Co-Writing with AI, on Human Terms: Aligning Research with User Demands Across the Writing Process

May combine two claims.

In the Formal context, reviewing support was limited and focused on evaluation using quantitative feedback like readability metrics and visual feedback such as progress bars.

“Finally, reviewing support was limited, and focused on evaluation using quantitative feedback like readability metrics [88], and visual feedback such as progress bars [90].”

• cites [88] 88. Seyed Parsa Neshaei, Roman Rietsche, Xiaotian Su, and Thiemo Wambsganss. 2024. Enhancing Peer Review with AI- Powered Suggestion Generation Assistance: Investigating the Design Dynamics. In Proceedings of the 29th International Conference on Intelligent User Interfaces (Greenville, SC, USA) (IUI '24). Association for Computing Machinery, New York, NY, USA, 88-102. https://doi.org/10.1145/3640543.3645169 sourceDisputed

  The citation [90] is attached to the claim about visual feedback such as progress bars. Section 3.2 describes Hamta’s interface as providing inline or list-based text suggestions, a word count, and a help button—not progress-bar feedback. Section 6.2 says the study did not measure review quality, contrary to the characterization of its support as focused on evaluation. (LLM, conf=0.83)

  Checked by a reasoning model — may be wrong.

  Experimental — accuracy not yet measured
• cites [90] 90. Eric Nichols, Leo Gao, and Randy Gomez. 2020. Collaborative Storytelling with Large-scale Neural Language Models. In Proceedings of the 13th ACM SIGGRAPH Conference on Motion, Interaction and Games (Virtual Event, SC, USA) (MIG '20). Association for Computing Machinery, New York, NY, USA, Article 17, 10 pages. https: //doi.org/10.1145/3424636.3426903 sourceCould not verify

  Citation [90] is attached to the claim about visual feedback such as progress bars, not to the readability-metrics example attributed to [88]. The paper’s evaluation discussion and Figure 5 describe stacked bar graphs showing evaluators’ preferences between systems, but do not describe progress bars or visual feedback for reviewing a story (searched: progress bars, reviewing support, feedback). (LLM, conf=0.76)

  Experimental — accuracy not yet measured

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Rezwana et al.'s work highlighted that ownership perceptions depend on both contribution levels and leadership in the writing process, suggesting that interaction designs that maximize user agency can enhance ownership.

“Rezwana et al. 's [105] work highlighted that ownership perceptions depend on both contribution levels and leadership in the writing process, suggesting that interaction designs that maximize user agency can enhance ownership.”

• cites [105] 105. Mohi Reza, Nathan M Laundry, Ilya Musabirov, Peter Dushniku, Zhi Yuan "Michael" Yu, Kashish Mittal, Tovi Grossman, Michael Liut, Anastasia Kuzminykh, and Joseph Jay Williams. 2024. ABScribe: Rapid Exploration & Organization of Multiple Writing Variations in Human-AI Co-Writing Tasks using Large Language Models. In Proceedings of the CHI Conference on Human Factors in Computing Systems. 1-18. sourceCould not verify

  The Results (sections 5 and 5.1; page unavailable) discuss perceptions of AI-assisted revision and the ability to manage writing variations. The passages do not address whether ownership perceptions depend on contribution levels or leadership in writing. Searches for ownership, leadership, contribution levels, and agency found no passage establishing that relationship. (LLM, conf=0.72)

  Experimental — accuracy not yet measured

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

The coding framework was established early in the process, allowing inductive identification of rich qualitative themes from the data.

“Following King & Brooks [67] and Braun & Clarke [12] we established our coding framework early in the process allowing us to inductively identify rich qualitative themes from the data.”

• cites [67] 67. Tae Soo Kim, Yoonjoo Lee, Minsuk Chang, and Juho Kim. 2023. Cells, Generators, and Lenses: Design Framework for Object-Oriented Interaction with Large Language Models. In Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology (San Francisco, CA, USA) (UIST '23). Association for Computing Machinery, New York, NY, USA, Article 4, 18 pages. https://doi.org/10.1145/3586183.3606833 sourceCould not verify

  The cited sentence describes the citing authors’ own coding process and names King & Brooks and Braun & Clarke, not Kim et al., as its methodological citations. Our copy of Kim et al. discusses an LLM interface design framework and a designer workshop (Sections 6 and 6.2), but no passage found addresses establishing a coding framework early or inductively identifying qualitative themes; searches covered those terms and qualitative coding methods. Page numbers were unavailable in our copy. (LLM, conf=0.72)

  Experimental — accuracy not yet measured
• cites [12] 12. Kyle Booten and Katy Ilonka Gero. 2021. Poetry Machines: Eliciting Designs for Interactive Writing Tools from Poets. In Proceedings of the 13th Conference on Creativity and Cognition (Virtual Event, Italy) (C&C '21). Association for Computing Machinery, New York, NY, USA, Article 51, 5 pages. https://doi.org/10.1145/3450741.3466813 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

The systematic literature review followed PRISMA guidelines.

“We conducted a systematic literature review following PRISMA guidelines [93] to identify and analyze research on AI writing support systems.”

• cites [93] 93. Hiroyuki Osone, Jun-Li Lu, and Yoichi Ochiai. 2021. BunCho: AI Supported Story Co-Creation via Unsupervised Multitask Learning to Increase Writers' Creativity in Japanese. In Extended Abstracts of the 2021 CHI Conference on Human Factors in Computing Systems (Yokohama, Japan) (CHI EA '21). Association for Computing Machinery, New York, NY, USA, Article 19, 10 pages. https://doi.org/10.1145/3411763.3450391 sourceCould not verify

  The Introduction discusses AI-assisted creative writing, and Section 4 describes a BunCho user study, not systematic-review methods. Searches for PRISMA guidelines and a systematic literature review found no passage addressing either in our copy; the paper therefore cannot verify the citation attached to “PRISMA guidelines.” Page numbers were not available. (LLM, conf=0.77)

  Experimental — accuracy not yet measured

May combine two claims.

Kobiella et al. found that participants who viewed AI as an enhancement tool rather than a replacement reported stronger feelings of accomplishment and ownership, while those who perceived their contributions as minimal experienced diminished ownership.

“Kobiella et al. [69] found that participants who viewed AI as an enhancement tool rather than a replacement reported stronger feelings of accomplishment and ownership, while those who perceived their contributions as minimal experienced diminished ownership.”

• cites [69] 69. Megan Knittel, Shelby Pitts, and Rick Wash. 2019. " The Most Trustworthy Coin" How Ideological Tensions Drive Trust in Bitcoin. Proceedings of the ACM on Human-Computer Interaction 3, CSCW (2019), 1-23. sourceCould not verify

  Split into 2 atoms — 0 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 2 unverifiable. Expand the claim to see per-atom verdicts.

  Experimental — accuracy not yet measured

May combine two claims.

In the Academic context, translating support provides preliminary drafts or helps the user to restructure their ideas in a different form, but encourages the user to write and integrate ideas on their own.

“Translating support provides preliminary drafts or helps the user to restructure their ideas in a different form (e.g., point form to prose), but encourages the user to write and integrate ideas on their own [21,122,124].”

