Choose an activity that really measures the objective: three cases
A lively activity is not necessarily useful for the stated objective. Asking a group to “discuss the risks” may stimulate ideas; it does not prove that each person can correct a risky document. Start with the expected evidence at the end of the session, then choose practice that produces comparable evidence, with support that gradually decreases.
A four-question filter
-
What exact action does the objective require? Identifying, deciding, writing, correcting and justifying are not interchangeable.
-
What will each person do? Listening and a collective vote do not necessarily show what each person can do.
-
What record will remain? An annotation, a justified decision or a revised document can be reviewed. An impression of participation is not enough.
-
Does the record use the assessment criteria? The final situation may change, but the action and criteria must remain comparable.
Our alignment check for objectives, activities and assessment presents the complete matrix. Here, let us focus on choosing the activity.
Three situations and their evidence
1. Classify anomalies in a case file. Reading a list of categories and then voting on the trainer’s answer proves little. Using a fictional file, each pair classifies three anomalies; each person then writes their decision in two sentences. Keep both the classification and the individual justification.
2. Correct unchecked claims in an AI-assisted text. Watching a correction on screen can introduce the method. Then, working from a short text and its source notes, each person highlights doubtful claims, marks them “confirmed”, “corrected” or “to confirm”, and submits an annotated version.
3. Turn an incomplete brief into a usable instruction. An oral list of “good prompts” does not use a shared brief. Give each participant the same fictional brief: they separate known and unknown information, then write an instruction that asks for the missing elements without inventing them. Our five-question teaching brief can serve as a resource after its publication scheduled for 28 September.
These situations are fictional examples, not test results. The outputs make the action inspectable without guaranteeing success. Pair work may hide an individual difficulty: request a short personal record before concluding that someone has mastered the skill.
Three common selection errors
Confusing talk with evidence. A discussion prepares people to act, but if the objective says “correct”, allow time to correct a piece of material.
Providing the answer before practice. A worked example helps participants get started. For the final task, change the data without changing the announced criteria.
Choosing the tool before the skill. First describe the teaching action and the constraints. AI can then speed up the preparation of a case file or propose variants for the trainer to review.
A quick check before the session
Write one sentence in each box: “objective”, “practice evidence”, “support provided”, “final individual evidence”, “criteria”. If the two records do not show the same action, adjust the activity. If the support is never removed, do not conclude that participants can work independently. If the criteria cannot be applied to the record, rewrite them.
To frame an AI request without losing these checks, the free pack of 25 prompts provides framing and review templates. Always adapt suggestions to the participants, equipment and your organisation’s rules.
