pith. sign in

arxiv: 2605.29090 · v1 · pith:ZSMS2OUJnew · submitted 2026-05-27 · 💻 cs.HC

"It's OK Because...": The Wild West of Student Rationalization of AI Use in Academic Writing

Jiyoon Kim , Kentaro Toyama , Sangmi Kim , John M. Carroll This is my paper

Pith reviewed 2026-06-29 09:55 UTC · model grok-4.3

classification 💻 cs.HC
keywords AI ethicsacademic integritystudent rationalizationgenerative AIwriting assistancethematic analysisAI policy
0
0 comments X

The pith

Students justify AI use in academic writing with over twenty distinct rationalizations that often contradict course policies.

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

The paper investigates how students make sense of generative AI in their writing assignments by conducting semi-structured interviews with twenty students along with reviewing their chat logs and course documents. It identifies five different sites where students form their understanding of acceptable AI use, separate from what faculty intend. Students draw on a taxonomy of more than twenty rationalizations, including the ideas that AI text is victimless to copy, that output matching their own style counts as their writing, and that heavy AI reliance actually improves learning. These justifications appear both before and after use and do not need to be consistent with each other. The findings indicate that current AI tools create an ethical path that leads students well beyond typical pedagogical aims.

Core claim

Analysis of interviews, chat logs, and syllabi shows students employ at least twenty rationalizations to frame AI assistance as acceptable, such as treating copied AI text as victimless, claiming ownership of any output that matches their beliefs or style, and asserting that extensive AI use produces more learning than independent work. These rationalizations operate in ad hoc and post hoc ways and support conscious policy violations in some cases.

What carries the argument

A taxonomy of over twenty student rationalizations for AI use in writing, identified through thematic analysis of interview data and logs.

If this is right

  • Faculty AI policies reach students through multiple conceptual sites that can diverge from original intent.
  • Some rationalizations enable students to violate policies while preserving a sense of ethical consistency.
  • Rationalizations can shift between ad hoc and post hoc application within the same student.
  • Educational interventions must address the full range of justifications rather than single rules.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Course designs could incorporate explicit discussion of common rationalizations to surface them before assignments begin.
  • Similar justification patterns may emerge in non-academic settings where AI assists professional output.
  • Tool interfaces might insert brief prompts at generation points to require users to articulate their intended use.

Load-bearing premise

The rationalizations described in the interviews accurately reflect students' real-time moral reasoning during actual AI use rather than responses shaped by the interview situation.

What would settle it

Direct comparison of students' stated rationalizations in interviews against their unprompted AI usage patterns in submitted assignments that violate the course syllabus.

read the original abstract

Generative AI challenges academic integrity not only by enabling students to delegate substantial portions of their academic work, but also by blurring the ethical boundaries by which students distinguish acceptable assistance from misconduct. Drawing on semi-structured interviews (n=20), AI chat logs, and course documents (syllabi, submitted assignments), we investigated how students themselves make moral sense of AI use in academic writing. Our analysis results in a range of novel findings: First, there are at least five distinct sites of AI-use conceptualization, ranging from faculty's intended AI policy, to students' actual AI use. Second, students use over 20 distinct rationalizations to justify AI use, such as that copying AI-generated text is victimless; that any AI text reflecting their own beliefs or their own style is their own writing; or that they are learning more by using AI -- even extensively -- than otherwise. We present a taxonomy of these rationalizations, and show how some of them are employed to justify conscious violations of course policies. Third, student rationalizations occur in both an ad hoc and post hoc manner, and they are not necessarily self-consistent. These and other findings suggest that modern AI presents a steep, ethical, slippery slope which students conceptually slide down, landing far outside the pedagogical goals and expectations of instructors. We discuss implications for educational design and AI policy.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript reports findings from a qualitative study using semi-structured interviews (n=20), AI chat logs, and course documents (syllabi, assignments) to examine how students conceptualize and rationalize AI use in academic writing. It identifies five sites of AI-use conceptualization, presents a taxonomy of over 20 distinct rationalizations (e.g., viewing AI text as victimless or as 'own writing' if it matches beliefs/style, or claiming extensive AI use enhances learning), shows these occur ad hoc and post hoc without self-consistency, and concludes that AI creates a steep ethical slippery slope leading students far from pedagogical goals. Implications for educational design and policy are discussed.

