Pith. sign in

REVIEW 3 major objections 6 minor 41 references

What Shapes Writers' Decisions to Disclose AI Use?

T0 review · 3 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 Writers' decisions to disclose AI involvement are shaped by twelve factors in three groups—procedural, social, and personal—according to this synthesis of prior work.

desk verdict Useful curated taxonomy of AI-disclosure factors, clearly mapped to prior work, but the procedural factors are inferred from attribution studies rather than disclosure studies and should be flagged as such. read the letter →

arxiv 2505.20727 v1 pith:SR5NAL7L submitted 2025-05-27 cs.HC

Jingchao Fang , Mina Lee This is my paper
classification cs.HC
keywords human-AIcollaborativewritingAIdisclosuretransparencycontentco-creationlargelanguagemodelswriterdecision-makingfactorsynthesis
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

This paper asks why writers disclose or hide that AI helped them write, and answers with a curated list of twelve factors organized into three groups. The procedural group covers how AI was used; the social group covers how readers, rules, and norms react; the personal group covers the writer's own confidence, values, and demographics. The authors synthesize prior findings rather than test the list empirically, and they present it as a foundation for studying how the factors combine. If the list holds up, transparency policies could target specific levers rather than treating disclosure as a single binary choice.

What carries the argument

The load-bearing object is the curated factor list itself, which separates binary disclosure from fine-grained attribution and gives each factor a definition with contrasting poles, such as low versus high replaceability or direct versus indirect use. The list carries the argument by turning an understudied question, why disclose AI use, into a vocabulary of measurable variables that can later be manipulated or surveyed in a combined design.

What would settle it

A factorial vignette study in which writers report their disclosure intention across all twelve factors would settle whether each factor moves decisions in the predicted direction; if any factor shows no effect or an effect opposite to its stated polarity, the claim is wrong. A systematic search with explicit inclusion criteria that surfaces a category of factors outside the three groups would also falsify the list's completeness.

Watch

Extended reading notes

Core claim

The central claim is that a set of twelve factors, grouped into procedural, social, and personal categories, can shape a writer's decision to disclose AI involvement in co-created text. Procedural factors are replaceability, effortfulness, intentionality, and directness. Social factors are detectability, regulatory penalty for misclaim, perceived negative and positive effects of disclosure, and prevalence of disclosure. Personal factors are self-efficacy, internal judgments, and demographics. The paper does not claim these have been measured jointly; it curates them from prior studies and theories so that future work can investigate combined effects.

Load-bearing premise

The load-bearing premise is that the 39 prior works selected for this synthesis, chosen without a systematic search protocol, include all major factors that shape writers' disclosure decisions.

Editorial extensions

If this is right

  • The twelve factors give future studies a shared set of variables, so results on AI disclosure can be compared instead of remaining isolated observations.
  • Transparency interventions can operate at three levels: how writers use AI, how the social context frames disclosure, and how individual writers' confidence and values are addressed.
  • Writers' disclosure decisions are likely to involve trade-offs, because factors can push in opposite directions, such as high detectability encouraging disclosure while strong perceived stigma discourages it.
  • The curated list can serve as the independent-variable set for an integrative study that measures the relative importance of all twelve factors together.

Reading between the lines

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

  • If the paper is right, disclosure behavior could be modeled as a weighted trade-off among these factors, and policies could shift the weights rather than simply mandating disclosure.
  • The authors do not test this, but the procedural factors suggest that disclosure labels could distinguish direct copying from indirect inspiration, which is finer and more truthful than a single 'AI-generated' tag.
  • The social factors imply a feedback loop the paper leaves implicit: as disclosure becomes prevalent, the perceived cost of disclosing may fall, making further disclosure easier.
  • A natural next experiment, vignettes that vary all twelve factors at once, would measure which combinations dominate, something the paper does not claim to have done.
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

3 major / 6 minor

Summary. The paper asks what factors shape writers' decisions to disclose AI use in writing. It synthesizes prior work in human-AI interaction and behavioral science into a curated list of twelve factors grouped into three categories: procedural (Replaceability, Effortfulness, Intentionality, Directness), social (Detectability, Regulatory penalty for misclaim, Perceived negative effects of disclosure, Perceived positive effects of disclosure, Prevalence of disclosure), and personal (Self-efficacy, Internal judgments, Demographics). The authors explicitly frame this list as a starting point for future holistic investigation, not as an empirically validated model. The paper includes definitions for each factor and an appendix mapping each factor to prior sources.

