Course workbench (Phase 2 prototype)
What it is, how to build it, and how to run the review session with a course director
A single self-contained HTML file per course-year, for that course’s director. It opens from a laptop, SharePoint or email with no server or login. Search runs in the browser, and file links go to SharePoint, so the usual permissions apply.
just workbench PCV AY25-26 # → site/workbench-PCV-AY25-26.html (~9 MB)
Warning
The file embeds the extracted text of every slide in the course (that’s how search works offline). Treat it like the course materials themselves: share it only with people who already have access to that course’s SharePoint library. Generated workbenches are gitignored for the same reason.
Views
| Tab | What it shows | Evidence behind it |
|---|---|---|
| Overview | Events, learner/scheduled hours, objectives, Bloom level as written, objective flags (click to filter) | Course Build workbooks; ADR 0003 |
| Schedule | Week-by-week sessions with time, method, objective and file counts, flagged objectives. Click a session for its objectives, evidence slides and files. | Planning and build sheets; file links with the linking method |
| Objectives | Every objective, filterable by week, Bloom level, flag and text. Each expands to: its best evidence slides; the share of key terms found in the session’s material; the MEPO → Foundational Competency / EPA chain. | scripts/alignment.py; crosswalks |
| Search | Keyword search over all slide text and speaker notes. Results are grouped by session and can be ordered by date (“first introduced → revisited”). Duplicate slides collapse, and role tags show (answer key, restricted, facilitator…). | BM25 in the browser |
| Matrix | Weeks × USMLE system / MEPO / chief concern / discipline, as objective counts. Click a cell for its objectives. | Objective tags |
| Change vs last year | For each session, slides unchanged / added / removed compared with the same-titled session last year, with examples of added text | Slide text comparison |
| AMC vs FCB | For each session, the share of AMC slides with a close or exact match in FCB’s materials | Slide text comparison |
| Data quality | This course’s items from the backlog DQ table | Source records |
Objective flags
These are prompts for review. Every flag links to the evidence behind it.
| Flag | Meaning | Caveat |
|---|---|---|
| repeated | The same text appears in 3+ different weeks, usually a recurring format (chief concern, VISTA, DOCS skills). Excluded from counts. Repeats within one week (prework → cases → small group) count normally. | Some may be intentional skill repetition, e.g. hand hygiene. |
| no material | Nothing for this session is in the library. | DOCS and H&S small-group materials probably live in Canvas. |
| restated only | The only matching pages restate the objectives. | |
| low overlap | Fewer than half of the objective’s key terms appear anywhere in the session’s material. | The teaching may be in images, video or the room. |
| no practice found | Labelled Analysis or above, and no question or case slide on its content. | Higher-order work often happens in discussion. |
| verb ≠ label | The Bloom label differs from the usual level of the leading verb. | Our verb list is one reasonable reading. |
Numbers for PCV AY25-26 (build of 2026-09-26)
- 276 events (both campuses), 402 files, 11,767 slides/pages.
- 735 objectives (691 + 44 repeated).
- AMC: 171 learner hours (174 h scheduled).
- Flags: no material 95 · low overlap 30 · no practice found 50 · verb ≠ label 61.
- Change vs AY24-25: 104 sessions compared. 6 titles are new; 7 titles weren’t repeated.
- AMC vs FCB, 91 sessions:
- 41 separate builds;
- 12 partly shared, 4 largely shared;
- 33 with no FCB material, 1 with no AMC material.
Review session with the course director (about 45 min)
Goal: learn what is useful, what is wrong, and what is missing. It is not a demo. Record answers in docs/research/pilot-feedback.md.
- Orientation (5 min). State the ground rule: evidence, not scores. Show where every number comes from.
- Their questions first (10 min).
- “What would you want to know about your course that you can’t easily find today?”
- Try those in Search and Objectives. Note misses.
- Walk the views (15 min). For each tab: “Is this right? Is it useful? What would you do with it?” Particular checks:
- Hours: do they match their own understanding?
- Five “low overlap” objectives: real gap, taught elsewhere, or a matching failure?
- Five “no practice found” objectives: where does the higher-order work actually happen?
- Change vs last year: does it match what they revised?
- AMC vs FCB: is the divergence intended?
- Trust (5 min). “Which of these would you be comfortable showing a curriculum committee? What would make you distrust it?”
- Missing (5 min).
- “Where do the DOCS and H&S small-group materials live?”
- “What else should be here?” (exam blueprint? student feedback?)
- Wrap (5 min). Could they use this while revising a session? That is the lead-in to the lesson-design assistant (IDEAS §14).
Known limitations
- Word matching, not meaning: synonyms and images defeat it. Embeddings are deferred (Phase 1 memo).
- Objective sheets and schedules sometimes disagree on titles (DQ1). Six PCV objectives have no week.
- Year-over-year and campus comparisons use exact or near-exact slide text. A reworded slide counts as removed plus added.
- Deep links go to the file, not the slide. The page number is shown next to each link.