Version: 2.0. Audience: Classes 8–10 first, with adaptable notes for Classes 11–12. Status: school pilot resource.
> Educational resource only. Not legal advice, not regulatory certification, not emergency or mental-health support.
GUARD is a WhizzStep classroom method for teaching judgement about AI use — not an externally accredited standard, and not a substitute for your school's academic-integrity, safeguarding or IT-acceptable-use policies. It teaches five habits: get clear on the Goal, Understand limits, Assess evidence, Respect people, and Decide responsibly and disclose. It does not require a public AI account, a student login, or any live language model — every tool runs curated content in the browser.
What GUARD is not: a certification of AI competence, a fairness or safety audit tool for real AI systems, a crisis-response service, or a replacement for teaching subject content. It is a structured way to practise decisions students already face.
The core pathway targets Classes 8–10. Scenarios assume the reading level, social situations (group assignments, class WhatsApp groups, school announcements) and stakes typical of this age range. Section 17 below covers adapting the same missions for Classes 11–12.
| Letter | Outcome |
|---|---|
| G — Goal | States the real learning or safety goal behind a task, separate from the convenience of a ready-made AI answer. |
| U — Limits | Names specific ways an AI output could be wrong, invented, missing context, or falsely confident — not just "AI can be wrong" in the abstract. |
| A — Evidence | Traces a claim to an independent, authoritative source rather than accepting fluency, repetition or self-confirmation as proof. |
| R — People | Identifies who could be exposed, excluded, misrepresented or harmed, including people not directly asking the question. |
| D — Decide | Chooses and justifies one of: use, revise, verify further, reject, disclose, or escalate to an adult — and can explain why the alternatives were weaker. |
All tools are static pages with no server dependency once loaded, so a shared computer lab with one load per class works. Where devices are scarce:
/guard-focused-labs/ → Printable and offline resources) for a pen-and-paper lesson./guard-toolkit/) work fully offline once printed."Today we are not learning how to make AI do everything for us. We are learning how to judge AI responsibly.
A strong student is not the one who always uses AI fastest. A strong student is the one who knows when to use it, when to check it, when to disclose it, and when to stop.
We're going to use a method called GUARD: Goal, Understand limits, Assess evidence, Respect people, Decide. You'll practise it on realistic school situations — not quiz questions with one right answer, but decisions where reasonable people could disagree, and where I want to hear your reasoning, not just your choice.
Nothing you type here needs your name, goes to any company, or gets shared outside this room unless you choose to print or export it. If anything in today's session reminds you of something real that's worrying you, come and talk to me afterwards — separately from the activity."
Two ready-made pathways are available as separate documents: the Four-Session Plan (a compact first pilot route) and the Eight-Session Plan (deeper practice with one session per GUARD letter plus bias/fairness and capstone). Both are linked from the Educator Hub. Section 9 below gives the detailed facilitation sequence referenced by each session.
Use this structure for any session, whether from the four- or eight-session plan:
| Element | What to prepare |
|---|---|
| Objective | One sentence: which GUARD letter(s) and which decision skill this session targets. |
| Preparation | Open the relevant Scenario Lab mission or Focused Lab in a browser tab; check the Answer Key section for that mission; print worksheets if needed. |
| Time | Foundation missions run 6–8 minutes of interaction; Intermediate/Advanced run 7–9. Budget the remainder for discussion — discussion should be at least as long as the activity. |
| Activity | Students work individually or in pairs through the mission or Focused Lab, in Practice-with-hints mode for a first pass. |
| Teacher prompts | Use the mission's "what to look for" cue and teacher discussion prompts from the Answer Key rather than asking "what did you pick?" — ask "what evidence would change your mind?" |
| Expected misconceptions | See Section 15 — most missions have one predictable wrong turn (e.g. treating fluency as proof, treating consistency as fairness). |
| Evidence of learning | A completed Decision Record entry, or a spoken justification that names the specific risk in the scenario rather than a generic "AI can be wrong." |
| Safeguarding note | Flag any session using the viral-audio or worried-friend missions — see Section 16 before running these. |
The 10-question Diagnostic (/guard-diagnostic/) gives a baseline out of 20, with two questions per GUARD letter (out of 4 each). Run it in Session 1 before any mission, so students have a personal reference point. It autosaves in the browser and can be retaken — treat the first attempt as baseline, not as a mark. Print the result for a portfolio if your pilot needs a paper record.
