conducted a similar experiment with 53 users. Results indicate that gradual self-disclosure significantly enhances perceived social intimacy, regardless of persona reciprocity. Moreover, participants perceived utterances generated with self-criticism as more natural compared to those of vanilla LLMs; self-criticism fostered higher intimacy in early stages. Also, we observed that excessive empathetic expressions occasionally disrupted immersion, pointing to the importance of response calibration during intimacy formation.

Figures

Figures reproduced from arXiv: 2505.24658 by the authors.

Figure 1
Figure 1. Chatroom interface used in the experiment consisted of two main components: (i) the chat thread (ii) the input box where [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Experimental workflow illustrating the examination of self-disclosure and intimacy during Human-LLM interactions, consisting [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Caption be applied after observing the responses of a human partner. Therefore, we used a fourth condition based on empathy alone when the LLM had to answer after the human partner’s response. Accordingly, we provided both the original Korean prompt and its English translation. 4.3 Procedure Although the procedure followed was similar to that in Study 1, we made subtle changes in the main experiment and post-intervi… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

74 extracted references · 65 canonical work pages

  1. [1]

    Irwin Altman. 1973. Social penetration: The development of interpersonal relationships. Rinehart, & Winston (1973)

    work page 1973

  2. [2]

    Arthur Aron, Edward Melinat, Elaine N Aron, Robert Darrin Vallone, and Renee J Bator. 1997. The experimental generation of interpersonal closeness: A procedure and some preliminary findings. Personality and social psychology bulletin 23, 4 (1997), 363–377

    work page 1997

  3. [3]

    Elliot Aronson. 1969. Some antecedents of interpersonal attraction.. In Nebraska symposium on motivation . University of Nebraska Press

    work page 1969

  4. [4]

    John H Berg and Valerian J Derlega. 1987. Themes in the study of self-disclosure. In Self-disclosure: Theory, research, and therapy . Springer, 1–8

    work page 1987

  5. [5]

    Ellen Berscheid, Mark Snyder, and Allen M Omoto. 1989. The Relationship Closeness Inventory: Assessing the closeness of interpersonal relationships. Journal of personality and Social Psychology 57, 5 (1989), 792

    work page 1989

  6. [6]

    Timothy Bickmore and Justine Cassell. 2005. Social dialongue with embodied conversational agents. Advances in natural multimodal dialogue systems (2005), 23–54

    work page 2005

  7. [7]

    Bonferroni

    C.E. Bonferroni. 1936. Teoria statistica delle classi e calcolo delle probabilità . Seeber. https://books.google.co.kr/books?id=3CY-HQAACAAJ

    work page 1936

  8. [8]

    Petter Bae Brandtzaeg, Marita Skjuve, and Asbjørn Følstad. 2022. My AI friend: How users of a social chatbot understand their human–AI friendship. Human Communication Research 48, 3 (2022), 404–429

    work page 2022

Show all 74 references
  1. [9]

    D Byrne. 1971. The Attraction Paradigm Academic Press. New York, NY, USA (1971)

    1971

  2. [10]

    Donn Erwin Byrne. [n. d.]. The attraction paradigm. (No Title) ([n. d.])

  3. [11]

    Nancy L Collins and Lynn Carol Miller. 1994. Self-disclosure and liking: a meta-analytic review. Psychological bulletin 116, 3 (1994), 457

    1994

  4. [12]

    Emmelyn Croes, Marjolijn L Antheunis, and Linwei He. 2023. You go first: The effects of self-disclosure reciprocity in human-chatbot interactions. In 2023 11th International Conference on Affective Computing and Intelligent Interaction Workshops and Demos (ACIIW) . IEEE, 1–4

    2023

  5. [13]

    Edward L Deci and Richard M Ryan. 2012. Self-determination theory. Handbook of theories of social psychology 1, 20 (2012), 416–436. Manuscript submitted to ACM Uncovering factors affecting human-AI intimacy formation 23

    2012

  6. [14]

    Malinda Desjarlais. 2022. The socially poor get richer, the rich get poorer: The effect of online self-disclosure on social connectedness and well-being is conditional on social anxiety and audience size. Cyberpsychology: Journal of Psychosocial Research on Cyberspace 16, 4 (2022)

    2022

  7. [15]

    Thomas Palmeira Ferraz, Kartik Mehta, Yu-Hsiang Lin, Haw-Shiuan Chang, Shereen Oraby, Sijia Liu, Vivek Subramanian, Tagyoung Chung, Mohit Bansal, and Nanyun Peng. 2024. LLM self-correction with DeCRIM: Decompose, critique, and refine for enhanced following of instructions with...

