Facilitator Guide: One Task, Three Conversations

Status: Draft for learning-design, accessibility, and pilot review

Learning purpose

Help beginners experience prompting as a repairable conversation rather than a test of whether they know a perfect formula. Learners should leave able to compare a natural request, a clearer request, and one deliberate follow-up while keeping privacy, verification, and human ownership visible.

Preparation

• Allow 20–30 minutes for the lab and 10 minutes for debrief.
• Open the 🧪 Lab · Activity, 📎 Resource · worksheet, and choose an example from the 📎 Resource · Gallery.
• Decide whether learners may use an approved chatbot, observe a demonstration, or use only the supplied scenario and illustrative responses.
• Test display size and keyboard navigation if the local notebook will be used.
• Do not require an account, fast typing, personal disclosure, or spontaneous speaking.

Suggested flow

1. Frame the comparison — 3 minutes. Explain that the task stays steady while the request changes.
2. Choose or supply a task — 3 minutes. Screen learner tasks for privacy and consequence before anyone enters them into a tool.
3. Round one — 4 minutes. Let learners ask naturally and mark assumptions without repairing the response.
4. Round two — 6 minutes. Add goal, relevant context, desired output, and a useful boundary. Use a fresh thread when practical.
5. Round three — 4 minutes. Continue the round-two thread with one focused follow-up based on a visible need.
6. Compare and decide — 5 minutes. Record what improved, what remained unsupported, what was withheld, and what needs human review.

Keep it low stakes

Do not use medical, legal, financial, employment, safety, account, client, or other consequential decisions. If a learner begins disclosing private or another person’s information, pause before submission and move to the supplied fictional scenario. Do not ask a learner to explain the sensitive detail they chose not to share.

Watch for

• Changing the task between rounds. Bring the learner back to one steady goal so the comparison remains useful.
• Adding more detail merely because it is available. Ask whether each detail changes the work and is appropriate to share.
• Restarting before round three. Round three tests deliberate continuation from the clearer request.
• Treating longer as better. Compare fit and support, not word count.
• Smoothing over missing facts. Keep placeholders or questions visible instead of accepting invented details.
• Hunting for universal prompt wording. Remind learners that model behavior varies and the durable skill is inspecting and choosing the next move.

Accessibility and participation options

• Read prompts aloud while also displaying or printing them.
• Let learners type, dictate to a partner, mark a printed transcript, or observe.
• Give additional processing time before group discussion.
• Describe changes in words; do not rely on color or side-by-side vision alone.
• Invite people to pass, use the fictional example, or share only observations rather than personal content.

Debrief

Ask:

• Which detail or boundary materially changed the response?
• What did the chatbot assume or present without support?
• Why did the round-three follow-up belong in the existing conversation?
• What did you decide not to share?
• What would you verify outside the chatbot before using the result?
• What did you keep, reject, or change yourself?

Evidence boundary

Completion shows that a learner can perform and explain this comparison in one setting. It does not establish durable skill, factual correctness, safe use in consequential work, or that one prompt will behave consistently across tools and time. Record pilot observations separately from artifact approval.

Provisional timing for facilitator planning

These inherited estimates are starting points for facilitation, not completion targets for learners. Refine them after observing setup, access needs, discussion, and actual participation.

• Lab: 20–30 minutes (short: 12; extended: 45)