The eight Focused Labs (/guard-focused-labs/) are short, single-concept activities — classification, multi-select, builder or threshold-tuning tasks — each mapped to one part of GUARD. Use them for a warm-up before a full Scenario Lab mission on the same theme, or standalone when a session is short. All three modes (Practice with hints, Independent challenge, Teacher-led discussion) are available from the lab's mode selector.
Each of the six missions takes students through all five GUARD stages for one realistic situation, then shows a printable learning snapshot. Start the class together on the situation and evidence pack, then let students work the five decisions at their own pace. Reconvene for the model decision and the two discussion prompts in the Answer Key — these are written to surface disagreement, not close it down quickly.
The Decision Record (/guard-decision-record/) is a short structured reflection students complete after any mission or real assignment: what they decided, why, and what evidence mattered. It autosaves, prefills the mission title/task from a mission's handoff link, and exports to TXT/JSON or print — never automatically to WhizzStep. The AI-Use Log (/ai-use-log/) is a running personal record of where a student used AI on real schoolwork, for disclosure. Introduce the Decision Record after the first mission and the AI-Use Log once you set a real assignment that permits AI assistance (Section 19).
The capstone (/guard-capstone/) asks students to complete a full Decision Record, evidence trail, affected-people analysis, a final decision, a disclosure or escalation statement, and a 150–250 word reflection on one ambiguous case. Assess it against the Assessment Rubric's five competencies (Goal clarity, Limitations, Evidence, Privacy and people, Decision), not against whether the student reached the same final decision as the model answer — a well-reasoned "verify further" and a well-reasoned "reject" can both be Proficient.
The scenarios are written so that the score-1 option in most decisions is a genuinely defensible partial answer, not a distractor. When a student picks it and argues well:
The viral-audio and worried-friend missions are the two most likely to surface a real concern. If a student discloses something real — a real deepfake or impersonation incident, a real friend at risk — stop treating it as the simulation immediately:
See the School Pilot Kit's Safeguarding Protocol for the full procedure.
The six missions and eight Focused Labs work unchanged for older students; adapt the discussion rather than the content:
| Subject | Application |
|---|---|
| English | Use the quotation mission alongside a real citation-checking exercise on a set text; discuss AI-assisted brainstorming versus AI-written analysis. |
| Science | The finished-science-report mission maps directly onto lab-report write-ups; require a one-line AI-use disclosure on any submitted report. |
| Social science | The scholarship-shortlist mission supports units on statistics, policy or civics; discuss real examples of algorithmic decision-making in admissions or lending at a level appropriate to the class. |
| Computing | Use the Focused Labs' builder/threshold activities alongside a lesson on how classifiers and thresholds actually work, then connect back to the fairness stakes in mission 5. |
When you set a real assignment, state explicitly what is permitted (e.g. brainstorming, grammar checking, structure feedback) and what is not (e.g. AI-generated analysis submitted as the student's own reasoning), matching the finished-science-report mission's framing. Require a short AI-Use Log entry or one-line disclosure statement with every submission where AI assistance was used at all — this makes disclosure routine rather than an admission of guilt.
End any session with a short round: "Where did GUARD change what you would have done a week ago?" Keep the answers ungraded. Point strong students toward the capstone and certificate-track content on the GUARD page; point the whole class toward the AI-Use Log as an ongoing habit rather than a one-off activity.
Educational resource. GUARD is a practical WhizzStep classroom method, not an externally certified or accredited standard. Not legal, safeguarding-policy or emergency-response advice.