Prepare questions for a coaching conversation

Where the evidence comes from

• PBLWorks Gold Standard teaching practices
• EEF Metacognition and Self-Regulated Learning

Strong for teacher coaching and metacognition; AI benefit unvalidated.

1. Confirm the conversation evidence

   Bring teacher-selected, non-identifying evidence, excluding rare identifying combinations. Remove personal data before using any AI tool. Confirm observations yourself before continuing; do not ask AI to diagnose students.

   Prompt 1 · Evidence check Copy

   CONTEXT
   Driving question and current goal: [paste them].
   Observed team evidence, without names: [paste it].
   Possible misconception or stuck point: [add it as your hypothesis].
   Next decision the team owns: [name it].
   
   REQUEST
   Separate observations from teacher inferences and list missing information needed before the conversation.
   
   QUALITY BAR
   Use only the supplied evidence.
   Do not infer motivation, ability, emotion or intent.
   Keep the next decision with the team.
   
   FORMAT
   Use a table with observation, teacher inference, evidence gap and teacher check.

2. Generate a question sequence

   Ask for questions that move from explanation to evidence, monitoring and the team's next choice.

   Prompt 2 · Question sequence Copy

   CONTEXT
   Checked observations and evidence gaps: [paste them].
   Driving question and current goal: [paste them].
   Next decision the team owns: [name it].
   Conversation time available: [add it].
   Teacher's hypothesis and instruction threshold: [paste them; label hypotheses as unconfirmed].
   
   REQUEST
   Draft three alternative sequences, each usable within the supplied conversation time. Show all three in full; do not select one or omit alternatives. Include open, probing and metacognitive questions.
   
   QUALITY BAR
   Do not embed an answer or assume a fact.
   Include questions that help the team plan, monitor and evaluate.
   Do not transfer the next decision to the teacher or AI.
   
   FORMAT
   Three labelled tables A, B and C, each with four complete questions: opening, probe, monitor/evaluate, next decision. Columns: exact question | purpose | listen for | time allocation. Each sequence must fit the total time. Add the supplied instruction threshold separately or mark teacher decision required.

3. Audit hidden steering

   A question can sound open while pushing students toward the teacher's preferred answer. Check before using it.

   Prompt 3 · Steering audit Copy

   CONTEXT
   All three complete question sequences: [paste them].
   Teacher-checked observations and evidence gaps: [paste them].
   Teacher's current hypothesis: [state it as unconfirmed].
   Decision that belongs to the team: [state it].
   Teacher's direct-instruction threshold: [paste an observable condition or missing].
   
   REQUEST
   Audit every question across all three sequences for leading wording, false choices, hidden assumptions and solving the task. Check narrowing lists and words such as "yet" that may smuggle in the hypothesis.
   
   QUALITY BAR
   Quote the exact wording that creates each problem.
   Do not replace the questions.
   Mark where direct instruction would be clearer than another question.
   
   FORMAT
   Twelve-row table: sequence/question ID | exact question | steering risk or none found | exact observed evidence or absence | teacher action (retain/reject/review).
   Separate direct-instruction line: exact supplied threshold or missing | why explanation may be clearer | teacher confirmation required. Do not claim the condition has occurred. Missing threshold blocks card preparation.

4. Make the pocket card

   Reject steering yourself. Select and edit four questions, explicitly approve each exact string and confirm the instruction threshold. Without that checkpoint, stop. Keep teacher edits separate from raw outputs; fictional test approvals are never real approval.

   Prompt 4 · Coaching card Copy

   CONTEXT
   Teacher-approved questions labelled opening, probe, monitor/evaluate and next decision: [paste exact strings].
   Teacher selection/edit and approval record: [paste it or missing; label synthetic records TEST-ONLY].
   Driving question: [paste it].
   Conversation goal and time: [add them].
   Teacher-selected observed evidence: [paste it].
   Teacher's unconfirmed hypothesis: [paste it].
   Team-owned next decision: [paste it].
   Evidence to listen for: [add it].
   Teacher-confirmed conditions for direct instruction or stopping: [add them or missing].
   
   REQUEST
   Arrange the questions as a concise coaching card.
   
   QUALITY BAR
   Copy each approved question character for character, including punctuation. Do not shorten, reorder words or replace phrases. Do not add questions.
   If approval, a required question or the confirmed threshold is missing, return blocked with the missing inputs only.
   Do not add advice, answers or a judgement about the team. Never upgrade a hypothesis to an observation.
   Leave blank space for teacher follow-up. Mark any TEST-ONLY input prominently; no claim of real human approval.
   
   FORMAT
   Context header: driving question | current goal/time | observed evidence | unconfirmed teacher hypothesis | team-owned next decision.
   Four labelled blocks: opening, probe, monitor/evaluate, next decision. Each has the exact approved question and a separate listening note based only on supplied input.
   Footer: exact confirmed direct-instruction/stopping condition; blank teacher follow-up space.
   Verification table: approved input question | copied card question. Quote both completely for teacher comparison.

Check before you use it

• Is every question grounded in something you observed?
• Does any question contain the answer you want?
• Can students act on the answer during this project?
• Does the sequence help students plan, monitor or evaluate?
• Have you decided when a clear explanation is better than another question?
• Does the final decision still belong to the team?