• cites [21] 21. Kiroong Choe, Seokhyeon Park, Seokweon Jung, Hyeok Kim, Ji Won Yang, Hwajung Hong, and Jinwook Seo. 2024. Supporting Novice Researchers to Write Literature Review using Language Models. In Extended Abstracts of the 2024 CHI Conference on Human Factors in Computing Systems (CHI EA '24). Association for Computing Machinery, New York, NY, USA, Article 307, 9 pages. https://doi.org/10.1145/3613905.3650787 sourceNot checked — no full text of the source
• cites [122] 122. Nikhil Singh, Guillermo Bernal, Daria Savchenko, and Elena L. Glassman. 2023. Where to Hide a Stolen Elephant: Leaps in Creative Writing with Multimodal Machine Intelligence. ACM Trans. Comput.-Hum. Interact. 30, 5, Article 68 (Sept. 2023), 57 pages. https://doi.org/10.1145/3511599 sourcePartly supported

  Section 6.2.1 describes Editor-Red providing words, lines, and ideas that helped participants continue writing; Section 6.3 describes participants actively extending and altering suggestions to fit their own text. The paper studies creative writing, however, and does not establish support for academic drafting or restructuring ideas from point form into prose. (LLM, conf=0.86)

  Experimental — accuracy not yet measured
• cites [124] 124. Sangho Suh, Meng Chen, Bryan Min, Toby Jia-Jun Li, and Haijun Xia. 2024. Luminate: Structured Generation and Exploration of Design Space with Large Language Models for Human-AI Co-Creation. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems (Honolulu, HI, USA) (CHI '24). Association for Computing Machinery, New York, NY, USA, Article 644, 26 pages. https://doi.org/10.1145/3613904.3642400 sourcePartly supported

  Section 4.2.4 (p. 11) describes a writer selecting AI-generated ideas and text, then editing them into a story draft; Section 6.1 (p. 12) likewise describes a participant editing and merging generated paragraphs. These passages support draft-making and user integration in creative writing, but do not establish translating support in an academic context, conversion from point form to prose, or an explicit encouragement to write ideas independently. (LLM, conf=0.82)

  Experimental — accuracy not yet measured

Probably not a substantial claim: may not pass S3 (load-bearing).

The codebook thematic analysis developed initial codes from Flower & Hayes' cognitive process model and Lee et al.'s design space framework.

“We employed a codebook thematic analysis approach, developing our initial codes from Flower & Hayes' cognitive process model [35] and Lee et al.'s design space framework [73].”

• cites [35] 35. 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 Interaction 31, 2 (2024), 1-40.Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites [73] 73. Philippe Laban, Jesse Vig, Marti Hearst, Caiming Xiong, and Chien-Sheng Wu. 2024. Beyond the chat: Executable and verifiable text-editing with llms. In Proceedings of the 37th Annual ACM Symposium on User Interface Software and Technology. 1-23. sourceCould not verify

  The cited paper’s A.6 Study 1: Annotation Detail describes generating initial annotation labels with GPT-4 and manually reviewing them, but does not describe a codebook thematic analysis or codes derived from Flower and Hayes or Lee et al. Searches for those names, “codebook,” and “thematic analysis” found no passage addressing the claimed method. The citing sentence describes the citing authors’ own analysis rather than clearly attributing that method to this paper. (LLM, conf=0.76)

  Experimental — accuracy not yet measured

May combine two claims.

Dhillon et al.'s research demonstrated that while next-paragraph suggestions improved writing quality, longer AI text completions decreased satisfaction by undermining writers' independence.

“Similarly, Dhillon et al. 's [29] research demonstrated that while next-paragraph suggestions improved writing quality, longer AI text completions decreased satisfaction by undermining writers' independence.”

• cites [29] 29. Ali Darvishi, Hassan Khosravi, Solmaz Abdi, Shazia Sadiq, and Dragan Gašević. 2022. Incorporating Training, Self-monitoring and AI-Assistance to Improve Peer Feedback Quality. In Proceedings of the Ninth ACM Conference on Learning @ Scale (New York City, NY, USA) (L@S '22). Association for Computing Machinery, New York, NY, USA, 35-47. https://doi.org/10.1145/3491140.3528265 sourceCould not verify

  The AI-assistance and RQ2 sections discuss improving student peer-review feedback, not next-paragraph suggestions or longer AI text completions. Searches for Dhillon, text completions, writer independence, and satisfaction found no passage addressing the attributed finding. Page numbers were not available in our copy. (LLM, conf=0.78)

  Experimental — accuracy not yet measured

Classified as probably not a substantial claim.

The review focused on papers published between 2018-2024, corresponding to the emergence and widespread adoption of transformer-based language models.

“Our review focused on papers published between 2018-2024, corresponding to the emergence and widespread adoption of transformer-based language models [129].”

• cites [129] 129. Jakob Tholander and Martin Jonsson. 2023. Design Ideation with AI -Sketching, Thinking and Talking with Generative Machine Learning Models. In Proceedings of the 2023 ACM Designing Interactive Systems Conference (Pittsburgh, PA, USA) (DIS '23). Association for Computing Machinery, New York, NY, USA, 1930-1940. https: //doi.org/10.1145/3563657.3596014 sourcePartly supported

  The Introduction (page unavailable) describes the rapid development of generative AI and large language models, including GPT-3 and ChatGPT. It does not establish widespread adoption, identify a transformer-based timeline, or support 2018–2024 as the period for those developments. The review’s publication-date range is the citing authors’ own choice. (LLM, conf=0.83)

  Experimental — accuracy not yet measured

May combine two claims.

In the Formal context, translating support is focused on writing efficiency, enabling the AI to write in the same interaction location as the user, or to make suggestions that are integrated directly in the text.

“Translating support is focused on writing efficiency, enabling the AI to write in the same interaction location as the user, or to make suggestions that are integrated directly in the text [8,36].”

• cites [8] 8. Karim Benharrak, Tim Zindulka, Florian Lehmann, Hendrik Heuer, and Daniel Buschek. 2024. Writer-Defined AI Personas for On-Demand Feedback Generation. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems (Honolulu, HI, USA) (CHI '24). Association for Computing Machinery, New York, NY, USA, Article 1049, 18 pages. https://doi.org/10.1145/3613904.3642406 sourcePartly supported

  The Conclusion (p. 13) identifies direct editor integration as a benefit of AI feedback, and Section 3.2 (p. 4) describes requesting feedback on selected text within the editor. But the feedback appears in a sidebar comment card; the paper does not describe the AI writing alongside the user or inserting suggestions directly into the text, and Section 6.5.1 (p. 12) frames its focus as the writer-reader relationship rather than writing efficiency. (LLM, conf=0.82)

  Experimental — accuracy not yet measured
• cites [36] 36. Linda Flower and John R. Hayes. 1981. A Cognitive Process Theory of Writing. College Composition and Communication 32, 4 (1981), 365-387. http://www.jstor.org/stable/356600 sourceNot checked — no full text of the source

May combine two claims.

In the General context, systems commonly provided suggestions for revisions which could be integrated directly into the writing area, aiding efficiency and idea exploration in the text.

