Skip to main content

For facilitators and community learning hosts

Teach the reasoning.
Make room for practice.

You do not need to present yourself as an AI expert. Prepare the examples, understand the evidence standards, and help learners explain and improve their work. These resources support facilitation; they do not authorize you to issue an IHA credential.

Prepare before you teach

  1. Complete the independent tasks yourself. Compare your work with the benchmark and identify a question you cannot yet answer.
  2. Rehearse a source-to-draft demonstration. Show one failure, how you noticed it, and what you changed.
  3. Check devices, reading needs, language, and access. Offer paper or spoken work with the same criteria. Do not require a paid AI account.
  4. Agree on privacy and consent. Use the provided fictional briefs; invite personal examples only when learners choose and can replace sensitive details with a fictional case.
  5. Arrange specialist support for questions beyond your competence. Saying “we need to verify that” models the judgment being taught.

A sample 75-minute session

This is a planning example, not a required pace. Split the session when learners need more time.

  • 0–10: For a first session, offer the short baseline before instruction; for a later session, recall a prior decision. Record access support and preserve the attempt.
  • 10–25: Inspect a flawed output, invite observations, and demonstrate the repair.
  • 25–40: Guided practice in pairs; coach with questions rather than supplying every answer.
  • 40–60: Independent changed case. Learners produce an artifact and explain a choice.
  • 60–70: Review against criteria, make one revision, and record why.
  • 70–75: Exit question, unresolved issues, and a short recall task for tomorrow.

Facilitator protocol 2026-09-24.1

Review what the learner can show.

Use the lesson-specific criteria and these shared checks. Judge each criterion separately; do not average away an important error or turn a completed field into a passing grade.

Evidence
Claims stay within the supplied facts; missing or conflicting information remains visible.
Useful output
The artifact answers the requested task and any calculation is traceable.
Judgment and boundaries
The next step fits the uncertainty, protects private information, and names necessary human checking.
Explanation and revision
The learner explains a choice and changes the artifact when the evidence changes.
  • Not observed: No usable evidence for this criterion, or the task was not attempted. Record the reason; do not count this as a pass or silently drop it.
  • Needs revision: The work contains an unsupported claim, incorrect calculation, missing boundary, or explanation that does not meet the criterion. Name one correction to try.
  • Supported in this artifact: The learner identifies relevant evidence and the artifact meets this criterion. Record any help; this alone does not establish independent transfer.
  • Demonstrated on a changed task: The learner meets this criterion on an unfamiliar case and explains the choice without a copied answer, generative answer, or content hint. Accessibility accommodations are allowed and recorded.

Record the help, not just the result.

  • Independent: no content hint, worked answer, or generated answer used for this attempt.
  • Access support: reading aloud, dictation, larger text, interpretation, or navigation help; describe it separately from content coaching.
  • Content hint: a person or AI pointed toward a concept, missing fact, or next step; record the hint.
  • Worked example or answer: a person, source, peer, or model supplied part of the response; describe what the learner contributed.

Access support does not lower a judgment. When a hint or generated answer supplied the reasoning, record an assisted attempt and use another unfamiliar task before reporting independent transfer.

Observe a starting point. Then test a new case.

Before teaching, offer the short baseline in the pack. Preserve the first attempt before feedback. After practice, use the different transfer brief and its changed condition. These are public practice cases: if someone has seen the answer, record that and prepare an equivalent reviewed variant.

Build your business: baseline task

Fictional case: a florist says, “Customers ask when bouquets will be ready.” The shop already has a collection-time board. No customer has agreed to pay for a new service. An AI proposes “Sell a $200 monthly chatbot; ten shops will buy.” In 5–10 minutes, separate facts from assumptions, suggest one smaller useful deliverable, and write one question to ask the owner. Explain what you would check before offering it. Use paper or speech if preferred.

Grow your organization or mission: baseline task

Fictional case: a neighborhood group records 12 registrations for a workshop; attendance has not been counted. The draft flyer says Thursday, while an organizer note says the date is still awaiting confirmation. An AI writes, “Twelve residents attended our Thursday workshop.” In 5–10 minutes, correct the statement, name what needs confirmation, and explain who should check a public version. Do not invent an event result.

AI for everyday life: baseline task

Fictional case: travel option A costs $12 per trip plus a $6 booking fee; option B costs $14 per trip with no booking fee. You need one trip. An AI says A is cheaper because $12 is less than $14. Accessibility and cancellation terms are not supplied. In 5–10 minutes, show the totals, correct the comparison, and name one thing to check before choosing.

Close the feedback loop.

  1. Before hosting, name the primary and backup reviewers and confirm an actual feedback window, submission channel, support contact, and access arrangements.
  2. Have two reviewers independently review weak, borderline, and strong samples. Compare criterion-level evidence, record decision rules, and resolve disagreements.
  3. Review the dated learner artifact, quote the evidence, and give one specific revision. Keep the original and revised versions; ask the learner to explain the change.
  4. Use a second reviewer for disputed or borderline judgments. Tell learners when an agreed feedback window cannot be met.

The downloadable pack includes a copyable record for artifact version, assistance, criterion, evidence, review level, revision, reviewer, and follow-up consent. Use an approved restricted storage location; downloading the pack does not create an instructor inbox.

Find out what lasts and what gets used.

One week: retained understanding
With consent, ask for an explanation without the example and a short unfamiliar task. Record assistance, time since instruction, and missing responses.
Thirty days: use in context
Ask whether the skill was used, what work can safely be shown, what changed, and what still needed correction. Separate observed evidence from self-reported benefit. “Not used” and “not measured” are valid findings.

Keep attendance, attempted work, assisted practice, independently reviewed performance, confidence, and reported outcomes separate. Report actual numerators and denominators, including missed follow-ups. These are human-run follow-ups, not reminders sent by this page.

Grow delivery with evidence

Begin with a small facilitated group. Check whether people can complete the work, explain it later, and use it in a changed situation. Track access barriers and revision needs alongside satisfaction. Expand only when teaching and feedback capacity can keep pace.

Learners can keep work on their device or explicitly save a private draft to their Academy account. Account-saved drafts contribute to Academy activity reporting and are accessible to authorized staff. These guides do not enroll a group or assign a reviewer. Agree how you will review artifacts before hosting a group.

Discuss a hosted learning group

A learner or facilitator needs help?

For a hosted program, use its agreed support contact. Otherwise, send the pathway/page, attempted step, expected result, safe error text, device/browser, and access need. Remove private URL tokens and do not attach credentials, learner records, or sensitive screenshots. Public inquiries do not promise an immediate response.

Ask for learning or access support

Our design draws on worked examples, retrieval, and spaced practice described by IES, and the human agency and application dimensions in UNESCO’s AI competency framework. These references do not constitute endorsement or accreditation.