SWANK AI Guidance Note 04
Education · AI Literacy · Academic Integrity · Critical Verification

Core Standard

TEACH USE. DEFINE BOUNDARIES. REQUIRE VERIFICATION.

Educational institutions increasingly need to achieve two objectives at the same time:

protect genuine academic authorship

and

prepare students to use artificial intelligence responsibly.

These objectives are compatible.

AI literacy should not require abandoning academic integrity.

Academic integrity should not require avoiding the technology students will increasingly encounter in higher education, employment and everyday life.

Purpose

Generative AI can assist students with:

understanding difficult concepts;
generating practice questions;
brainstorming;
research planning;
explaining terminology;
comparing approaches;
checking understanding;
organising ideas;
improving accessibility;
and identifying areas requiring further study.

It can also be misused to replace work that a student is required to produce independently.

The central governance challenge is therefore not simply:

Should students use AI?

It is:

What kind of AI use is appropriate for this particular learning or assessment activity?

Clear institutional boundaries are more useful than either unrestricted use or blanket prohibition.

The Central Distinction

Educational institutions should distinguish between:

AI Used to Support Learning

and

AI Used to Substitute for Required Independent Work

These are different activities.

A student asking AI to explain a mathematical concept is not necessarily equivalent to submitting AI-generated answers in an assessment requiring independent authorship.

Likewise, asking for feedback on structure is different from asking a system to produce an entire assessed submission.

Institutional policy should make those distinctions understandable in advance.

AI Literacy Is More Than Prompting

AI literacy is not simply the ability to obtain useful outputs from a generative system.

Students should learn to evaluate those outputs.

Relevant questions include:

Is this accurate?

What is the source?

What has the system missed?

What assumptions is it making?

Is the information current?

Can I verify this independently?

Does another reliable source disagree?

Is AI appropriate for this particular task?

These are AI-literacy skills.

They are also critical-thinking skills.

Verification

Generative AI can produce information that is:

accurate;
incomplete;
outdated;
misleading;
biased;
unsupported;
or confidently fabricated.

Students should therefore understand that fluent language is not evidence of accuracy.

Where AI produces material claims, students may need to verify:

factual assertions;
quotations;
statistics;
legal or technical propositions;
historical information;
references;
citations;
and source claims.

Verification should be taught as a normal part of responsible AI use rather than only as a response to misconduct.

A Green / Amber / Red Model

One practical approach is to classify AI use according to the requirements of the learning activity.

The exact boundaries should be adapted to the institution, subject and assessment.

GREEN — Learning Support

AI use may ordinarily be permitted or encouraged for activities such as:

explaining difficult concepts;
generating practice questions;
brainstorming;
developing research questions;
comparing possible approaches;
checking understanding;
identifying knowledge gaps;
practising revision;
organising non-assessed ideas;
exploring examples.

The purpose is to strengthen learning rather than substitute for it.

AMBER — Guidance or Disclosure Required

Some uses may be permitted only with:

teacher guidance;
explicit disclosure;
acknowledgement;
verification;
or assessment-specific conditions.

Examples may include:

editing assistance;
summarisation;
drafting support;
structural feedback;
translation;
coding assistance;
AI-supported research;
assistance connected with assessed work.

The institution should make clear:

whether the use is permitted;
what must be disclosed;
what the student remains responsible for;
and how AI assistance should be acknowledged.

Ambiguous rules increase the risk of inconsistent enforcement.

RED — Prohibited Use

Examples may include:

submitting prohibited AI-generated work as the student’s own;
concealing AI assistance where disclosure is required;
fabricating citations or evidence;
using AI during an assessment where the conditions expressly prohibit it;
asking AI to complete work intended to demonstrate the student’s independent competence;
presenting generated analysis as independently researched where this is prohibited.

The prohibited conduct should be defined clearly enough that students can understand it before assessment begins.

Assessment-Specific Rules

A single institution-wide AI rule may be insufficient.

Appropriate use can vary according to:

subject;
learning objective;
age;
assessment design;
professional requirements;
accessibility needs;
and the competence being tested.

For example, an assessment designed to test unaided writing may reasonably restrict AI drafting.

A research-methods exercise may instead require students to demonstrate responsible AI use, source verification and disclosure.

Good policy should therefore define boundaries at the level where the educational purpose becomes clear.

Academic Integrity Requires Prior Clarity

Students should not first discover the institution’s AI rules after being accused of breaching them.

Before consequential assessment, institutions should communicate:

whether AI is permitted;
what forms of assistance are permitted;
what forms are prohibited;
whether disclosure is required;
how AI use should be acknowledged;
what evidence may be considered if misuse is suspected;
and what review or appeal pathway exists.

Clear rules support both students and staff.

AI Detection

AI-detection technology may provide an indicator.

It should not automatically be treated as proof of authorship.

Detection systems can produce:

false positives;
false negatives;
uncertain probability scores;
inconsistent results;
and misclassification of highly structured human writing.