“Systems commonly provided suggestions for revisions which could be integrated directly into the writing area [14,18,75], aiding efficiency and idea exploration in the text.”

• cites [14] 14. Oğuz 'Oz' Buruk. 2023. Academic Writing with GPT-3.5 (ChatGPT): Reflections on Practices, Efficacy and Transparency. In Proceedings of the 26th International Academic Mindtrek Conference (Tampere, Finland) (Mindtrek '23). Association for Computing Machinery, New York, NY, USA, 144-153. https://doi.org/10.1145/3616961.3616992 sourcePartly supported

  Section 4.2, “Research Buddy” (page unavailable), says GPT-3.5 can suggest ideas and help writers explore them efficiently; Section 4.1 describes using it to produce text quickly. But the paper describes one author’s use of GPT-3.5, not systems commonly offering revision suggestions that can be integrated directly into a writing area. Section 5.6 proposes GPT-supported text editors as a future tool rather than reporting that feature in existing systems. (LLM, conf=0.84)

  Experimental — accuracy not yet measured
• cites [18] 18. Tsung-Shu Chang, Yitong Li, Hui-Wen Huang, and Beth Whitfield. 2021. Exploring EFL Students' Writing Performance and Their Acceptance of AI-based Automated Writing Feedback. In Proceedings of the 2021 2nd International Conference on Education Development and Studies (Hilo, HI, USA) (ICEDS '21). Association for Computing Machinery, New York, NY, USA, 31-35. https://doi.org/10.1145/3459043.3459065 sourceNot checked — no full text of the source
• cites [75] 75. Mina Lee, Percy Liang, and Qian Yang. 2022. Coauthor: Designing a human-ai collaborative writing dataset for exploring language model capabilities. In Proceedings of the 2022 CHI conference on human factors in computing systems. 1-19. sourcePartly supported

  Section 4.2.3 says accepted suggestions were automatically appended to the current text, and sections 5.1.2 and 5.2.1 discuss idea generation and potential productivity benefits. The paper describes suggestions for continuing writing, not specifically revisions to existing text, and its account of CoAuthor does not establish that systems commonly use this design. (LLM, conf=0.85)

  Experimental — accuracy not yet measured

May combine two claims.

In the Creative context, support for reviewing features a mix of quantitative and qualitative feedback, with a focus on the AI evaluating text and providing suggestions rather than revising it directly.

“Support for reviewing features a mix of quantitative and qualitative feedback, with a focus on the AI evaluating text and providing suggestions rather than revising it directly [20,126].”

• cites [20] 20. Yixin Cheng, Kayley Lyons, Guanliang Chen, Dragan Gašević, and Zachari Swiecki. 2024. Evidence-centered Assessment for Writing with Generative AI. In Proceedings of the 14th Learning Analytics and Knowledge Conference (Kyoto, Japan) (LAK '24). Association for Computing Machinery, New York, NY, USA, 178-188. https://doi.org/10. 1145/3636555.3636866 sourcePartly supported

  The ECD Task Model (§2.2; page unavailable) says writers can request AI suggestions and choose whether to accept or modify them. But the Discussion and Conclusions (§5; page unavailable) specifies that these suggestions only continue the current text. The paper does not describe AI reviewing or evaluating text, or providing a mix of quantitative and qualitative feedback. (LLM, conf=0.85)

  Experimental — accuracy not yet measured
• cites [126] 126. Reid Swanson and Andrew Gordon. 2012. Say Anything. ACM Transactions on Interactive Intelligent Systems 2 (09 2012), 1-35. https://doi.org/10.1145/2362394.2362398 sourceNot checked — no full text of the source

May combine two claims.

In the Personal context, planning and translating support are intermingled due to the frequent use of longer AI outputs that directly ideate and write for the user, though these are generally presented as suggestions in order to preserve the user's engagement with the text.

“Planning and translating support are intermingled due to the frequent use of longer AI outputs that directly ideate and write for the user, though these are generally presented as suggestions in order to preserve the user's engagement with the text [15,65].”

• cites [15] 15. Daniel Buschek, Martin Zürn, and Malin Eiband. 2021. The Impact of Multiple Parallel Phrase Suggestions on Email Input and Composition Behaviour of Native and Non-Native English Writers. In Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems (Yokohama, Japan) (CHI '21). Association for Computing Machinery, New York, NY, USA, Article 732, 13 pages. https://doi.org/10.1145/3411764.3445372 sourcePartly supported

  Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  Experimental — accuracy not yet measured
• cites [65] 65. Taewook Kim, Jung Soo Lee, Zhenhui Peng, and Xiaojuan Ma. 2019. Love in lyrics: An exploration of supporting textual manifestation of affection in social messaging. Proceedings of the ACM on Human-Computer Interaction 3, CSCW (2019), 1-27. sourceNot checked — no full text of the source

May combine two claims.

In the Personal context, there is transformation of user inputs between modalities like speech to text or visuals to text, or between textual forms like keywords to prose.

“We also see transformation of user inputs between modalities like speech to text or visuals to text [79,144], or between textual forms like keywords to prose [65].”

• cites [79] 79. Antonios Liapis, Christian Guckelsberger, Jichen Zhu, Casper Harteveld, Simone Kriglstein, Alena Denisova, Jeremy Gow, and Mike Preuss. 2023. Designing for Playfulness in Human-AI Authoring Tools. In Proceedings of the 18th International Conference on the Foundations of Digital Games (Lisbon, Portugal) (FDG '23). Association for Computing Machinery, New York, NY, USA, Article 75, 4 pages. https://doi.org/10.1145/3582437.3587192 sourceCould not verify

  Citation [65] is attached to the claim about transforming keywords into prose. Section 4.1 discusses changing words in a text-to-image prompt, and Section 4.3 discusses playful interpretation of user input, but neither describes keywords being converted into prose. Searches for “keywords” and “prose” found no passage addressing that transformation. (LLM, conf=0.72)

  Experimental — accuracy not yet measured
• cites [144] 144. Ann Yuan, Andy Coenen, Emily Reif, and Daphne Ippolito. 2022. Wordcraft: Story Writing With Large Language Models. In Proceedings of the 27th International Conference on Intelligent User Interfaces (Helsinki, Finland) (IUI '22). Association for Computing Machinery, New York, NY, USA, 841-852. https://doi.org/10.1145/3490099.3511105 sourceSupported
• cites [65] 65. Taewook Kim, Jung Soo Lee, Zhenhui Peng, and Xiaojuan Ma. 2019. Love in lyrics: An exploration of supporting textual manifestation of affection in social messaging. Proceedings of the ACM on Human-Computer Interaction 3, CSCW (2019), 1-27. sourceNot checked — no full text of the source

May combine two claims.

In the Creative context, support for planning takes the form of generative ideation, usually presented as suggestions, although some systems have a more equal and collaborative storytelling focus that weaves the AI ideas into the story text.

“Support for planning takes the form of generative ideation, usually presented as suggestions [74,135], although some systems have a more equal and collaborative storytelling focus that weaves the AI ideas into the story text [70,92].”