    2024 arXiv

  8. [16]

    Milton Friedman. 1937. The use of ranks to avoid the assumption of normality implicit in the analysis of variance. Journal of the american statistical association 32, 200 (1937), 675–701

    1937

  9. [17]

    ER Girden. 1992. ANOV A: Repeated measures. Vol. 84. Sage

    1992

  10. [18]

    Ulrich Gnewuch, Meng Yu, and Alexander Maedche. 2020. The effect of perceived similarity in dominance on customer self-disclosure to chatbots in conversational commerce. (2020)

    2020

  11. [19]

    Annabell Ho, Jeff Hancock, and Adam S Miner. 2018. Psychological, relational, and emotional effects of self-disclosure after conversations with a chatbot. Journal of Communication 68, 4 (2018), 712–733

    2018

  12. [20]

    Jess Hohenstein, Rene F Kizilcec, Dominic DiFranzo, Zhila Aghajari, Hannah Mieczkowski, Karen Levy, Mor Naaman, Jeffrey Hancock, and Malte F Jung. 2023. Artificial intelligence in communication impacts language and social relationships. Scientific Reports 13, 1 (2023), 5487

    2023

  13. [21]

    Shih-Hong Huang, Ya-Fang Lin, Zeyu He, Chieh-Yang Huang, and Ting-Hao Kenneth Huang. 2024. How Does Conversation Length Impact User’s Satisfaction? A Case Study of Length-Controlled Conversations with LLM-Powered Chatbots. In Extended Abstracts of the CHI Conference on Human F...

    2024

  14. [22]

    Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perelman, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Welihinda, Alan Hayes, Alec Radford, et al. 2024. Gpt-4o system card. arXiv preprint arXiv:2410.21276 (2024)

    2024 arXiv

  15. [23]

    Becky Inkster, Shubhankar Sarda, Vinod Subramanian, et al. 2018. An empathy-driven, conversational artificial intelligence agent (Wysa) for digital mental well-being: real-world data evaluation mixed-methods study. JMIR mHealth and uHealth 6, 11 (2018), e12106

    2018

  16. [24]

    Maurice Jakesch, Megan French, Xiao Ma, Jeffrey T Hancock, and Mor Naaman. 2019. AI-mediated communication: How the perception that profile text was written by AI affects trustworthiness. In Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems . 1–13

    2019

  17. [25]

    Oliver P John, Sanjay Srivastava, et al. 1999. The Big-Five trait taxonomy: History, measurement, and theoretical perspectives. (1999)

    1999

  18. [26]

    Leon OH Kroczek, Alexander May, Selina Hettenkofer, Andreas Ruider, Bernd Ludwig, and Andreas Mühlberger. 2024. The influence of persona and conversational task on social interactions with a LLM-controlled embodied conversational agent. arXiv preprint arXiv:2411.05653 (2024)

    2024 arXiv

  19. [27]

    Jean-Philippe Laurenceau, Lisa Feldman Barrett, and Paula R Pietromonaco. 1998. Intimacy as an interpersonal process: the importance of self- disclosure, partner disclosure, and perceived partner responsiveness in interpersonal exchanges. Journal of personality and social psyc...

    1998

  20. [28]

    I hear you, I feel you

    Yi-Chieh Lee, Naomi Yamashita, Yun Huang, and Wai Fu. 2020. "I hear you, I feel you": encouraging deep self-disclosure through a chatbot. In Proceedings of the 2020 CHI conference on human factors in computing systems . 1–12

    2020

  21. [29]

    Kai-Hui Liang, Weiyan Shi, Yoo Jung Oh, Hao-Chuan Wang, Jingwen Zhang, and Zhou Yu. 2024. Dialoging Resonance in Human-Chatbot Conversation: How Users Perceive and Reciprocate Recommendation Chatbot’s Self-Disclosure Strategy. Proceedings of the ACM on Human- Computer Interact...