Significance. The paper makes a useful contribution by assembling scattered findings into a structured taxonomy and by explicitly distinguishing disclosure from attribution. Its strength is the clear, concise definitions of candidate factors and the recognition that disclosure decisions are jointly shaped by procedural, social, and personal considerations. The paper is honest about the preliminary nature of the synthesis, stating that the list may not cover every factor and that the next step is to measure relative significance. However, the contribution is primarily conceptual: there is no new empirical data, no systematic review protocol, and the paper deliberately stops short of testable predictions. If the goal is to seed an integrative research agenda, the paper succeeds in providing a readable map, but its scientific value depends entirely on the subsequent empirical work it motivates.

major comments (3)
  1. [Section 3.1, Section 3.4, footnote 2] The four procedural factors (Replaceability, Effortfulness, Intentionality, Directness) are justified primarily by the collective-centered creation (CCC) framework and studies of credit attribution, psychological ownership, and contribution taxonomies (Refs. 16, 18, 30, 36, 37). As the paper itself notes in footnote 2, disclosure is a binary indication of AI involvement, distinct from fine-grained attribution. None of the cited procedural-factor studies measures disclosure as a dependent variable. Yet Section 3.4 claims that 'existing studies suggest that each factor may independently influence writers' disclosure decisions,' which overstates the evidence for the procedural factors. This unstated inference is load-bearing because it supports the inclusion of these four factors as determinants of disclosure. I recommend adding a paragraph that explicitly acknowledges the inferential leap from credit-attribution studies to disclosure decisions and proposes a testable hypothesis or an empirical route connecting each procedural factor to disclosure behavior.
  2. [Section 3, Appendix A] The synthesis is described as 'thematically identified' (Section 3) and 'systematically synthesize[d]' (end of Section 2), but the paper provides no search protocol, inclusion criteria, or rationale for selecting the 39 cited works. Without a systematic method, the curated list may reflect the authors' selective reading rather than a representative mapping of the literature. This is a specific concern for the central claim that the list is a foundation for holistic investigation. Please either add a brief description of the search and selection process, or temper the 'systematic' phrasing and acknowledge that the synthesis is scoping in nature.
  3. [Section 3.4, factor definitions in Sections 3.1-3.3] The paper states that 'each factor may independently influence writers' disclosure decisions,' but several definitions overlap and likely interact. For example, Effortfulness and Intentionality both capture the degree of human involvement; Detectability and Regulatory penalty for misclaim are explicitly noted to interact (Section 3.2(2)); and Perceived negative effects of disclosure likely correlate with Detectability. If these factors are not orthogonal, the proposed integrative approach will need to account for collinearity and interaction effects. Please discuss the assumed independence of the factors or clarify which combinations are expected to co-occur.
minor comments (6)
  1. [References [16, 17, 10, 21, 22, 23, 24, 25, 38, 39]] Many cited sources are arXiv preprints rather than peer-reviewed publications. For empirical claims, please mark the publication status or note where results have been formally peer-reviewed, as this affects the strength of the evidence.
  2. [Table 1, Appendix A.1] The table maps each procedural factor to source-specific terms, but the mapping for 'Effortfulness' to 'Amount of effort' and 'Intentionality' to 'Leadership'/'Control' appears to conflate related but distinct constructs. Consider adding a short explanation of how the source terms were interpreted.
  3. [Reference [3]] The citation for self-efficacy lists 'Albert Bandura and Sebastian Wessels.' The canonical 1997 work is Bandura's 'Self-Efficacy: The Exercise of Control' (W.H. Freeman), authored by Bandura alone; please verify the reference.
  4. [Section 3.2(2)] The parenthetical note that regulatory penalty does not always lead to disclosure is important and could be expanded into a sentence of its own, as it highlights a non-obvious interaction with Detectability.
  5. [Section 2, footnote 2] The definition of disclosure as 'a binary yes or no' sits in tension with the later discussion of 'extent' of disclosure in the introduction (e.g., 'detailing how AI contributed and to what extent' under attribution). Clarify whether disclosure is viewed strictly as binary or as a graded construct, as this affects measurement.
  6. [General] The paper would benefit from a short 'Limitations' section that explicitly acknowledges the non-systematic selection of sources, the reliance on preprints, and the absence of empirical validation, in addition to the existing caveat that the list 'may not cover every possible factor.'

