Draft your institutional AI-use guideline. Seven mandatory components, optional eighth section, red-team backlog, and AI-use disclosure. ≤2 pages when printed. [LOCAL] = institution-specific body to be named by the owner.
What counts as "AI use"; who and what is covered. Settle definition fights here — most policy fights are secretly definition fights.
Beneficence · Non-maleficence · Autonomy · Justice · Explicability — localized. Principles survive model churn; rules below get versioned.
The three-lane vocabulary, institutionalized: every course declares lanes per assessment (🔴 Restricted / 🟡 Permitted with disclosure / 🟢 Required). Default when a syllabus is silent: ____________.
One canonical AI-use statement format — used by students AND staff. Symmetry is credibility; a double standard erodes trust. Mirror-test required: audit your own AI use before drafting policy.
Process evidence primary (version trails, vivas, in-class anchors). Detector output: at most a screening signal, never sole evidence (Liang et al., Patterns 2023 — citation stays in the policy).
PDPA 2010 floor: what student data may enter which class of tool under what agreement. Institutional accounts over personal ones. [LOCAL: data-protection officer / relevant committee]. Platform governance: logging and auditing of institutional AI-platform usage, anonymization protocols, mental-health flag procedures (per MIT report §3.3.8–9).
Named owner · version number · review date · student representation in review. Monitoring metrics: track AI-use adoption, campus engagement, student satisfaction, post-graduation feedback sources. [LOCAL: MQA/MOHE alignment; senate/faculty approval path].
Designated AI Lead(s) per department or faculty · AI Implementation Team or committee · AI Fellows or equivalent training capacity · Pilot Fund for AI-curricular experiments. Without infrastructure, the policy is aspirational. (Per MIT report §3.3.1–4.)
Findings that could not be patched in the sprint. Severity-ranked, with the persona who raised each finding. Backlogs are how products stay honest about incompleteness.
Your guideline must carry its own AI-use statement. Drafted with [tool] for [what]; verified and authored by the team named above. Credibility demands transparency.
This is a 45-minute drafting sprint for teams (or individuals working solo). The output is a ≤2-page institutional AI-use guideline with 7 mandatory components, an optional 8th section, a red-team backlog, and an AI-use disclosure.
[LOCAL] slots. These are placeholders for institution-specific bodies (senate, faculty committee, data-protection officer, MQA/MOHE alignment). Your team must name them.The facilitator circulates with three interventions only: "Which component is that?" · "Can a first-year act on this sentence?" · "Where's your review date?"
[LOCAL] slots marked, owner namedThe guideline is part of the portfolio presented in Session 8 and scored against the S8 rubric: Evidence discipline, Design over policing, Deployability, Ethical reasoning, Falsifiability. The guideline tests the "Design over policing" dimension — a policy that bans more than it enables will score below "Proficient."
INSTITUTIONAL GUIDELINE FOR ETHICAL AI USE IN TEACHING AND LEARNING v0.1 · Owner: Dr. Aisyah Rahman, Faculty AI Lead · Review date: 3 Sep 2027
This guideline governs the use of AI tools — including generative, analytical, and assistive — across all teaching, learning, assessment, and academic-support activities at the Faculty of Social Sciences, [UNIVERSITY]. AI use includes any instance where machine-generated content or analysis forms part of an academic or administrative process. Excluded: IT-administered infrastructure tools (LMS analytics, plagiarism detection) governed by existing IT policy. [LOCAL: Senate Teaching & Learning Committee to review exclusions annually.]
Beneficence (AI serves learning outcomes) · Non-maleficence (AI does not disadvantage any student) · Autonomy (choice preserved) · Justice (equitable access) · Explicability (all AI use is disclosed and auditable). These principles are invariant across model versions.
All course syllabi shall declare a lane per assessment component: 🔴 Restricted (no AI, process evidence required), 🟡 Permitted with Disclosure (AI allowed when disclosed), 🟢 Required (AI use is part of the task). Default when silent: 🟡 Permitted with Disclosure. [LOCAL: Faculty Board to confirm default lane by [DATE].]
"I used [TOOL] for [PURPOSE]. I verified outputs for accuracy, relevance, and bias. I take full responsibility." Same format for students and instructors. Instructor AI-use disclosure published on course site.
Process evidence (version trails, vivas, in-class drafts) is primary. AI-detector output is a screening signal only, never sole evidence (Liang et al., Patterns 2023 — 61%+ false-positive rate against non-native writers).
PDPA 2010 floor: no student data enters unlicensed tools. Institutional accounts mandatory. [LOCAL: Data Protection Officer — Dr. Chong Wei Ming] to maintain approved-tool register. Quarterly audit of AI-platform usage by IT.
Owner: Dr. Aisyah Rahman. Review: ≤12 months. Student representation: 2 seats on review panel. Monitoring: AI-use adoption, campus engagement, student satisfaction. [LOCAL: MQA/MOHE alignment — crosswalk to national guidance when published.]
AI Lead per faculty · AI Implementation Committee (quarterly) · Pilot Fund (MYR 10,000) for AI-curricular experiments.
[High] Laundering Student: Default 🟡 lane lets students disclose without reducing AI use — consider minimum 🔴 components per module.
[High] Overworked Lecturer: Process-evidence requirement adds grading time — companion staffing note needed.
[Medium] Falsely Accused: No appeals timeline or adjudicator named.
[Medium] Auditor: Quarterly data audit planned but auditor body unnamed.
Drafted with Claude for structural suggestions. Content reviewed, revised, and approved by Dr. Aisyah Rahman and the Faculty AI Task Force. Final text is the responsibility of the named owner.
Sample adapted from the course-pack exemplar (Appendix D5). Every team's guideline will differ by institution, discipline, and local context.