Academic Integrity in the Age of AI: Assessment Design and Disclosure Practices

Clarify what integrity means in the task

Academic integrity requires an honest account of how work was produced and a credible demonstration of the learning being assessed. AI changes the assistance available, but the central question remains: does the submission represent the student's achievement under the stated conditions? A vague instruction to use AI responsibly does not answer that question.

TEQSA's assessment reform guidance addresses the challenge of assuring learning in an AI-enabled environment.^[1] Institutions can respond by making assessment expectations explicit and examining whether existing tasks provide sufficient evidence of student capability. The appropriate design will differ across subjects and learning stages.

Write task-specific rules

For each assessment, state whether AI may be used for brainstorming, language assistance, coding, analysis or content generation. Explain the educational reason for any restriction. A task intended to assess independent writing may have different conditions from one assessing critical evaluation of AI output. Make rules available in the brief and discuss examples before submission.

Specify disclosure requirements in usable language. Students might identify the tool, purpose, extent of use and verification undertaken. Where process records are requested, explain what to retain and avoid collecting unnecessary personal or confidential information. Ensure that rules are consistent with accessibility arrangements and other authorised support.

Strengthen the evidence of learning

Map the programme's essential capabilities and identify where students demonstrate them with sufficient assurance. A combination of written work, practical performance, supervised tasks and structured dialogue may provide richer evidence than a single unsupervised submission. Select methods according to outcomes rather than adopting one universal response.

Process checkpoints can reveal developing understanding. Ask students to explain a choice, revise a response to feedback or apply an idea to a new situation. These activities should be purposeful and proportionate. Requiring extensive surveillance or a record of every keystroke can create privacy and accessibility problems without resolving the underlying assessment question.

Teach disclosure and judgement

Give examples of acceptable assistance and misrepresentation. A student may legitimately use a permitted tool to improve clarity while remaining responsible for the argument and references. Presenting generated analysis as personally verified research, or concealing prohibited assistance, raises a different concern. Discuss how to verify tool outputs and acknowledge uncertainty.

UNESCO's AI guidance places ethical and educational considerations within institutional adoption.^[2] Translate these principles into concrete classroom expectations. Provide an equivalent learning route where a tool is optional or inaccessible. Avoid assuming that every student has the same account, digital confidence or ability to pay for additional features.

Illustrative example: a staged policy assignment

Imagine a course in which students write a policy memo. The first stage requires an independently prepared problem statement during class. At the second stage, students may use an approved AI tool to generate counterarguments, but must verify evidence and disclose the assistance. The final stage includes a short discussion of one decision in the memo.

The rubric evaluates the quality of reasoning, evidence and response to alternatives. The conversation is planned as part of assessment for all students rather than imposed only when writing appears unusually polished. This makes the evidence requirements predictable and reduces reliance on subjective suspicion.

Handle concerns through a fair process

An automated indication, writing-style change or unusual phrase may justify a question, but should not be treated as conclusive proof of misconduct. Follow the institution's procedures for evidence, notice, response, impartial decision-making and review. Consider legitimate explanations, including authorised assistance and changes in a student's writing process.

Keep teaching support distinct from disciplinary judgement. Staff need training in how to discuss concerns without accusation and how to document relevant evidence. Students should understand the available advice and appeal routes. Decisions should explain the rule, evidence and reasoning rather than invoke a tool's score as an unquestionable authority.

Review the whole assessment system

Monitor student understanding of the rules, staff workload, accessibility and the quality of learning evidence. Review cases for inconsistent interpretation and update examples accordingly. TEQSA's later implementation resource offers examples of institutions putting assessment reform into practice.^[3] Such examples can inform discussion without being copied uncritically into a different context.

A credible approach combines assessment design, student education and fair procedures. The objective is to understand and verify learning while preparing students to use emerging tools honestly and thoughtfully, with clear responsibility for the work they submit.