Asking AI for an activity: the right question

Teachers asking AI for activities are often disappointed: the draft comes back generic, age-inappropriate or disconnected from the real classroom. The problem is usually not the tool but the request. AI is only as good as the context it is given, and a context-free request produces a context-free answer.

Five pieces of information

• Age group and group size: an activity for 36–48 months differs entirely from one for 60–72.
• The children's interest: “they've been talking about insects for two weeks” anchors the draft to a real room.
• Time and place: 20 minutes or 45, indoors or in the garden?
• Material constraints: what you have and don't — “paper, paint and natural materials only”.
• Developmental focus: which domain(s) do you want to see?

With those five, the output approaches a usable draft. But a draft is a draft: the final decision is the teacher's. Rather than running the text as-is, put it through three filters.

Three filters

First, the child-centredness filter: is there anything in this activity the child can choose, or is it an adult-determined sequence from start to finish? Unless asked otherwise, models tend to produce teacher-led flows with right answers.

Second, the safety filter: small parts, sharp tools, allergens, swallowable objects. No material list should enter a classroom without an adult reviewing it.

Third, the open-endedness filter: do the questions lead to a single correct answer? Replacing “how many are there?” with “how could we group these?” moves the same activity to another level.

The teacher who gets the best out of AI is not the one who finds the best tool but the one who describes their classroom best. Context comes from the teacher, the draft from the machine; good activities are born where the two meet.