Addressing student non-compliance in AI use declarations: implications for academic integrity and assessment in higher

Synthesis: Fear, ambiguity, inconsistent enforcement, and peer influence drive students to avoid AI use declarations — even when declaration is mandatory. Gonsalves (2025), a mixed-methods study at King's Business School where 74% of students failed to declare AI use despite it being required on a coursework coversheet, uses the Theory of Planned Behavior (TPB) to explain non-compliance. Students view declaration as risky self-incrimination rather than a neutral administrative task, and the blurring of authorship by generative AI challenges traditional Academic Integrity norms. Clear, consistent, and Trust-based policies are needed to foster ethical AI use.

Key Findings

• Non-compliance is high and strategic. 74% of students failed to declare AI use despite a mandatory declaration on the coursework coversheet. Non-compliance is often a deliberate decision driven by fear of academic repercussions, not ignorance or negligence.
• Four barriers to disclosure: (1) fear of academic penalties, (2) ambiguous guidelines about what counts as "AI use," (3) perceived inconsistency in enforcement across courses/instructors, and (4) peer influence and competition ("everyone is using it"; "ChatGPT is like the fourth man" in group work).
• The declaration felt like self-incrimination. Placing the AI declaration alongside plagiarism statements fostered suspicion; students described it as "admitting to something wrong." Many believe AI use should remain private, like using a calculator.
• TPB helps but needs extension. Attitudes, subjective norms, and perceived behavioral control explain compliance, but GenAI's ethical ambiguity, internalized peer norms, and the philosophical divide over whether AI is a personal tool or a resource requiring disclosure complicate the model.
• Students want better support. They recommended clearer guidelines, a "checklist" for AI declarations, workshops on ethical AI use, consistent enforcement, and a "friendly," trust-based policy rather than suspicion and punishment.

What this means for practice

• Educators. Replace the declaration-as-integrity-warning framing with a positive, checklist-style AI declaration: 74% of students failed to declare when the declaration sat alongside plagiarism statements and felt like "admitting to something wrong."
• Administrators. Standardize enforcement across courses and instructors; perceived inconsistency was one of the four barriers to compliance, alongside fear of penalties, ambiguous guidelines, and peer influence.
• Faculty developers. Train faculty to communicate AI policy consistently, so students hear the same rules on acceptable use and how to declare it in every course.
• Assessment designers. Introduce AI declarations in low-stakes formative tasks first to normalize the practice before attaching them to high-stakes graded coursework.
• Administrators. Reframe AI policy toward innovation and ethical responsibility — a friendly, trust-based policy backed by clear guidelines and workshops — rather than compliance and suspicion.

Limitations

• Single case study at King's Business School: the survey yielded 57 valid responses, 3.7% of the population, and the authors note that the institution's unique cultural and policy environment limits the applicability of these findings.
• Seventeen semi-structured interviews (56 minutes to 1 hour 43 minutes, £30 per participant) were conducted and analyzed by the sole researcher, whose reflexivity procedures — repeated readings and reflective journaling — stand in for independent coding.
• The survey instrument was a service evaluation rather than a purpose-built research measure, and the authors caution that the study captures only the early phase of AI policy implementation.
• The analysis stops at one school and one framework: the authors state that expanding TPB or integrating other theoretical perspectives is needed to capture GenAI's ethical ambiguity, so the attitude–behavior links reported here are not transferable as policy effects.

Citation

Gonsalves, C. (2025). Addressing student non-compliance in AI use declarations: implications for academic integrity and assessment in higher education. Assessment & Evaluation in Higher Education, 50(4), 592–606.