Six principles for classroom AI guidance
What good guidance should do, synthesized from the past year’s research and current institutional policy.
- State AI rules in every syllabus, per assignment where they differ.
- Protect integrity through process evidence, not detector scores.
- Faculty decide within shared institutional guardrails.
Say it in the syllabus
Every course should state what AI use is allowed, what is not, and what must be disclosed. Where assignments differ, say so per assignment.
Assess the process
Grade drafts and reflections along with the final product. Detector scores are not reliable enough to decide misconduct cases.
Transparency and human verification
Students say how they used AI and stay responsible for accuracy. Written rules also protect students who worry about being falsely accused.
Equity, privacy and AI literacy
Protect student data and give every student access to approved tools. AI literacy belongs in courses, not only in orientation.
Faculty discretion within a shared framework
Instructors decide AI use for their courses within institutional guardrails that adapt as tools change.
Set AI use for each assignment
One allow-or-ban rule for a whole course rarely fits every assignment. Set the level per task and tie it to what students should learn.
We drew these principles from the sources linked on each card. They are our summary, not a quotation.
Equity and access in practice
Principle 4 at course level. Equal access to tools is not the same as an equal chance to build AI skills. Students differ in paid subscriptions, disability, language, prior exposure to AI and whether they are first in their family at college.
Name a free tool before requiring AI
If an assignment requires AI, name a free or institution-provided tool. Do not assume students pay for premium versions.
HEPI (2026): 38% say their institution provides AI tools ↗Offer a route without AI
Some students opt out for privacy or personal reasons. Give them an equivalent task that does not need an AI account.
Crompton et al. (2026): privacy and equitable access as core areas ↗Teach the tool before grading its use
Show how to prompt, check output and cite AI in class. Students arrive with very different experience.
From Novelty to Normalisation (2026) ↗Check accessibility before adopting
Confirm an approved tool works with screen readers and other assistive technology. Campus accessibility reviews cover this.
Campus AI Registry: review evidence ↗Plan for language, devices and connectivity
Students writing in a second language may use AI for language support; say whether that is allowed. Make sure required tools work on campus devices and low-bandwidth connections.
Crompton et al. (2026): equitable access as a core area ↗Protect student work and privacy
Do not require students to upload their work or personal information to tools that keep or train on it. Prefer institution-approved tools with reviewed data terms.
Harvard: protect confidential data ↗Ask how students use AI
Use differs across student groups. A short survey in week one shows where support is needed.
HEPI (2026) ↗