Significance. If the taxonomy and slippery-slope inference hold, the work provides a detailed empirical map of student ethical reasoning around generative AI that could directly inform more effective academic integrity policies and pedagogical interventions. The multi-source data collection (interviews plus logs/documents) is a methodological strength that, if properly leveraged, strengthens the contribution to HCI and education research on AI ethics.

major comments (2)
  1. [Methods] Methods section: The manuscript provides no details on thematic analysis procedures (e.g., coding process, codebook development, or saturation criteria), inter-rater reliability, sample selection criteria, or steps to mitigate self-report and social desirability bias. This is load-bearing for the central claim because the taxonomy of >20 rationalizations and the ad hoc/post hoc distinction are derived entirely from the interview data; without these details the link between data and claims cannot be evaluated.
  2. [Results] Results/Discussion: Although the abstract and methods note collection of AI chat logs and course documents, there is no description of how (or whether) these were used for triangulation, to verify whether reported rationalizations match actual logged behaviors, or to distinguish interview-elicited justifications from in-situ reasoning. This directly engages the concern that rationalizations may be post-hoc interview artifacts rather than genuine decision processes, weakening the slippery-slope conclusion.
minor comments (2)
  1. [Abstract] Abstract: The claim of 'at least five distinct sites' is stated but the sites themselves are not enumerated; listing them would improve immediate clarity for readers.
  2. [Discussion] The paper correctly notes that rationalizations can be post hoc, but this acknowledgment should be more explicitly connected to limitations of the interview method in the discussion of validity.

Simulated Author's Rebuttal

2 responses · 0 unresolved

Thank you for the constructive feedback on our manuscript. We appreciate the emphasis on methodological transparency and data source integration, both of which are essential for substantiating the taxonomy of rationalizations and the slippery-slope claim. We respond to each major comment below.

read point-by-point responses

  1. Referee: [Methods] Methods section: The manuscript provides no details on thematic analysis procedures (e.g., coding process, codebook development, or saturation criteria), inter-rater reliability, sample selection criteria, or steps to mitigate self-report and social desirability bias. This is load-bearing for the central claim because the taxonomy of >20 rationalizations and the ad hoc/post hoc distinction are derived entirely from the interview data; without these details the link between data and claims cannot be evaluated.

    Authors: We agree that these procedural details are missing from the current Methods section and are necessary for readers to assess the derivation of the taxonomy and the ad hoc/post hoc distinction. In the revised version we will expand the Methods section with a full account of the thematic analysis (inductive open coding, iterative codebook refinement, saturation criteria), inter-rater reliability procedures (or justification for single-coder analysis), explicit sample selection criteria, and bias-mitigation steps such as confidentiality assurances, neutral phrasing of questions, and avoidance of leading prompts. revision: yes

  2. Referee: [Results] Results/Discussion: Although the abstract and methods note collection of AI chat logs and course documents, there is no description of how (or whether) these were used for triangulation, to verify whether reported rationalizations match actual logged behaviors, or to distinguish interview-elicited justifications from in-situ reasoning. This directly engages the concern that rationalizations may be post-hoc interview artifacts rather than genuine decision processes, weakening the slippery-slope conclusion.

    Authors: The observation is accurate: the manuscript notes collection of the logs and documents but does not describe their analytic use or any triangulation. In practice the logs and documents served mainly to contextualize interview prompts rather than for systematic verification against logged behavior or to separate ad-hoc from post-hoc reasoning. We will add an explicit subsection on multi-source data handling, clarify the limited role actually played by the logs and documents, and discuss this as a limitation that constrains strong claims about in-situ versus elicited rationalizations, thereby tempering the slippery-slope inference where appropriate. revision: yes