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper is a transparent synthesis of external prior work and makes no predictive or fitted claims.

full rationale

This paper is an explicitly integrative synthesis, not a derivation. It curates twelve factors from prior external works and presents them as candidates, hedged as 'factors that can potentially shape writers' disclosure decisions.' There is no fitted parameter, no predictive model, no equation, and no quantity computed from the paper's own inputs. The procedural factors are transparently drawn from the collective-centered creation framework and other cited studies, with Appendix Table 1 mapping each new factor name to its original name across sources; this is explicit synthesis, not renaming presented as discovery or unification. The skeptic's concern that some cited works concern credit attribution rather than disclosure behavior is an external-validity or inference-strength limitation, not circularity: the paper does not claim those studies prove the factors determine disclosure, only that the factors are plausible candidates informed by prior work. Similarly, the absence of a systematic literature search is a scope limitation, not an input-output equivalence. There are no self-citations carrying the argument, and no claim is made that a result is predicted from a fitted parameter. The paper is self-contained as a position piece and honest about the speculative status of its list.

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

The ledger is mostly empty because the paper makes no quantitative claims. The main assumptions concern the completeness and distinctness of the curated factor list.

assumptions (3)
  • domain assumption The factor list derived from the cited works is representative of the key factors shaping disclosure decisions.
    The paper presents no empirical validation and relies on a non-systematic sample of 39 references. If key factors are missing, the proposed integrative model would be incomplete.
  • domain assumption The factors are distinct and can be studied jointly without major overlap.
    The paper does not demonstrate discriminant validity among the 12 factors; some factors, such as perceived negative effects and regulatory penalty, may be confounded.
  • domain assumption The cited prior works, many of which are preprints, provide accurate and transferable findings.
    The paper leans heavily on arXiv preprints and recent conference papers that have not all completed peer review, yet uses them as the evidence base for the factor list.

how reviews work

0 comments
Cite this review

Pith. "Pith review of What Shapes Writers' Decisions to Disclose AI Use?." pith.science (2026). https://pith.science/paper/SR5NAL7L

@misc{pith2026250520727,
  author       = {Pith},
  title        = {Pith review of: What Shapes Writers' Decisions to Disclose AI Use?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SR5NAL7L}},
  note         = {Machine review of arXiv:2505.20727}
}
read the original abstract

Have you ever read a blog or social media post and suspected that it was written--at least in part--by artificial intelligence (AI)? While transparently acknowledging contributors to writing is generally valued, why some writers choose to disclose or withhold AI involvement remains unclear. In this work, we ask what factors shape writers' decisions to disclose their AI use as a starting point to effectively advocate for transparency. To shed light on this question, we synthesize study findings and theoretical frameworks in human-AI interaction and behavioral science. Concretely, we identify and curate a list of factors that could affect writers' decisions regarding disclosure for human-AI co-created content.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

41 extracted references · 28 canonical work pages

  1. [1]

    Abdullah Almaatouq, Thomas L Griffiths, Jordan W Suchow, Mark E Whiting, James Evans, and Duncan J Watts. 2024. Beyond playing 20 questions with nature: Integrative experiment design in the social and behavioral sciences. Behavioral and Brain Sciences 47 (2024), e33

    work page 2024

  2. [2]

    Joseph J Avery, Patricia Sánchez Abril, and Alissa del Riego. 2024. Attributing AI Authorship: Towards a System of Icons for Legal and Ethical Disclosure. Nw. J. Tech. & Intell. Prop. 22 (2024), 1

    work page 2024

  3. [3]

    Albert Bandura and Sebastian Wessels. 1997. Self-efficacy. Cambridge University Press Cambridge. 4 What Shapes Writers’ Decisions to Disclose AI Use?

    work page 1997

  4. [4]

    ACM Publications Board. n.d.. ACM Policy on Authorship. https://www.acm.org/publications/policies/new-acm-policy-on-authorship. Accessed: 2025-04-15

    2025

  5. [5]

    Amy Brand, Liz Allen, Micah Altman, Marjorie Hlava, and Jo Scott. 2015. Beyond authorship: Attribution, contribution, collaboration, and credit. Learned Publishing 28, 2 (2015)

    work page 2015

  6. [6]

    John Carvalho, Angie Chung, and Michael Koliska. 2021. Defying transparency: Ghostwriting from the Jazz Age to social media. Journalism 22, 3 (2021), 709–725

    work page 2021

  7. [7]

    Robert B Cialdini and Melanie R Trost. 1998. Social influence: Social norms, conformity and compliance. (1998)

    work page 1998

  8. [8]

    Justin Cranshaw and Aniket Kittur. 2011. The polymath project: lessons from a successful online collaboration in mathematics. In Proceedings of the SIGCHI conference on human factors in computing systems . 1865–1874

    work page 2011

Show all 41 references
  1. [9]