• cites [74] 74. Mina Lee, Katy Ilonka Gero, John Joon Young Chung, Simon Buckingham Shum, Vipul Raheja, Hua Shen, Subhashini Venugopalan, Thiemo Wambsganss, David Zhou, Emad A Alghamdi, et al. 2024. A Design Space for Intelligent and Interactive Writing Assistants. In Proceedings of the CHI Conference on Human Factors in Computing Systems. 1-35. sourcePartly supported

  Split into 2 atoms — 1 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.

  Experimental — accuracy not yet measured
• cites [135] 135. Jiyao Wang, Haolong Hu, Zuyuan Wang, Song Yan, Youyu Sheng, and Dengbo He. 2024. Evaluating Large Language Models on Academic Literature Understanding and Review: An Empirical Study among Early-stage Scholars. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems (Honolulu, HI, USA) (CHI '24). Association for Computing Machinery, New York, NY, USA, Article 12, 18 pages. https://doi.org/10.1145/3613904. 3641917 sourceNot checked — no full text of the source
• cites [70] 70. Charlotte Kobiella, Yarhy Said Flores López, Franz Waltenberger, Fiona Draxler, and Albrecht Schmidt. 2024. "If the Machine Is As Good As Me, Then What Use Am I?" -How the Use of ChatGPT Changes Young Professionals' Perception of Productivity and Accomplishment. In Proceedings of the CHI Conference on Human Factors in Computing Systems. ACM, Honolulu HI USA, 1-16. https://doi.org/10.1145/3613904.3641964 sourcePartly supported

  Split into 2 atoms — 0 confirmed, 1 partially supported, 0 disputed, 0 misattributed, 1 unverifiable. Expand the claim to see per-atom verdicts.

  Experimental — accuracy not yet measured

May combine two claims.

Some systems present suggestions directly in the user's workspace, creating a more integrated experience but potentially blurring boundaries between user and AI contributions.

“Some systems present suggestions directly in the user's workspace [8,14,18], creating a more integrated experience but potentially blurring boundaries between user and AI contributions.”

• cites [8] 8. Karim Benharrak, Tim Zindulka, Florian Lehmann, Hendrik Heuer, and Daniel Buschek. 2024. Writer-Defined AI Personas for On-Demand Feedback Generation. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems (Honolulu, HI, USA) (CHI '24). Association for Computing Machinery, New York, NY, USA, Article 1049, 18 pages. https://doi.org/10.1145/3613904.3642406 sourceSupported
• cites [14] 14. Oğuz 'Oz' Buruk. 2023. Academic Writing with GPT-3.5 (ChatGPT): Reflections on Practices, Efficacy and Transparency. In Proceedings of the 26th International Academic Mindtrek Conference (Tampere, Finland) (Mindtrek '23). Association for Computing Machinery, New York, NY, USA, 144-153. https://doi.org/10.1145/3616961.3616992 sourcePartly supported

  Section 5.4 (page not provided) describes uncertainty about the author's agency over GPT-generated text, supporting the concern about blurred contributions. Section 5.6 proposes GPT-supported text editors that show prompts, outputs, and edits, but does not say that existing systems present suggestions directly in the user's workspace or that this creates a more integrated experience. (LLM, conf=0.82)

  Experimental — accuracy not yet measured
• cites [18] 18. Tsung-Shu Chang, Yitong Li, Hui-Wen Huang, and Beth Whitfield. 2021. Exploring EFL Students' Writing Performance and Their Acceptance of AI-based Automated Writing Feedback. In Proceedings of the 2021 2nd International Conference on Education Development and Studies (Hilo, HI, USA) (ICEDS '21). Association for Computing Machinery, New York, NY, USA, 31-35. https://doi.org/10.1145/3459043.3459065 sourceNot checked — no full text of the source

May combine two claims.

Some empirical studies had experimental conditions where the user had no influence over AI-generated text, which was associated with a reduction in perceived ownership.

“However, some empirical studies [34] did have experimental conditions where the user had no influence over AI-generated text-which was associated with a reduction in perceived ownership.”

• cites [34] 34. 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 Trans. Comput.-Hum. Interact. 31, 2, Article 25 (Feb. 2024), 40 pages. https://doi.org/10.1145/3637875 sourceSupported

Probably not a substantial claim: may not pass S3 (load-bearing).

The theoretical framework may not fully describe Human-AI interaction in creative and professional writing.

“While the framework is widely used in AI writing research ( [9,73,101]) and provided a valuable lens for this study, it may not fully describe Human-AI interaction in creative and professional writing.”

• cites [9] 9. Advait Bhat, Saaket Agashe, Parth Oberoi, Niharika Mohile, Ravi Jangir, and Anirudha Joshi. 2023. Interacting with Next-Phrase Suggestions: How Suggestion Systems Aid and Influence the Cognitive Processes of Writing. In Proceedings of the 28th International Conference on Intelligent User Interfaces (Sydney, NSW, Australia) (IUI '23). Association for Computing Machinery, New York, NY, USA, 436-452. https://doi.org/10.1145/3581641.3584060 sourcePartly supported

  Section 5.1 says the Hayes writing model has no clear place for suggestion systems and proposes an extension to describe writer–suggestion interaction. That supports the framework-limitation point, but the paper grounds it in a study of next-phrase suggestions for movie reviews, not creative and professional writing broadly. Page numbers are unavailable in our copy. (LLM, conf=0.85)

  Experimental — accuracy not yet measured
• cites [73] 73. Philippe Laban, Jesse Vig, Marti Hearst, Caiming Xiong, and Chien-Sheng Wu. 2024. Beyond the chat: Executable and verifiable text-editing with llms. In Proceedings of the 37th Annual ACM Symposium on User Interface Software and Technology. 1-23. sourceCould not verify

  The citation appears after the statement that a theoretical framework is widely used in AI-writing research; the later suggestion that it may not fully describe creative and professional writing is the citing authors’ own qualification. In §4.3, this paper describes its Warn-Verify-Audit framework for checking edits, but the passages found do not establish use of the unspecified theoretical framework. Searches for “theoretical framework” and “framework” did not settle the attributed point. (LLM, conf=0.75)

  Experimental — accuracy not yet measured
• cites [101] 101. Naveen Ram, Tanay Gummadi, Rahul Bhethanabotla, Richard J Savery, and Gil Weinberg. 2021. Say What? Collabora- tive Pop Lyric Generation Using Multitask Transfer Learning. In Proceedings of the 9th International Conference on Human-Agent Interaction (Virtual Event, Japan) (HAI '21). Association for Computing Machinery, New York, NY, USA, 165-173. https://doi.org/10.1145/3472307.3484175 sourceCould not verify

  The citation is attached to the statement that a framework is widely used in AI writing research; the later qualification about its limits appears to be the citing authors’ own assessment. The Related Work section (p. 2) discusses lyric generation and language models, while the Discussion (p. 8) discusses human–AI collaboration. Neither passage identifies the theoretical framework or establishes its use across AI writing research (searched: theoretical framework, framework widely used). (LLM, conf=0.72)

  Experimental — accuracy not yet measured

Classified as probably not a substantial claim.

Shneiderman posits a multidimensional perspective where automation and control can increase concurrently.