    2024

  22. [30]

    Liangchen Luo, Zi Lin, Yinxiao Liu, Lei Shu, Yun Zhu, Jingbo Shang, and Lei Meng. 2023. Critique ability of large language models. arXiv preprint arXiv:2310.04815 (2023)

    2023 arXiv

  23. [31]

    Clifford Nass and Youngme Moon. 2000. Machines and mindlessness: Social responses to computers. Journal of social issues 56, 1 (2000), 81–103

    2000

  24. [32]

    The pandas development team. 2020. pandas-dev/pandas: Pandas. doi:10.5281/zenodo.3509134

    2020 doi

  25. [33]

    Iryna Pentina, Tyler Hancock, and Tianling Xie. 2023. Exploring relationship development with social chatbots: A mixed-method study of replika. Computers in Human Behavior 140 (2023), 107600

    2023

  26. [34]

    Marianne Promberger and Jonathan Baron. 2006. Do patients trust computers? Journal of Behavioral Decision Making 19, 5 (2006), 455–468

    2006

  27. [35]

    Shalom H Schwartz. 2009. Basic human values. sociologie 42 (2009), 249–288

    2009

  28. [36]

    Skipper Seabold and Josef Perktold. 2010. statsmodels: Econometric and statistical modeling with python. In 9th Python in Science Conference

    2010

  29. [37]

    Samuel Sanford Shapiro and Martin B Wilk. 1965. An analysis of variance test for normality (complete samples). Biometrika 52, 3-4 (1965), 591–611

    1965

  30. [38]

    Marita Skjuve, Asbjørn Følstad, and Petter Bae Brandtzæg. 2023. A longitudinal study of self-disclosure in human–chatbot relationships. Interacting with Computers 35, 1 (2023), 24–39

    2023

  31. [39]

    Guangzhi Sun, Xiao Zhan, and Jose Such. 2024. Building better ai agents: A provocation on the utilisation of persona in llm-based conversational agents. In Proceedings of the 6th ACM Conference on Conversational User Interfaces . 1–6

    2024

  32. [40]

    Xiaoyu Tan, Shaojie Shi, Xihe Qiu, Chao Qu, Zhenting Qi, Yinghui Xu, and Yuan Qi. 2023. Self-criticism: Aligning large language models with their understanding of helpfulness, honesty, and harmlessness. In Proceedings of the 2023 Conference on Empirical Methods in Natural Lang...

    2023

  33. [41]

    Raphael Vallat. 2018. Pingouin: statistics in Python. Journal of Open Source Software 3, 31 (Nov. 2018), 1026. doi:10.21105/joss.01026

    2018 doi

  34. [42]

    Oliphant, Matt Haberland, Tyler Reddy, David Cournapeau, Evgeni Burovski, Pearu Peterson, Warren Weckesser, Jonathan Bright, Stéfan J

    Pauli Virtanen, Ralf Gommers, Travis E. Oliphant, Matt Haberland, Tyler Reddy, David Cournapeau, Evgeni Burovski, Pearu Peterson, Warren Weckesser, Jonathan Bright, Stéfan J. van der Walt, Matthew Brett, Joshua Wilson, K. Jarrod Millman, Nikolay Mayorov, Andrew R. J. Nelson, E...

  35. [43]

    Jacob O Wobbrock, Leah Findlater, Darren Gergle, and James J Higgins. 2011. The aligned rank transform for nonparametric factorial analyses using only anova procedures. In Proceedings of the SIGCHI conference on human factors in computing systems . 143–146

    2011

  36. [44]

    Robert F Woolson. 2005. Wilcoxon signed-rank test. Encyclopedia of Biostatistics 8 (2005). Manuscript submitted to ACM Uncovering factors affecting human-AI intimacy formation 25 Appendix This appendix provides supplementary materials used to ensure the transparency and reprod...