Where suspected AI misuse could affect:

grades;
progression;
disciplinary records;
qualifications;
or future opportunities,

the conclusion should be proportionate to the total evidence available.

Relevant evidence may include:

version history;
drafts;
previous work;
research notes;
source materials;
assessment conditions;
discussion with the student;
and the student’s ability to explain the reasoning and sources used.

Indicator is not determination.

Student Responsibility

Responsible AI literacy also requires clear student accountability.

Students should understand that they remain responsible for material they submit.

AI assistance does not excuse:

fabricated sources;
inaccurate claims;
prohibited assistance;
plagiarism;
or failure to comply with assessment requirements.

Where AI is permitted, students should be taught to use it critically rather than defer responsibility to the system.

Staff AI Literacy

Student guidance will be difficult to implement consistently if staff do not share a basic understanding of generative AI.

Staff training may need to address:

what generative AI can and cannot reliably do;
hallucinations;
source verification;
AI detection limitations;
appropriate educational uses;
disclosure rules;
accessibility;
assessment design;
data protection and confidentiality;
and consistent handling of suspected misuse.

AI governance is weakened when different staff members apply materially different assumptions to the same conduct.

Accessibility

AI may function as an assistive tool for some learners.

It may help students who experience difficulty with:

language;
writing;
speech;
cognitive load;
executive function;
information organisation;
fatigue;
or navigating complex instructions.

Educational policies should therefore distinguish between:

legitimate accessibility support

and

substitution for the academic competence being assessed.

That distinction may require context rather than a universal rule.

Privacy and Safeguarding

AI literacy should also include understanding what information should not be entered into unapproved systems.

Students and staff may need guidance concerning:

personal data;
sensitive information;
safeguarding material;
confidential records;
intellectual property;
unpublished research;
assessment content;
account security.

Responsible AI use includes understanding both the output and the data environment surrounding the tool.

Preparing Students for the Workplace

AI policy should consider the environment students are being prepared to enter.

Increasingly, professional competence may include knowing:

when AI can improve productivity;
when human judgment must remain primary;
how to verify generated material;
how to disclose use appropriately;
how to protect confidential information;
how to recognise automation bias;
and when not to use AI at all.

An educational system that teaches only avoidance may leave students poorly prepared for environments in which responsible AI use is expected.

The objective should be informed judgment.

Assessment Design

The growth of generative AI may require institutions to reconsider how learning is assessed.

Useful questions include:

What competence is this assessment intended to measure?
Does the assessment require independent recall, analysis, writing or problem-solving?
Could responsible AI use itself form part of the competence being assessed?
Can students explain their reasoning?
Can process evidence be incorporated?
Are assessment conditions clear?
Is disclosure practical?
Are rules consistent with the educational objective?

AI governance should support assessment design rather than rely entirely upon detection after submission.

Questions for Educational Institutions

Institutions developing AI policy may ask:

What do we want students to learn?
Where can AI support that learning?
Where would AI substitute for the competence being assessed?
Are permitted and prohibited uses clear?
Are disclosure requirements understandable?
Do staff apply the rules consistently?
Are students taught verification?
Are AI detection results treated proportionately?
Are accessibility needs considered?
Are privacy and safeguarding risks addressed?
Can students challenge an allegation of misuse?
Are we preparing students for responsible future AI use?

SIAAF Relevance

This Guidance Note principally relates to:

Domain 07 — AI Literacy & Organisational Readiness

Do the people using and governing AI understand it well enough to exercise judgment?

It may also engage:

Domain 01 — Governance & Accountability

Where responsibility for institutional AI policy and assessment rules must be identifiable.

Domain 03 — Evidence & Traceability

Where allegations of prohibited AI use require an appropriate evidential basis.

Domain 05 — Escalation, Challenge & Contestability

Where students require fair routes to question or challenge consequential findings.

Domain 06 — Risk, Harm & Operational Resilience

Where poor AI policy could create educational, privacy, safeguarding or progression consequences.

SWANK EDUCATION Standard

TEACH USE. DEFINE BOUNDARIES. REQUIRE VERIFICATION.

Students should leave education understanding not simply how to obtain an AI-generated answer, but how to judge whether that answer should be trusted or used.

Educational institutions should:

teach responsible use

define assessment boundaries

require appropriate disclosure

teach source verification

preserve academic authorship

apply evidence fairly

and

prepare students for informed AI use beyond education.

Related SWANK AI Guidance

Guidance Note 03 — Human Oversight in AI-Assisted Decision Systems

Guidance Note 06 — AI Detection: Indicators Are Not Proof

Guidance Note 07 — Public-Sector AI and Accessibility

Guidance Note 13 — Automation Bias in Professional Decision-Making

Guidance Note 20 — Governing AI in High-Stakes Environments

SWANK EDUCATION / SWANK AI

AI Literacy · Academic Integrity · Critical Verification

The appropriate response to powerful technology is not ignorance. It is informed judgment.