Circularity Check

0 steps flagged

No circularity: qualitative empirical study with no derivations or self-referential reductions

full rationale

This is a qualitative study relying on semi-structured interviews (n=20), thematic analysis, AI chat logs, and course documents to produce a taxonomy of student rationalizations. No equations, fitted parameters, model predictions, or mathematical derivations exist that could reduce claims to inputs by construction. No self-citation load-bearing steps, uniqueness theorems, or ansatzes are invoked. The central findings (five sites of conceptualization, >20 rationalizations, ad hoc/post hoc occurrence, slippery-slope inference) are presented as emerging from the data analysis itself, with no evidence of the taxonomy being presupposed or the results being forced by prior self-citations. The paper is self-contained as an empirical report against external benchmarks of interview data.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claims rest on standard qualitative-research assumptions about the validity of interview data and thematic analysis; no free parameters or invented entities are introduced.

axioms (1)
  • domain assumption Semi-structured interviews and thematic analysis can reliably surface students' authentic rationalizations about AI use.
    Core premise of the study design; assumes limited social-desirability bias and accurate self-reporting.

pith-pipeline@v0.9.1-grok · 5783 in / 1312 out tokens · 32152 ms · 2026-06-29T09:55:43.736761+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

7 extracted references · 2 canonical work pages

  1. [1]

    In Proceedings of the 2025 CHI Conference on Human Fac- tors in Computing Systems, 1–17

    Examining Student and Teacher Perspectives on Undisclosed Use of Generative AI in Academic Work. In Proceedings of the 2025 CHI Conference on Human Fac- tors in Computing Systems, 1–17. Ammari, T.; Chen, M.; Zaman, S.; and Garimella, K

    2025

  2. [2]

    arXiv preprint arXiv:2505.24126 (2025)

    How students (really) use ChatGPT: Uncovering ex- periences among undergraduate students.arXiv preprint arXiv:2505.24126. Ardito, C. G. 2025. Generative AI detection in higher educa- tion assessments.New Directions for Teaching and Learn- ing, 2025(182): 11–28. Balalle, H.; and Pannilage, S. 2025. Reassessing academic integrity in the age of AI: A systema...

    work page arXiv 2025

  3. [3]

    Journal of Academic Ethics, 24(1): 1–23

    Cheating Writing with Generative AI: Exploring Stu- dent Motivations Using the Theory of Planned Behavior. Journal of Academic Ethics, 24(1): 1–23. Gruenhagen, J. H.; Sinclair, P. M.; Carroll, J.-A.; Baker, P. R.; Wilson, A.; and Demant, D. 2024. The rapid rise of generative AI and its implications for academic integrity: Students’ perceptions and use of ...

    2024

  4. [4]

    Khosrowi, D.; Finn, F.; and Clark, E

    Redesigning assessments for AI-enhanced learning: A framework for educators in the generative AI era.Education Sciences, 15(2): 174. Khosrowi, D.; Finn, F.; and Clark, E. 2023. Diffusing the creator: Attributing credit for generative AI outputs. InPro- ceedings of the 2023 AAAI/ACM Conference on AI, Ethics, and Society, 890–900. K¨obis, N. C.; Van Prooije...

    2023

  5. [5]

    Lin, X.; Huang, H.; Shi, S.; and Vines, J

    GPT detectors are biased against non-native English writers.Patterns, 4(7). Lin, X.; Huang, H.; Shi, S.; and Vines, J. 2026. Rely- ing on LLMs: Student Practices and Instructor Norms are Changing in Computer Science Education.arXiv preprint arXiv:2602.05506. Lund, B. D.; Lee, T. H.; Mannuru, N. R.; and Arutla, N

    work page arXiv 2026

  6. [6]

    Turing Test

    AI and academic integrity: Exploring student percep- tions and implications for higher education.Journal of Aca- demic Ethics, 23(3): 1545–1565. Maramark, S. 1993.Academic dishonesty among college students. US Department of Education, Office of Educa- tional Research and Improvement . . . . McCutchen, D. 2000. Knowledge, processing, and working memory: Im...

    1993

  7. [7]

    V olokh, E

    AI’s learning paradox: how business students’ en- gagement with AI amplifies moral disengagement-driven misconduct.Studies in Higher Education, 1–18. V olokh, E. 2002. The mechanisms of the slippery slope. Harv. L. Rev., 116: 1026. Waqas, M.; Hania, A.; and Chunyan, X. 2026. Understand- ing AIgiarism in higher education: the lens of general AI attitudes a...

    2002