After the Count / Review status and human-pilot guide

Review status and human-pilot guide

Development reviews used independent AI-simulated student, faculty, design and learning perspectives. These are design and source critiques, not enrolled students or real faculty research participants. Browser and automated checks establish specific technical behavior. They do not measure learning gains.

A practical human pilot

  1. Recruit learners with varying prior knowledge and document consent and course arrangements under your institution’s rules. Use task instructions only, without a walkthrough that explains the intended answers.
  2. Ask each learner to start, state what the task asks, find a definition, complete a case, operate two settings, save a JSON, restore it and produce a readable export. Note the exact screen, instruction and observed behavior for confusion.
  3. Keep the first unassisted explanation, any requested hints or worked examples, and a later independent transfer response distinct. Do not treat a revealed example as independent success.
  4. Record actual active time by case, pauses and export/restore failures. Report the range and context, rather than presenting the authoring estimate as measured time.
  5. Have faculty examine rule accuracy, denominators, model boundaries, rival claims and evidence-to-recommendation reasoning. Compare rubric judgments before changing the task.
  6. Revise avoidable friction while retaining the need to calculate, explain, compare and defend a cost. Re-test the exact changed tasks.

Review reports

The publication edition adds the new cover and the two use routes after those reports. Its verification record accompanies the complete release. All four election datasets and calculation functions remain unchanged.