“Instead, Shneiderman posits a multidimensional perspective where automation and control can increase concurrently”

• cites [118] 118. Donghoon Shin, Subeen Park, Esther Hehsun Kim, Soomin Kim, Jinwook Seo, and Hwajung Hong. 2022. Exploring the Effects of AI-assisted Emotional Support Processes in Online Mental Health Community. In Extended Abstracts of the 2022 CHI Conference on Human Factors in Computing Systems (New Orleans, LA, USA) (CHI EA '22). Association for Computing Machinery, New York, NY, USA, Article 300, 7 pages. https://doi.org/10.1145/3491101.3519854 sourceCould not verify

  Section 2.2 (pp. 2–3) discusses AI-mediated communication and AI assistance in an online mental health community, but not Shneiderman’s perspective on automation and control. Searches for Shneiderman and for concurrent increases in automation and human control found no passage addressing the claim. (LLM, conf=0.76)

  Experimental — accuracy not yet measured

Probably not a substantial claim: may not pass S3 (load-bearing).

Shneiderman's HAI framework argues that automation and human control need not be at odds on a unidimensional spectrum.

“such as Ben Shneiderman's HAI framework, which argues that automation and human control need not be at odds on a unidimensional spectrum [118]”

• cites [118] 118. Donghoon Shin, Subeen Park, Esther Hehsun Kim, Soomin Kim, Jinwook Seo, and Hwajung Hong. 2022. Exploring the Effects of AI-assisted Emotional Support Processes in Online Mental Health Community. In Extended Abstracts of the 2022 CHI Conference on Human Factors in Computing Systems (New Orleans, LA, USA) (CHI EA '22). Association for Computing Machinery, New York, NY, USA, Article 300, 7 pages. https://doi.org/10.1145/3491101.3519854 sourceCould not verify

  The AI-mediated Communication section (pp. 2–3) describes AI-assisted interpersonal communication, and the Conclusion (p. 6) describes the paper’s emotional-support system. Neither addresses Shneiderman’s HAI framework or the claim that automation and human control are not opposites on a single spectrum. Searches for those terms and for independent dimensions of automation and human control found no supporting passage. (LLM, conf=0.76)

  Experimental — accuracy not yet measured

Probably not a substantial claim: may not pass S4 (based).

Content-centric writers focus more on planning and ideation.

“content-centric writers focus more on planning and ideation”

May combine two claims.

The synthesis revealed encouraging alignments as well as notable gaps between current AI writing support systems and writers' expressed needs.

“We then synthesized findings from both studies, revealing encouraging alignments as well as notable gaps between current AI writing support systems and writers' expressed needs.”

Probably not a substantial claim: may not pass S4 (based). May combine two claims.

Monitoring is a crucial yet underexplored cognitive process in AI-assisted writing.

“Monitoring is a crucial yet underexplored cognitive process in AI-assisted writing”

Probably not a substantial claim: may not pass S3 (load-bearing).

The initial search yielded 1,676 papers.

“Our initial search yielded 1,676 papers.”

Classified as probably not a substantial claim. May combine two claims.

The full-text review process resulted in the exclusion of 110 papers: 71 for not providing co-writing support, 10 for purely technical contributions, 15 for non-natural language output, and 14 for not meeting paper type criteria.

“This process resulted in the exclusion of 110 papers: 71 for not providing co-writing support (criterion 1), 10 for purely technical contributions (criterion 2), 15 for non-natural language output (criterion 3), and 14 for not meeting our paper type criteria (criterion 4).”

May combine two claims.

Research attention is uneven across cognitive processes and contexts, with strong representation of Generating and Translating activities in Creative and Academic settings and limited focus on Monitoring across all contexts.

“The table reveals uneven research attention across cognitive processes and contexts-for example, strong representation of Generating and Translating activities in Creative and Academic settings, and limited focus on Monitoring across all contexts.”

May combine two claims.

The semi-structured interviews with 15 writers characterized key themes describing what writers want to own when ownership matters to them and how AI interactions shape ownership during writing.

“Characterized key themes describing what writers want to own when ownership matters to them and how AI interactions shape ownership during writing.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The initial screening identified 219 papers for full-text review.

“One researcher conducted the initial screening, applying our exclusion criteria to titles and abstracts, which identified 219 papers for full-text review.”

May combine two claims.

S3 (Active Co-Writing) was the most commonly deployed approach (23 systems, 37.1%), followed by S1 (Structured Guidance) (19 systems, 30.6%), S2 (Guided Exploration) (11 systems, 17.7%), and S4 (Critical Feedback) (9 systems, 14.5%).

“This coding revealed that S3 (Active Co-Writing) was the most commonly deployed approach (23 systems, 37.1%), followed by S1 (Structured Guidance) (19 systems, 30.6%),S2 (Guided Exploration) (11 systems, 17.7%), and S4 (Critical Feedback) (9 systems, 14.5%), with the full distribution shown in Table 2.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The distribution of ratings reflects a range of perspectives on the extent to which writers want AI to intervene across different processes.

“The distribution of ratings reflects a range of perspectives on the extent to which writers want AI to intervene across different processes.”

May combine two claims.

Academic writing favored S1 (61.1%), Creative writing employed S3 (50%), and Formal writing preferred S2 (42.9%).

“Context-specific preferences were also evident, with Academic writing favoring S1 (61.1%),Creative writing employing S3 (50%), and Formal writing preferring S2 (42.9%).”

May combine two claims.

Such tasks tend to be perceived by writers as more functional than creative, making AI tools more acceptable for generating content without affecting their sense of ownership.

“Such tasks tend to be perceived by writers as more functional than creative, making AI tools more acceptable for generating content without affecting their sense of ownership.”

Classified as probably not a substantial claim.

The four strategies are distinguished by the AI's role, intended user behaviors, interaction outcomes, interface design, and usage of AI outputs.

“These strategies are distinguished by the AI's role, intended user behaviors, interaction outcomes, interface design, and usage of AI outputs.”

May combine two claims.

When task delegation is done this way, writers' sense of autonomy and joy in creating something novel is maintained-or even enhanced in cases where the AI frees up their focus.

“When task delegation is done this way, writers' sense of autonomy and joy in creating something novel is maintained-or even enhanced in cases where the AI frees up their focus.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The ability to critically evaluate AI output helped writers like W10 maintain a sense of separation between looking at AI output and feeling like it inadvertently influenced them.

“The ability to critically evaluate AI output helped writers like W10 maintain a sense of separation between looking at AI output and feeling like it inadvertently influenced them.”

May combine two claims.

S1 dominated Evaluating (54.5%) and S3 led in Generating (52.9%) and Translating (50%) processes.

“Clear patterns emerged across writing processes, with S1 dominating Evaluating (54.5%) and S3 leading in Generating (52.9%) and Translating (50%) processes.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

While ideas and themes can feel culturally shared, the form in which those ideas are presented is where the writer's individuality comes through.

“They further noted that while ideas and themes can feel culturally shared, the form in which those ideas are presented is where the writer's individuality comes through”

May combine two claims.

The content-form distinction connects writing contexts with cognitive processes: academic and non-fiction writers prioritized content to convey ideas clearly, while fiction writers emphasized form, valuing their unique voice and style.

