Jev Skill Evaluator

Simulates scoring during a lesson: each teacher turn is rated using only that turn and what came before it, the way the simulator would have to decide how students and the coach respond next.

1Skills

Look-fors that will be checked each turn:
None yet
Click a look-for to skip it. Lesson-wide look-fors (faded) aren't rated live.

What happens each turn

  • Single-utterance / responsive look-fors: rated as in the full run (the turn, plus 2 turns before for responsive).
  • Multi-turn look-fors, two calls in parallel:
    • Where the sequence is: Jev says whether this turn starts or continues a sequence, which of the look-for's observable behaviors it carries out, and whether the sequence is finished. A turn continues a sequence only if Jev names a step (or is over 70% sure). Two teacher turns without one end it.
    • How it's going: a five-band rating: Strong, Attempting, Building (partway, on track), Insufficient, Non-Example.
  • When a sequence completes or ends, it gets one whole-sequence rating.
  • Coach: a turn is ineffective if any look-for is Insufficient with confidence over 50%. After 2 or 3 of them (your setting), the coach steps in. A separate harm check on each turn alone (not tied to any look-for, and the only place harm is rated) over 70% means it steps in right away; 50–70% counts as ineffective. The count resets when the teacher fixes what went wrong (those look-fors come back Strong or Attempting), or after 4 turns with nothing ineffective. Good ratings on unrelated look-fors don't reset it. Look-fors a transcript can't show don't count.
  • Jev never sees the students' reaction to the turn being rated, or anything after it.

2Transcript

Simulator CSV exports import directly. Or paste, one turn per line, like Teacher: Take out your notebooks.

Password