EDU Prompt Lab by La EJE for designing educational prompts with pedagogical judgement and artificial intelligence
EDU PROMPT LAB

From teaching intention to powerful prompts.

Create educational prompts step by step by defining your intention, your group and your evidence first. Use this tool to think better and direct AI with pedagogical judgement, not to outsource your decisions.

1. What do you want to activate?

Before designing the activity, decide what observable action you want to see.
If you write something generic, AI will give you something generic.

Do not write “learn about…”. Write an action you can observe: compare, justify, detect, apply, create, revise, transfer. Hint: start with an observable verb. If it cannot be seen, reviewed or discussed, it is still too abstract.
A submitted product is not enough. Look for evidence of quality, improvement, decision-making, application or transfer. Hint: evidence should show quality, improvement or decision-making. A polished output is not enough.
See evidence examples
Weak: “They will show me their prompts.” Minimum: “They will show an initial prompt and an improved version, explaining what they changed and why.” Strong: “They will submit a prompt for a real task, including context, constraints, quality criteria and one next iteration after reviewing the AI response.”
Before moving on: does this evidence show learning, or just completion?

2. The reality AI cannot ignore

Time, group size, technology and constraints change the design. AI does not know your classroom or context. You need to give it that reality.

Include age, stage, approximate number, attention level, group energy or educator profile.
Write the content, problem or situation you want to activate.
A hint is enough: comparison, online challenge, cards, response audit, exhibition, prototype... If you do not know yet, say it.
Time, modality, resources, technology, space, participation or attention limits.
Think of real risks: boredom, AI dependency, impossible logistics, flashy but empty activity. Hint: name what you want to avoid. AI often repeats clichés if you do not set limits.

3. What do you need from AI right now?

You are not choosing the type of activity. You are choosing the kind of help you need now.

Key idea: AI will use an expert role to be more useful, but you are only choosing the help you need at this moment. You can change the type of help later.
Do not ask for everything. Ask for the next useful step: ideas, review, sequence, material, adaptation or criteria.

4. Ask, audit and iterate

The prompt is only the beginning of the conversation. Copy it into your AI tool, come back with the response and review it before accepting it.

Instruction: this prompt works with any generative AI tool you have access to. You can paste it into ChatGPT, Claude, Gemini, Copilot or another LLM. It does not depend on a specific platform.
Before sending: read it once. What would you change before letting AI answer?
Come back here with the AI response. Do not accept it blindly: use these filters to decide what to keep, correct or discard.
1. Is it verifiable? Are there claims, data, sources, authors or examples that need checking or traceability?
2. Is it contextual? Does it fit your group, level, available time, resources and real purpose?
3. Is it precise? Does it clearly define what the group will do and what evidence would show learning, improvement or decision-making?
4. Is it justified? Does it have pedagogical sense, or is it only an entertaining activity? Does it activate real thinking?
5. What risk is hidden? Could it contain bias, errors, omissions, dangerous simplifications or false certainty?
6. What do I decide? What will you use, correct, discard or refuse to outsource? Does it help you decide or give you a closed product?
Evidence and learning
Scope and length
Classroom reality
Verification
Copy this prompt to improve the response:
Criteria first. Format second. AI can organise the container, but it must not decide the pedagogy.
Save context in this browser

You can save the information you have written on this device so you do not need to start from scratch next time. It is stored only in this browser on this computer. It is not sent to La EJE, shared with anyone or transferred outside your device. As a general rule when working with AI: never include identifiable personal data about yourself or your learners.