“This content-form distinction connects writing contexts with cognitive processes: academic and non-fiction writers prioritized content to convey ideas clearly, while fiction writers emphasized form, valuing their unique voice and style.”

Probably not a substantial claim: may not pass S3 (load-bearing).

W3 uses AI to generate writing prompts and background settings, allowing them to focus on what they consider the core writing process.

“They describe using AI to generate writing prompts and background settings, allowing them to focus on what they consider the core writing process”

Probably not a substantial claim: may not pass S3 (load-bearing).

In contrast, when writers have freedom over form, the idea of AI models intervening in the translation process becomes less appealing.

“In contrast, when writers have freedom over form, the idea of AI models intervening in the translation process becomes less appealing.”

May combine two claims.

When the ideas are established, writers are open to using AI tools for translating ideas into clear language or reviewing their work to enhance clarity and polish.

“When the ideas are established, writers are open to using AI tools for translating ideas into clear language or reviewing their work to enhance clarity and polish.”

May combine two claims.

Writers feel a stronger sense of ownership when they perceive themselves as having substantial control over the AI's contributions, and interaction design can shape these perceptions.

“We find a common theme across these interactions: writers feel a stronger sense of ownership when they perceive themselves as having substantial control over the AI's contributions, and that interaction design can shape these perceptions.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The theme highlights situations where the risk of writers' overreliance on AI is particularly prevalent.

“It also highlights situations where the risk of writers' overreliance on AI is particularly prevalent.”

May combine two claims.

The authors identified two main types of contributions-content and form-each linked to specific cognitive processes.

“We identified two main types of contributions-content and form-each linked to specific cognitive processes.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The theme 'How AI Interactions Shape Ownership' explores how interaction design impacts writers' senses of agency and ownership.

“How AI Interactions Shape Ownership: This theme explores how interaction design impacts writers' senses of agency and ownership.”

Classified as probably not a substantial claim.

Writers feel it is important that decisions are primarily made by the writer and not by the AI.

“so it's important that those decisions are primarily made by you and not by the AI.”

Classified as probably not a substantial claim. May combine two claims.

Content contributions involve generating ideas and setting goals, aligning with the planning process in the Flower and Hayes model.

“Content contributions involve generating ideas and setting goals, aligning with the planning process in the Flower and Hayes model [35].”

• cites [35] 35. 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 Interaction 31, 2 (2024), 1-40.Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

Form contributions focus on style, tone, and flow, aligning with translation and revision.

“Form contributions focus on style, tone, and flow, aligning with translation and revision.”

Probably not a substantial claim: may not pass S3 (load-bearing).

For W7, intentional rule-breaking was a distinctive aspect of their voice.

“For W7, intentional rule-breaking was a distinctive aspect of their voice.”

Classified as probably not a substantial claim.

When the purpose of a piece of writing is to convey pre-existing ideas or information with clarity-as is common in academic and non-fiction writing-the content itself becomes the primary contribution and the central focus of ownership for writers.

“When the purpose of a piece of writing is to convey pre-existing ideas or information with clarity-as is common in academic and non-fiction writing, the content itself Legend: Pink -Little Support Yellow -Moderate Support Blue -Significant Support becomes the primary contribution and the central focus of ownership for writers.”

Classified as probably not a substantial claim.

The theme 'What Writers Want to Own' characterizes the aspects of the composition process and product from which writers derive their sense of ownership and prioritize as their primary contribution.

“What Writers Want to Own: This theme characterizes the aspects of the composition process and product from which writers derive their sense of ownership and prioritize as their primary contribution.”

May combine two claims.

S1 and S2 systems represented 61% of systems in Academic contexts and 57% in Formal contexts.

“As shown in Table 2, S1 and S2 systems represented 61% of systems in Academic contexts and 57% in Formal contexts”

Probably not a substantial claim: may not pass S3 (load-bearing).

Even if language models were trained to mirror an author's personal style, form-oriented writers could find this prospect wholly unappealing.

“Even if the language models were trained to mirror an author's personal style, form-oriented writers could find this prospect wholly unappealing”

May combine two claims.

In these writing contexts, writers derive a sense of ownership by engaging in cognitive processes involved in ideation and organization of ideas, making planning the dominant process that they seek to control.

“In these writing contexts, writers derive a sense of ownership by engaging in cognitive processes involved in ideation and organization of ideas, making planning the dominant process that they seek to control.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The desire to emphasize ownership over content rather than form is influenced by external constraints, such as word counts or stylistic conventions.

“The desire to emphasize ownership over content rather than form is influenced by external constraints, such as word counts or stylistic conventions.”

May combine two claims.

Form-centric contexts emphasize ownership over translation and revision while being more open to AI assistance with planning and ideation.

“Form-centric contexts emphasize ownership over translation and revision while being more open to AI assistance with planning and ideation.”

Classified as probably not a substantial claim.

AI suggestions (as opposed to direct insertion) do not change writers' sense of ownership.

“If it's sort of making suggestions, then it would not change my sense of ownership over the text, because I'd still feel like that's just sort of this pop-up window.”

Probably not a substantial claim: may not pass S3 (load-bearing).

For them, style, rhythm, and structure are the personal touches they are not willing to delegate to AI or any other external influence.

“For them, style, rhythm, and structure are the personal touches they are not willing to delegate to AI or any other external influence.”

May combine two claims.

Content-oriented writers derive their ownership from generating original ideas and prioritize information structure over style.

“These writers derive their ownership from generating original ideas, and prioritize information structure over style.”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

S3 facilitates efficient collaboration while maintaining user control.

“S3 Active Co-Writing facilitates efficient collaboration while maintaining user control”

May combine two claims.

Although participants had experience with a variety of AI writing tools, all had used ChatGPT, with fewer using alternatives.

“Additionally, although our participants had experience with a variety of AI writing tools, all had used ChatGPT, with fewer using alternatives.”

May combine two claims.

S1 supports users by providing step-by-step AI-assistance while fostering skill development.

“S1: Structured Guidance supports users by providing step-by-step AI-assistance while fostering skill development”

Probably not a substantial claim: may not pass S3 (load-bearing).

S4 promotes strategies that facilitate user reflection and engagement through analysis and feedback.

“S4: Critical Feedback promotes strategies that facilitate user reflection and engagement through analysis and feedback”

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

S2 enables creative control through systematic exploration.

“S2: Guided Exploration enables creative control through systematic exploration”

May combine two claims.

Research attention is uneven across processes and contexts, with strong representation of AI support during the Generating subprocess within the Planning phase in Creative and Academic writing contexts, but limited focus on Monitoring support across all writing contexts.

“This mapping revealed uneven research attention across processes and contexts (shown in Table 1), such as strong representation of AI support during the Generating subprocess within the Planning phase in Creative and Academic writing contexts, but limited focus on Monitoring support across all writing contexts.”

Probably not a substantial claim: may not pass S4 (based).

Studies on suggestions do not allow participants to disable AI assistance entirely.

“For instance, studies on suggestions do not allow participants to disable AI assistance entirely.”

