To disclose or not to disclose: Peer influence and psychological factors in students' use of generative artificial intelligence

Synthesis: To disclose or not to disclose — a mixed-methods study of 409 Singaporean undergraduates examining why students conceal their Generative AI use despite institutional disclosure mandates. Guided by Bandura's Social Cognitive Theory, it finds disclosure is primarily predicted by relational and social variables — perceived peer disclosure and comfort with instructors were the strongest predictors, while moral disengagement had weaker effects. Non-disclosure reflects strategic adaptation to perceived peer norms and low interpretive trust in instructors, not moral negligence.

Qu & Wang (2026) reframe GenAI non-disclosure in higher education as a relationally embedded practice rather than simple cheating. Despite mandates, students frequently conceal their GenAI use, reflecting ethical uncertainty and relational risk. The study moves the focus from compliance toward the social environments that make transparency possible.

Method

• Sample: 409 undergraduates at a Singaporean university who reported using GenAI tools for academic purposes.
• Theory: Bandura's Social Cognitive Theory — how cognitive, social, and emotional factors jointly shape disclosure willingness.
• Design: Mixed-methods. Quantitative hierarchical regression tested the effects of moral disengagement, perceived peer disclosure, and relational comfort with peers/instructors on self-reported disclosure likelihood. Qualitative thematic analysis of open-text responses explored students' interpersonal reasoning and perceived risks.

Key Findings

• Relational and social predictors dominate. Perceived peer disclosure and comfort with instructors were the strongest predictors of disclosure likelihood; moral disengagement showed weaker effects.
• Non-disclosure is strategic, not negligent. Qualitative findings revealed non-disclosure reflected strategic adaptation to perceived peer norms and low interpretive trust in instructors, rather than a lack of moral concern.
• Transparency depends on trust. Disclosure is a relationally embedded practice where ethical decisions are co-constructed within specific social and institutional contexts. Non-disclosure may reflect moral adaptation in environments with normative ambiguity and interpretive uncertainty.
• Institutional implication: Transparency depends less on compliance and more on environments that enhance trust and positive normative climates.

What this means for practice

• Instructors. Invest in relational safety: comfort with instructors was among the strongest predictors of disclosure, so make one-to-one conversations about GenAI use routine and non-punitive.
• Instructors. Frame disclosure as a metacognitive explanation of how the work was done rather than a confession of wrongdoing, which lowers interpretive risk without adding grading burden.
• Instructors. Run structured peer discussions about GenAI use to make private uncertainty visible — perceived peer disclosure was the strongest predictor, and students overestimate how much their peers conceal.
• Administrators. Embed routine declaration mechanisms in assessment workflows and keep policy communication consistent across departments, since variability amplifies uncertainty and avoidance; detection-and-punishment approaches alone are unlikely to work.
• Instructors. Recognize the "to disclose or not" dilemma as a real source of student stress and Trust pressure rather than evidence of moral negligence.

Limitations

• Conducted at a single Asian university with high technological adoption and an evolving GenAI policy, where strong academic pressure and emphasis on authority may intensify relational caution; comparative work in lower-power-distance contexts is needed.
• Several constructs were measured with single-item indicators, which precludes estimates of internal consistency reliability and may inflate measurement error, attenuating effect sizes.
• The sample of 409 undergraduates was determined by practical constraints rather than a priori power planning, and post-hoc registration weakens confirmatory inference.
• The cross-sectional design captures disclosure intentions at one point in time and cannot establish the temporal ordering of norms, moral disengagement, and relational trust.

Citation

Qu, Y., & Wang, J. (2026). To disclose or not to disclose: Peer influence and psychological factors in students' use of generative artificial intelligence. British Journal of Educational Psychology. https://doi.org/10.1111/bjep.70086