    Leslie M Dawson. 1997. Ethical differences between men and women in the sales profession. Journal of Business Ethics 16 (1997), 1143–1152

    1997

  2. [10]

    Fiona Draxler, Daniel Buschek, Mikke Tavast, Perttu Hämäläinen, Albrecht Schmidt, Juhi Kulshrestha, and Robin Welsch. 2023. Gender, age, and technology education influence the adoption and appropriation of LLMs. arXiv preprint arXiv:2310.06556 (2023)

    2023 arXiv

  3. [11]

    Fiona Draxler, Anna Werner, Florian Lehmann, Matthias Hoppe, Albrecht Schmidt, Daniel Buschek, and Robin Welsch. 2024. The AI ghostwriter effect: When users do not perceive ownership of AI-generated text but self-declare as authors. ACM Transactions on Computer-Human Interacti...

    2024

  4. [12]

    Katy Ilonka Gero and Lydia B Chilton. 2019. Metaphoria: An algorithmic companion for metaphor creation. InProceedings of the 2019 CHI conference on human factors in computing systems . 1–12

    2019

  5. [13]

    Peter C Gøtzsche, Jerome P Kassirer, Karen L Woolley, Elizabeth Wager, Adam Jacobs, Art Gertel, and Cindy Hamilton. 2009. What should be done to tackle ghostwriting in the medical literature? PLoS Medicine 6, 2 (2009), e1000023

    2009

  6. [14]

    Philipp Hacker, Andreas Engel, and Marco Mauer. 2023. Regulating ChatGPT and other large generative AI models. In Proceedings of the 2023 ACM conference on fairness, accountability, and transparency . 1112–1123

    2023

  7. [15]

    Jessica He, Stephanie Houde, Gabriel E Gonzalez, Darío Andrés Silva Moran, Steven I Ross, Michael Muller, and Justin D Weisz. 2024. AI and the Future of Collaborative Work: Group Ideation with an LLM in a Virtual Canvas. In Proceedings of the 3rd Annual Meeting of the Symposiu...

    2024

  8. [16]

    Jessica He, Stephanie Houde, and Justin D Weisz. 2025. Which Contributions Deserve Credit? Perceptions of Attribution in Human-AI Co-Creation. arXiv preprint arXiv:2502.18357 (2025)

    2025 arXiv

  9. [17]

    It was 80% me, 20% AI

    Angel Hsing-Chi Hwang, Q Vera Liao, Su Lin Blodgett, Alexandra Olteanu, and Adam Trischler. 2024. " It was 80% me, 20% AI": Seeking Authenticity in Co-Writing with Large Language Models. arXiv preprint arXiv:2411.13032 (2024)

    2024 arXiv

  10. [18]

    Donal Khosrowi, Finola Finn, and Elinor Clark. 2023. Diffusing the creator: Attributing credit for generative AI outputs. In Proceedings of the 2023 AAAI/ACM Conference on AI, Ethics, and Society . 890–900

    2023

  11. [19]

    Dmitry Kobak, Rita González-Márquez, Emőke-Ágnes Horvát, and Jan Lause. 2024. Delving into ChatGPT usage in academic writing through excess vocabulary. arXiv preprint arXiv:2406.07016 (2024)

    2024 arXiv

  12. [20]

    Lori NK Leonard, Timothy Paul Cronan, and Jennifer Kreie. 2004. What influences IT ethical behavior intentions—planned behavior, reasoned action, perceived importance, or individual characteristics? Information & Management 42, 1 (2004), 143–158

    2004

  13. [21]

    Zhuoyan Li, Chen Liang, Jing Peng, and Ming Yin. 2024. How Does the Disclosure of AI Assistance Affect the Perceptions of Writing? arXiv preprint arXiv:2410.04545 (2024)

    2024 arXiv

  14. [22]

    Weixin Liang, Zachary Izzo, Yaohui Zhang, Haley Lepp, Hancheng Cao, Xuandong Zhao, Lingjiao Chen, Haotian Ye, Sheng Liu, Zhi Huang, et al

  15. [23]

    Weixin Liang, Yaohui Zhang, Zhengxuan Wu, Haley Lepp, Wenlong Ji, Xuandong Zhao, Hancheng Cao, Sheng Liu, Siyu He, Zhi Huang, et al. 2024. Mapping the increasing use of LLMs in scientific papers. arXiv preprint arXiv:2404.01268 (2024)

    2024 arXiv

  16. [24]

    Q Vera Liao and Jennifer Wortman Vaughan. 2023. Ai transparency in the age of llms: A human-centered research roadmap. arXiv preprint arXiv:2306.01941 10 (2023)