    2005

  37. [46]

    (1) Year of birth ( 1996 - 2005 ) (2) Gender ( Male / Female )

    Demographics. (1) Year of birth ( 1996 - 2005 ) (2) Gender ( Male / Female )

    1996

  38. [47]

    (1) I usually feel energized when I socialize with others

    Personality and Behavior — Big Five Inventory [25] (7-point Likert scale). (1) I usually feel energized when I socialize with others. (2) I find it easy to cooperate with others. (3) I handle tasks responsibly and with care. (4) I believe I manage stress well. (5) I adapt easi...

  39. [48]

    (1) I place great importance on achieving personal goals

    Values and Beliefs — Schwartz’s Basic Human Values [35] (7-point Likert scale). (1) I place great importance on achieving personal goals. (2) I believe it is important to have control over my own life. (3) I think it’s important to follow social norms and traditions. (4) I tre...

  40. [49]

    we" statements each. For instance,

    Hobby and Interest — Self-Determination Theory [13] (7-point Likert scale). (1) I feel a strong interest in learning new things. (2) I enjoy exploring new places or traveling. (3) I find it meaningful to try new activities and gain new experiences. (4) I try to find balance an...

  41. [50]

    3.실험의절차 이연구를위해우리는이미 사전설문조사를 통해 참가자인 당신의성향을파악했고,이것은 당신의페르소나가 됩니다.이후 당신과 참가자들의성향을고려하여파트너를 매칭했고,두 사람이 함께 참여하는종류의공유실험을 준비했습니다

    참가인원및연령 만 19세 29세 사이의성인 남녀 90명을대상으로 하나,조기에종료될수있습니다. 3.실험의절차 이연구를위해우리는이미 사전설문조사를 통해 참가자인 당신의성향을파악했고,이것은 당신의페르소나가 됩니다.이후 당신과 참가자들의성향을고려하여파트너를 매칭했고,두 사람이 함께 참여하는종류의공유실험을 준비했습니다. 채팅을돕기위해제공되는총 33개질문이순차적으로제공될예정이며, 당신은파트너와질문에 해당하는 답변을 공유하는작업을 해주시면됩니다. 한질문에두 명의 피험자 모두가 답변해야 다음질문으로넘어갈수...

  42. [51]

    또한본연구 가진행되는동안귀하의 답변과, 매세트이후진행될설문에서의 답변이수집되지만,이정보는연구진을제외한 외부에공개되지않으며,연구종료 3년 후폐기됩니다.본연구팀에서는이연구를 통해얻은 모든정보를익명화하 여 분석하며, 학회지나 학회에발표할경우귀하를유추할수있는개인정보는 사용되지않을것입니다

    기밀유지와 피험자정보수집및 활용에대한검토 연구진행에앞서저희는 참가자들의성별및 나이에관한정보와 사전설문 내용을수집하게됩니다. 또한본연구 가진행되는동안귀하의 답변과, 매세트이후진행될설문에서의 답변이수집되지만,이정보는연구진을제외한 외부에공개되지않으며,연구종료 3년 후폐기됩니다.본연구팀에서는이연구를 통해얻은 모든정보를익명화하 여 분석하며, 학회지나 학회에발표할경우귀하를유추할수있는개인정보는 사용되지않을것입니다. 다만법이 요구하면귀하의개인정보가제공될수도있습니다. 또한 모니터요원,점검요원, 생명윤리위원회...

  43. [52]

    The session is expected to last about 1 hour, though it may extend by approximately 10 minutes depending on the flow of the experiment

    Purpose and Overview of the Study Manuscript submitted to ACM Uncovering factors affecting human-AI intimacy formation 29 This study aims to understand how relationships develop between two individuals through online conversations. The session is expected to last about 1 hour,...

  44. [53]

    However, recruitment may end earlier

    Number and Age Range of Participants We plan to recruit 90 male and female adults aged between 19 and 29. However, recruitment may end earlier

  45. [54]

    persona

    Study Procedure Before the study, we conducted a preliminary survey to understand your traits, which were used to create your “persona. ” Based on your responses and those of other participants, we matched you with a partner and prepared a shared conversational task. A total o...