Probably not a substantial claim: may not pass S3 (load-bearing).

There are two primary contribution types: content and form.

“The second theme, covered in Section 4.2.2, distinguishes between the two primary contribution types: content and form.”

May combine two claims.

Gero et al.'s research revealed that the idea generation phase can particularly threaten ownership, with some writers considering the struggle with writer's block as integral to their writerly identity.

“Gero et al.'s research [39] revealed that the idea generation phase can particularly threaten ownership, with some writers considering the struggle with writer's block as integral to their writerly identity.”

• cites [39] 39. Katy Ilonka Gero, Vivian Liu, and Lydia Chilton. 2022. Sparks: Inspiration for Science Writing using Language Models. In Proceedings of the 2022 ACM Designing Interactive Systems Conference (Virtual Event, Australia) (DIS '22). Association for Computing Machinery, New York, NY, USA, 1002-1019. https://doi.org/10.1145/3532106.3533533 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S2 (checkable).

Maintaining user agency and ownership requires careful consideration of interaction design, user control mechanisms, and the balance between AI support and user autonomy.

“These findings indicate that maintaining user agency and ownership requires careful consideration of interaction design, user control mechanisms, and the balance between AI support and user autonomy.”

• AssumptionRests on an assumption: Writers must maintain agency over the cognitive processes they value most through careful task delegation in order to preserve their sense of ownership over the final text.

Probably not a substantial claim: may not pass S3 (load-bearing).

The Flower and Hayes (1981) model identifies four primary cognitive processes in writing: planning, translating, reviewing, and monitoring.

“Their model identifies four primary processes: planning”

• Also stated in: § 2.1 Theories on Writing Processes · plain

  “translating (i.e., transforming structured information into linear prose), reviewing (i.e., evaluating and revising text according to established goals), and monitoring (i.e., overseeing, regulating, and coordinating the writer's cognitive activities, such as deciding when to shift between planning, translating, and reviewing, and ensuring alignment with writing goals).”

• cites [35] 35. 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 Interaction 31, 2 (2024), 1-40.Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S2 (checkable).

Compared to other writing process theories, the Flower and Hayes model maintains an optimal balance between sophistication and analytical utility for our specific research context.

“Compared to other writing process theories, the Flower and Hayes model maintains an optimal balance between sophistication and analytical utility for our specific research context.”

• cites [35] 35. 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 Interaction 31, 2 (2024), 1-40.Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

S3 (Active Co-Writing) may challenge writers' sense of agency by blurring task delegation boundaries.

“However, unlike Proposal Integration, suggestions are integrated directly in the final text which may challenge writers' sense of agency by blurring task delegation boundaries.”

Classified as probably not a substantial claim.

Prior experience with AI writing tools was an inclusion criterion.

“Given our focus on AI-assisted writing, prior experience with AI writing tools was an inclusion criterion.”

Classified as probably not a substantial claim. May combine two claims.

The authors used the main writing processes-planning, translation, reviewing, and monitoring-as predefined codes to structure their interpretation, while also inductively identifying new patterns.

“Guided by the cognitive process theory of writing, we used the main writing processes-planning, translation, reviewing, and monitoring-as predefined codes to structure our interpretation, while also inductively identifying new patterns.”

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

The application of the framework highlights opportunities for more nuanced strategy implementation across cognitive processes.

“While the single-coder approach represents a limitation, this application of our framework highlights opportunities for more nuanced strategy implementation across cognitive processes.”

Classified as probably not a substantial claim.

The authors conducted a reflexive thematic analysis of transcripts through an inductive-deductive approach.

“To identify factors that influence writers' sense of ownership in the AI-assisted writing process, we conducted a reflexive thematic analysis [12] of transcripts through an inductive-deductive approach.”

• cites [12] 12. Kyle Booten and Katy Ilonka Gero. 2021. Poetry Machines: Eliciting Designs for Interactive Writing Tools from Poets. In Proceedings of the 13th Conference on Creativity and Cognition (Virtual Event, Italy) (C&C '21). Association for Computing Machinery, New York, NY, USA, Article 51, 5 pages. https://doi.org/10.1145/3450741.3466813 sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing).

Scores for negatively worded items (P1, P3, T1, R1, and M1) were reversed.

“Scores for negatively worded items (P1, P3, T1, R1, and M1) were reversed (see Figure 4).”

Classified as probably not a substantial claim. May combine two claims.

The goal was not to aggregate results into a single measure of ownership and agency but rather to examine distinct aspects of the writing process.

“Given the varied perspectives on ownership across writing elements, our goal was not to aggregate results into a single measure of ownership and agency but rather to examine distinct aspects of the writing process.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The single-coder approach represents a limitation.

“While the single-coder approach represents a limitation, this application of our framework highlights opportunities for more nuanced strategy implementation across cognitive processes.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The institution's research ethics board approved the study protocol.

“The institution's research ethics board approved the study protocol.”

May combine two claims.

Creative writing showed surprisingly high deployment of S3 systems, which are the most likely to threaten ownership.

“Creative writing showed surprisingly high deployment of S3 systems, which are the most likely to threaten ownership.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

This selective reliance on AI allows writers to focus their energy and sense of ownership on other parts of the work, while using AI to polish weaker areas.

“This selective reliance on AI allows writers to focus their energy and sense of ownership on other parts of the work, while using AI to polish weaker areas.”

Classified as probably not a substantial claim.

The theme highlights the critical role that Human-AI interaction design can play in maintaining writers' sense of agency and ownership in AI-assisted workflows.

“This theme highlights the critical role that Human-AI interaction design can play in maintaining writers' sense of agency and ownership in AI-assisted workflows.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Tang et al. found differential usage patterns between professional and non-professional users of AI tools.

“Perceived competence differences have been documented by Tang et al. [127], who found differential usage patterns between professional and non-professional users of AI tools”

• cites Tang et al. [127]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

Cao et al. investigated how AI support systems can mitigate the effects of time pressure on decision-making.

“and by Cao et al. [16], who investigated how AI support systems can mitigate the effects of time pressure on decision-making.”

• cites Cao et al. [16]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S3 (load-bearing).

Shakeri et al. addressed users' vulnerability from lack of confidence in creative writing contexts.

“and by Shakeri et al., who addressed users' vulnerability from lack of confidence in creative writing contexts.”

• cites Shakeri et al.Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

May combine two claims.

The theme showcases the flexible role that human agency plays in AI-assisted writing and how it responds to situational factors.

“Instead of assuming the desirability of ownership as an inherent or static prerequisite, this theme showcases the flexible role that human agency plays in AI-assisted writing and how it responds to situational factors.”

Classified as probably not a substantial claim.

AI interaction design plays a role in supporting writers with monitoring and decision-making.

“They also point to the role that AI interaction design plays in supporting writers with monitoring and decision-making.”

Probably not a substantial claim: may not pass S2 (checkable).

The three overarching themes offer a way to grapple with the complex interplay between writer agency, task demands, and AI functionality, helping AI system designers make sense of how ownership is negotiated and maintained in AI-assisted writing.