    2023 arXiv

  17. [25]

    Shervin Minaee, Tomas Mikolov, Narjes Nikzad, Meysam Chenaghlu, Richard Socher, Xavier Amatriain, and Jianfeng Gao. 2024. Large language models: A survey, 2024. arXiv preprint arXiv:2402.06196 (2024)

    2024 arXiv

  18. [26]

    Andrés Monroy-Hernández, Benjamin Mako Hill, Jazmin Gonzalez-Rivero, and Danah Boyd. 2011. Computers can’t give credit: How automatic attribution falls short in an online remixing community. InProceedings of the SIGCHI Conference on Human Factors in Computing Systems . 3421–3430

    2011

  19. [27]

    Jon L Pierce, Tatiana Kostova, and Kurt T Dirks. 2003. The state of psychological ownership: Integrating and extending a century of research. Review of general psychology 7, 1 (2003), 84–107

    2003

  20. [28]

    Travis C Pratt, Francis T Cullen, Christine S Sellers, L Thomas Winfree Jr, Tamara D Madensen, Leah E Daigle, Noelle E Fearn, and Jacinta M Gau

  21. [29]

    Elizabeth M Renieris, David Kiron, and Steven Mills. 2024. Artificial Intelligence Disclosures Are Key to Customer Trust. MIT Sloan Management Review (Online) (2024), 1–4

    2024

  22. [30]

    Jeba Rezwana and Mary Lou Maher. 2023. User perspectives on ethical challenges in human-AI co-creativity: A design fiction study. In Proceedings of the 15th Conference on Creativity and Cognition . 62–74. 5 Fang and Lee

    2023

  23. [31]

    Linda A Riley and Stuart C Brown. 1996. Crafting a public image: An empirical study of the ethics of ghostwriting. Journal of Business Ethics 15 (1996), 711–720

    1996

  24. [32]

    Alex Shaw, Vivian Li, and Kristina R Olson. 2012. Children apply principles of physical ownership to ideas.Cognitive science 36, 8 (2012), 1383–1403

    2012

  25. [33]

    Aram Sinnreich, Mark Latonero, and Marissa Gluck. 2009. Ethics reconfigured: How today’s media consumers evaluate the role of creative reappropriation. Information, Communication & Society 12, 8 (2009), 1242–1260

    2009

  26. [34]

    James T Tedeschi. 2013. Impression management theory and social psychological research . Academic Press

    2013

  27. [35]

    Lynn Walsh. 2024. New research: Journalists should disclose their use of AI. Here’s how. https://trustingnews.org/trusting-news-artificial- intelligence-ai-research-newsroom-cohort/. Accessed: 2025-05-25

    2024

  28. [36]

    Ruyuan Wan, Simret Araya Gebreegziabher, Toby Jia-Jun Li, and Karla Badillo-Urquiola. 2024. CoCo Matrix: Taxonomy of Cognitive Contributions in Co-writing with Intelligent Agents. In Proceedings of the 16th Conference on Creativity & Cognition . 504–511

    2024

  29. [37]

    Yuxin Xu, Mengqiu Cheng, and Anastasia Kuzminykh. 2024. What Makes It Mine? Exploring Psychological Ownership over Human-AI Co-Creations. In Proceedings of the 50th Graphics Interface Conference . 1–8

    2024

  30. [38]

    Qihui Zhang, Chujie Gao, Dongping Chen, Yue Huang, Yixin Huang, Zhenyang Sun, Shilin Zhang, Weiye Li, Zhengyan Fu, Yao Wan, et al. 2024. LLM-as-a-coauthor: Can mixed human-written and machine-generated text be detected? arXiv preprint arXiv:2401.05952 (2024)

    2024 arXiv

  31. [39]

    Zhiping Zhang, Chenxinran Shen, Bingsheng Yao, Dakuo Wang, and Tianshi Li. 2024. Secret Use of Large Language Model (LLM). arXiv preprint arXiv:2409.19450 (2024). A COMPARISON TO PRIOR WORK A.1 Procedural Factors Since each procedural factor we described has been mentioned in ...

    2024 arXiv

  32. [2010]

    Justice Quarterly 27, 6 (2010), 765–802

    The empirical status of social learning theory: A meta-analysis. Justice Quarterly 27, 6 (2010), 765–802

    2010

  33. [2024]

    arXiv preprint arXiv:2403.07183 (2024)

    Monitoring ai-modified content at scale: A case study on the impact of chatgpt on ai conference peer reviews. arXiv preprint arXiv:2403.07183 (2024)

    2024 arXiv

Pith tools

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