Annotated transcripts
All six transcripts are fictional and constructed for training purposes. They are not recordings of real children; tutor and student names are invented. They are built from documented middle-school algebra misconceptions so that the errors students make in them are the errors real tutors will meet — the misconception research behind that choice is cited in the evidence notes. A structured realism review (agent-based; practitioner review pending) checked each transcript against that research — see evaluation methods.
Why transcripts instead of videos
Videos of real sessions would show real children — a privacy problem, and a consent problem minors can’t meaningfully opt out of. Written fictional transcripts are private by construction, searchable, annotatable, and screen-reader accessible. They are also editable: trainers can adapt a transcript to their local curriculum without re-filming.
The annotation format
Each transcript is a dialogue between the tutor and 2–4 students, broken into exchanges. Numbered annotations follow each exchange — never floating in a margin — and name the move the tutor used (e.g., probing question, wait time, confirm and ask why), which competency domain it serves, and what an observer would look for. Move names and hint-ladder levels (0–5) are shared with the companion Socratic AI Tutor project — same moves, encoded for an AI.
The library
| # | Transcript | Verdict | The one thing it teaches |
|---|---|---|---|
| T1 | Routine holding together | Strong | The first 3 minutes + a closed data loop |
| T2 | A student stuck in the same spot | Strong | Worked example after repeated failure (L4), done right |
| T3 | Questioning strong, data loop broken | Mixed | Good teaching ≠ a good program: the ticket matters |
| T4 | Structure intact, feedback generic | Mixed | “Good job” ×20 teaches nothing; the M5 door opens and closes |
| T5 | The tutor explains everything | Weak | Over-explaining: prepared, warm — and the students never think |
| T6 | The tutor answers for the students | Weak | No wait time, ladder skipped, speed as the prize |
Suggested sequence for new tutors: T1 → T5 → T2 → T6 → then spot-the-move practice, then T3 + T4 when your first observation is scheduled (they show why the rubric scores domains separately).
Using transcripts in training
- Spot-the-move (interactive): practice tool — 10 exchanges from this library; you name the move and whether it worked; feedback explains why. Nothing you answer is stored.
- Live sessions: Session A and Session B facilitator guides use T1/T2/T5/T6 excerpts as role-play material.
- Calibration: coordinators score T1, T3 and T5 against the observation rubric and compare scores — the calibration activity protocol.
Table of contents
- Transcript T1: A strong session — the routine holding together
- Transcript T2: A strong session — a student stuck in the same spot
- Transcript T3: A mixed session — questioning strong, data loop broken
- Transcript T4: A mixed session — structure intact, feedback generic
- Transcript T5: A weak session — the tutor explains everything
- Transcript T6: A weak session — the tutor answers for the students
- Spot the move (practice)