Jiyoon Kim et al. (2026) — arXiv preprint. (catch-up — submitted 2026-05-27)
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. Through semi-structured interviews (n=20), analysis of AI chat logs, and course documents, the researchers identified at least five distinct conceptual sites where students' interpretation of AI policies diverges from faculty intent. Students employed over 20 distinct rationalizations — including 'copying AI-generated text is victimless' and 'any AI text reflecting their own beliefs is their own writing' — to justify conscious violations of course policies. Modern AI presents a steep, ethical, slippery slope which students conceptually slide down, landing far outside the pedagogical goals and expectations of instructors.
Key Contributions
- Five disconnect sites: Kim et al. identify at least five distinct conceptual sites where students' interpretation of AI use policies diverges from faculty intent — ranging from policy understanding to actual usage behaviors. This reveals systemic gaps in how academic-integrity policies are communicated and internalized.
- Taxonomy of 20+ rationalizations: Students employed over 20 distinct justifications for AI use, including victimless-crime framing ("copying AI-generated text is victimless"), ownership redefinition ("text reflecting my beliefs is my own writing"), and learning-optimization claims ("I learn more by using AI extensively"). These rationalizations were ad hoc, post hoc, and not self-consistent.
- Ethical slippery slope: The paper characterizes modern AI as presenting a "steep, ethical, slippery slope" where students conceptually slide far outside pedagogical goals. This finding extends work on over-reliance by documenting the metacognitive mechanisms students use to justify AI dependence.
- Implications for writing-education: As AI writing tools become ubiquitous, educators must design assignments and policies that account for these rationalization patterns — not just detection-based interventions. This connects to research on ai-assisted-writing-research-teams and the shifting norms of academic writing in the AI era.
- ai-literacy gap exposed: The study reveals that students' ethical frameworks for AI use are underdeveloped and self-serving, highlighting the urgent need for explicit AI literacy instruction that addresses ethical reasoning, not just technical capability. This aligns with the agentic-literacy-debt framework's emphasis on governance infrastructure for AI use.
- Cross-listed from cs.HC, this work brings human-computer interaction methods (semi-structured interviews, chat log analysis) to bear on higher-ed integrity challenges, complementing quantitative studies on ai-assistance-reduces-persistence with rich qualitative evidence of student reasoning.
Related Pages
- learning-analytics — Educational data infrastructure and analytics pipelines
- intelligent-tutoring — AI tutoring systems and student modeling
- academic-integrity — AI's impact on academic honesty and policy
- over-reliance — Risks of student dependence on AI assistance
- ai-literacy — Building student and educator competency with AI tools
- student-experience — How students interact with and perceive AI in education
- edtech-platform — Platform and infrastructure design for educational technology
- llm-detecting-llm-generated-content-education — Distinguishing Artificial from Authentic: Evaluating LLMs fo
Citation
APA: Jiyoon Kim, Kentaro Toyama, Sangmi Kim, & John M. Carroll (2026). "It's OK Because...": The Wild West of Student Rationalization of AI Use in Academic Writing. arXiv:2605.29090. arXiv preprint.