“Together, these three overarching themes offer a way to grapple with the complex interplay between writer agency, task demands, and AI functionality, helping AI system designers make sense of how ownership is negotiated and maintained in AI-assisted writing.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

W3's phrasing indicates a view of ideation as a necessary setup that can be delegated to AI, while the actual crafting of sentences is where they feel personal investment and ownership.

“Their phrasing indicates a view of ideation as a necessary setup that can be delegated to AI, while the actual crafting of sentences is where they feel personal investment and ownership.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Zhang et al.'s multi-level framework enables users to choose between different levels of AI creative involvement based on their individualized needs.

“Task importance considerations appear in Zhang et al. 's [143] multi-level framework, which enables users to choose between different levels of AI creative involvement based on their individualized needs”

• cites Zhang et al. [143]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Classified as probably not a substantial claim.

These factors represent a user's personal value-based context or external limitations that shape their willingness to delegate writing tasks to AI.

“These factors represent a user's personal value-based context or external limitations that shape their willingness to delegate writing tasks to AI.”

Probably not a substantial claim: may not pass S4 (based).

This paper is the first comprehensive study on designing for human agency within AI-assisted writing that combines a systematic review of generative AI-era research with an analysis of writers' perspectives on preserving agency and ownership.

“This paper, to our knowledge, is the first comprehensive study on designing for human agency within AI-assisted writing that combines a systematic review of generative AI-era research with an analysis of writers' perspectives on preserving agency and ownership.”

May combine two claims.

User-initiated AI interactions do not fulfill participants' desire for minimizing distractions or fine-grained control over how the AI interacts with their stylistic choices.

“It was common that systems had user-initiated AI interactions, however this design choice does not fulfill our participants' desire for minimizing distractions or fine-grained control over how the AI interacts with their stylistic choices.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The educational homogeneity of the sample may have influenced participants' perspectives on AI-assisted writing, since individuals with higher educational backgrounds might approach AI tools differently.

“This educational homogeneity may have influenced perspectives on AI-assisted writing, as individuals with higher educational backgrounds might approach AI tools differently than those with other experiences.”

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Gero et al. found that while authenticity and ownership are related, they are not directly correlated-users may not perceive a system that mimics their style as inauthentic.

“Gero et al. [39] found that while authenticity and ownership are related, they are not directly correlated-users may not perceive a system that mimics their style as inauthentic.”

• cites Gero et al. [39] sourceNot checked — no full text of the source

Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

Recent CSCW programs reflect this shift, with dedicated sessions on Human-AI Collaboration and AI and Trust at CSCW 2023 [1], and AI in Creativity Flows and Future Dialogues on Personal AI Assistants at CSCW 2024 [2].

“Recent CSCW programs reflect this shift, with dedicated sessions on Human-AI Collaboration and AI and Trust at CSCW 2023 [1], and AI in Creativity Flows and Future Dialogues on Personal AI Assistants at CSCW 2024 [2].”

• cites [1] 1. Revising [7, 13, 17, 20, 28, 46, 51, 55, 97, 114, 132] [11, 50, 142] [7, 79, 95, 132, 137, 144] [30, 40, 43, 79, 104] Monitoring [103, 130]Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites [2] 2. 2023. CSCW 2023 Program Overview. https://cscw.acm.org/2023/index.php/program-overview/. Accessed: 2025-04- 15.Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S4 (based).

The strategies do not offer guidance on writers' values tied to preserving human agency.

“While the strategies represent current research, they do not offer guidance on writers' values tied to preserving human agency.”

May combine two claims.

Users seek AI assistance in areas outside their primary contribution, suggesting a chessboard-like pattern.

“This view of ownership suggests a 'chessboard-like' pattern, where users seek AI assistance in areas outside their primary contribution.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Our findings are influenced by our choice of theoretical framework.

“Firstly, our findings are influenced by our choice of theoretical framework [35].”

• cites [35] 35. 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 Interaction 31, 2 (2024), 1-40.Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Probably not a substantial claim: may not pass S2 (checkable).

Our approach demonstrates the value of viewing creative tasks as multi-dimensional.

“Like Shneiderman's multi-dimensional characterization of automation and human agency, our approach demonstrates the value of viewing creative tasks as a multi-dimensional.”

Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

These interactions introduce new concerns around agency and ownership.

“Crucially, these interactions introduce new concerns around agency and ownership that our community now need to grapple with.”

• AssumptionRests on an assumption: Recent advances in AI have shifted CSCW's dynamic from human-human collaboration with computers as tools to AI taking semi-autonomous roles.
• AssumptionRests on an assumption: AI systems increasingly take on semi-autonomous roles, and interactions with them begin to mirror human collaboration.

Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

As AI systems advance, monitoring and management of the Human-AI collaborative relationship becomes increasingly important.

“As AI systems advance, monitoring and management of the Human-AI collaborative relationship becomes increasingly important.”

Probably not a substantial claim: may not pass S4 (based).

This high deployment of S3 in Creative writing highlights a potential misalignment between strategy use and user needs, as S3 can compromise the sense of ownership central to writers' creative expression.

“highlighting a potential misalignment between strategy use and user needs, as Active Co-Writing (S3) can compromise the sense of ownership central to writers' creative expression.”

Probably not a substantial claim: may not pass S3 (load-bearing).

The ACM's disciplinary focus may have oriented our findings toward certain perspectives in computing and human-computer interaction, underrepresenting interdisciplinary approaches or perspectives from adjacent fields.

“The ACM's disciplinary focus may have oriented our findings toward certain perspectives in computing and human-computer interaction, underrepresenting interdisciplinary approaches or perspectives from adjacent fields.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Monitoring is connected to the collaborative relationship through process management and a meta-level view of both individual and collaborative writing processes.

“While monitoring as a cognitive process is distinct from the collaborative relationship between human and AI, they are connected through process management and a meta-level view of both the individual and collaborative writing processes.”

Probably not a substantial claim: may not pass S2 (checkable).

The gap in monitoring research likely stems from research focusing on optimizing specific interactions rather than examining broader collaborative dynamics.

“The gap likely stems from research focusing on optimizing specific interactions rather than examining broader collaborative dynamics.”

Probably not a substantial claim: may not pass S2 (checkable).

Breaking down creative tasks into distinct cognitive processes and contexts highlights specific context × process dimensions where designers should focus AI support.

“By breaking it down into distinct cognitive processes and contexts, we move beyond a one-size-fits-all perspective and highlight specific context × process dimensions where designers should focus AI support.”

May combine two claims.

S3 streamlines AI integration but potentially creates tension around authorship of the final product.

“Result Ownership (S3) streamlines AI integration but potentially creates tension around authorship of the final product.”

May combine two claims.

Separation of suggestions creates a deliberate boundary that reinforces the writer's role as decision-maker and aligns with participants' desire for control over final text.

“This separation creates a deliberate boundary that reinforces the writer's role as decision-maker, aligning with our interview participants' desire to maintain control over what enters their final text.”

Probably not a substantial claim: may not pass S2 (checkable).

By aligning system design with the real-world needs of writers, this work lays the foundation for human-centered AI writing tools that enable true co-writing, on human terms.

“By aligning system design with the real-world needs of writers, this work lays the foundation for human-centered AI writing tools that enable true co-writing, on human terms.”