  46. [55]

    Therefore, there is no expected risk to your safety or privacy

    Risks and Benefits The conversation prompts do not ask for any personally identifiable or sensitive information. Therefore, there is no expected risk to your safety or privacy. Aside from a gift voucher provided at the end of the study, there are no additional benefits

  47. [56]

    During the study, your conversation responses and answers to follow-up surveys will be collected

    Confidentiality and Use of Collected Data Before the study, we collect information such as your gender, age, and responses to the preliminary survey. During the study, your conversation responses and answers to follow-up surveys will be collected. This data will not be shared ...

  48. [57]

    You may refuse to participate at any time without any penalty or disadvantage

    Voluntary Participation and Right to Withdraw • Your participation in this study is entirely voluntary. You may refuse to participate at any time without any penalty or disadvantage. • You may withdraw from the study at any time and request that your data be discarded. This de...

  49. [58]

    I was born in {st.session_state.userInfo[0]}

  50. [59]

    I am {st.session_state.userInfo[1]}

  51. [60]

    (scale: {st.session_state.userInfo[2]}) Manuscript submitted to ACM 30 Hong et al

    I usually feel energized when I socialize with others. (scale: {st.session_state.userInfo[2]}) Manuscript submitted to ACM 30 Hong et al

  52. [61]

    (scale: {st.session_state.userInfo[3]})

    I find it easy to cooperate with others. (scale: {st.session_state.userInfo[3]})

  53. [62]

    (scale: {st.session_state.userInfo[4]})

    I handle tasks responsibly and with care. (scale: {st.session_state.userInfo[4]})

  54. [63]

    (scale: {st.session_state.userInfo[5]})

    I believe I manage stress well. (scale: {st.session_state.userInfo[5]})

  55. [64]

    (scale: {st.session_state.userInfo[6]})

    I adapt easily to new situations. (scale: {st.session_state.userInfo[6]})

  56. [65]

    (scale: {st.session_state.userInfo[7]})

    I place great importance on achieving personal goals. (scale: {st.session_state.userInfo[7]})

  57. [66]

    (scale: {st.session_state.userInfo[8]})

    I believe it is important to have control over my own life. (scale: {st.session_state.userInfo[8]})

  58. [67]

    (scale: {st.session_state.userInfo[9]})

    I think it 's important to follow social norms and traditions. (scale: {st.session_state.userInfo[9]})

  59. [68]

    (scale: {st.session_state.userInfo[10]})

    I treat everyone equally and respect diverse cultures and perspectives. (scale: {st.session_state.userInfo[10]})

  60. [69]

    (scale: {st.session_state.userInfo[11]})

    I value helping others and contributing to their happiness. (scale: {st.session_state.userInfo[11]})

  61. [70]

    (scale: {st.session_state.userInfo[12]})

    I feel a strong interest in learning new things. (scale: {st.session_state.userInfo[12]})

  62. [71]

    (scale: {st.session_state.userInfo[13]})

    I enjoy exploring new places or traveling. (scale: {st.session_state.userInfo[13]})

  63. [72]

    (scale: {st.session_state.userInfo[14]})

    I find it meaningful to try new activities and gain new experiences. (scale: {st.session_state.userInfo[14]})

  64. [73]

    (scale: {st.session_state.userInfo[15]})

    I try to find balance and satisfaction in life through leisure activities. (scale: {st.session_state.userInfo[15]})

  65. [74]

    (scale: {st.session_state.userInfo[16]}) We set the scale values in the above prompt differently for each of the three conditions: fit, unfit, and neutral

    I believe that pursuing hobbies is important for my personal growth. (scale: {st.session_state.userInfo[16]}) We set the scale values in the above prompt differently for each of the three conditions: fit, unfit, and neutral. First, for the fit condition, we assigned a persona ...

    2000

  66. [2020]

    Nature Methods 17 (2020), 261–272

    SciPy 1.0: Fundamental Algorithms for Scientific Computing in Python. Nature Methods 17 (2020), 261–272. doi:10.1038/s41592-019-0686-2

    2020 doi

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.