Evaluation of these materials

The pilot report is the landing page for testing these materials with real people — currently pending. The full methods statement (what was evaluated, by what method, by whom, and what each piece of evidence can and cannot show) lives in evaluation/METHODS.md in the repository — summarized here:

Evaluation Method Status
Evidence verification Primary-source tracing (CALDER working paper, NSSA briefs, dosage-analysis press release) — research notes Complete
Automated accessibility axe-core 4.11.1, WCAG 2.1 AA, every page + the practice tool Complete
Keyboard & reflow Playwright 1.57 scripted keyboard-only task tests + 320px reflow + aria snapshots Complete
Design review Agent review against a structured form — two passes: in-session by the authoring agent (2026-10-25), then a separate fresh-context reviewer agent (2026-10-26, agent review, whose findings were applied the same day) Complete
Transcript realism review Agent review of the six fictional transcripts against the cited misconception research, in three passes: in-session (2026-10-25), a fresh-context reviewer (2026-10-26), and a single-agent re-review (2026-11-11, results) Complete (agent passes only; no practitioner review yet)
Simulated tutor-trainees Agent-simulated: one AI agent played 9 trainees (3 personas × 3 levels) on pre/post error-response tasks and scored the 54 responses against the rubric’s Domain 3 anchors itself; single model, not blind, descriptive only (results) Complete (2026-11-11); not human results
Pilot with real people 3–5 adults, pre/post error-response tasks, blind rubric scoring Not run — protocol prepared
Screen reader testing Guidepup/VoiceOver unavailable on the build OS (macOS 27) — no screen reader testing performed; manual protocol prepared Not run — protocol prepared

Honesty labels: every review on this site is labeled “Agent review” — never “educator review” or “expert feedback.” Pending items stay marked pending; nothing is fabricated; no learning gains are claimed anywhere.


Back to top

Content licensed CC BY 4.0; code licensed MIT. Bridge Math Tutoring is a fictional organization; all transcripts are fictional and constructed for training purposes.

This site uses Just the Docs, a documentation theme for Jekyll.