"Should I Tell My Teacher?" Student AI Disclosure Practices, Stigma, and Self-Regulated Learning in Higher Education

Synthesis: Student AI disclosure is not only a matter of policy compliance — it is an affective process intertwined with self-regulated learning. Chang, Lin, Huang, and Ryoo (2026), studying 78 undergraduates, find that students' worries about judgment, stigma, dependence, and replacement are associated with greater concealment and peer-only disclosure, while teacher-directed disclosure is consistent with adaptive Help-Seeking. A key finding: anxiety does not suppress disclosure wholesale — it redirects it toward safer peer outlets, cutting students off from the instructor Feedback that could calibrate their AI use. Disclosure norms vary by discipline and language, and do not develop automatically with academic progression.

Key Findings

• Worry redirects disclosure toward peers, not outright suppression. Higher worry was associated with a desire for total non-disclosure and peer-only sharing, but not with lower disclosure to teachers directly. Anxious students did not stop talking about AI — they moved the conversation to peers, who cannot give the calibrated feedback teachers can.
• Two interpretive response patterns emerge: a transparent tendency (disclosing to teachers, checking with instructors, treating instructor interaction as an external regulatory checkpoint) and an anxious-concealing tendency (frequent AI use for substantive tasks paired with elevated worry and withdrawal from formal feedback — consistent with maladaptive AI Regulation in Education).
• Heavy AI users are both most worried and least connected to feedback. Students who frequently uploaded drafts for AI review reported the highest worry, more secrecy, and no increase in teacher disclosure — the group at greatest risk of passive, uncritical reliance, and least likely to receive instructional attention.
• Discipline is the strongest predictor of disclosure to teachers. Education, Social Sciences, and STEM/Health students reported higher disclosure than Business students (with Humanities not differing from Business). Language status was marginal (monolingual students tended to disclose less); years of study did not predict disclosure.
• Disclosure ≠ help-seeking, but can make it visible. Help-seeking is obtaining assistance; disclosure makes that assistance visible to others. The two are logically and temporally separable.

What this means for practice

• Instructors. Replace behavior-only declaration requirements with low-barrier disclosure touchpoints — in-assignment reflection prompts, opt-in consultations, or embedded AI-use reflection tools — so students surface AI use as learning AI Regulation in Education rather than as confession.
• Instructors. Normalize AI use in class with instructor-led examples and an explicit statement that disclosure will not automatically be penalized, lowering the social cost of transparency that drove concealment in this study.
• Administrators. Scaffold disclosure norms deliberately from the first year instead of assuming transparency develops with academic progression: years of study did not predict disclosure to teachers here (β = −0.12, p = .303).
• Instructors. Give the heaviest AI users direct instructional attention: students who frequently uploaded drafts for AI review reported the highest Worry Index scores (r = .477) and more complete secrecy (r = .305) with no detectable increase in disclosure to teachers (r = –.146).
• Faculty developers. Build discipline-specific disclosure guidance: academic discipline was the largest unique predictor of disclosure to teachers (partial r = .36), so one cross-campus declaration policy will not fit Education, Business, and Humanities programs equally.

Limitations

• Single online education course at one Western Canadian university, housed in an Education department — a context the authors note may have been more favorable to disclosure than other academic settings, strongly limiting external validity.
• Participation was voluntary and incentivized with up to two percent of the course grade, tied to depth of response; students already comfortable discussing AI use may have self-selected, inflating reported transparent disclosure.
• Modest sample (N = 78) with an underpowered regression: the Business reference group held n ≈ 5 and the language-status effect was marginal (B = –0.51, p = .053), so the demographic findings are hypothesis-generating only.
• The 11-item instrument is newly developed without full psychometric validation: the proposed Disclosure to Peers (α = .34) and AI Usage (α = –.30) composites failed reliability and were analyzed item by item, and the cross-sectional design precludes causal claims about worry and concealment.

Citation

Chang, D. H., Lin, M. P. C., Huang, J.-Y., & Ryoo, J. (2026). "Should I Tell My Teacher?" Student AI Disclosure Practices, Stigma, and Self-Regulated Learning in Higher Education. Frontiers in Education, 11, 1826174.