cp-j0gw: senate-staff + public-commenter audience extension; Mechanism Library + depth toggle + Sources

content/citations.json

@@ -681,6 +681,13 @@
       "section_heading": "M17 — BD/CSD timelines under OBBBA §§ 85001–85002",
       "file": "content/mdocs/M17.html"
     },
+    "M18#FIVE-PASS-CONVERGENCE": {
+      "mdoc": "M18",
+      "convention": "panel-theme",
+      "title": "§ 668.16(t) institution-level pathway — five-pass convergence (panel)",
+      "section_heading": "M18 — five legal frames each independently testing the institution-level escalation design",
+      "file": "content/mdocs/M18.html"
+    },
     "R12#FIVE-REASONS": {
       "mdoc": "M12",
       "convention": "panel-theme",

content/mdocs/M18.html

@@ -0,0 +1,52 @@
+<article class="mdoc-panel" data-panel-id="M18" data-trigger="auto">
+  <header class="mdoc-header">
+    <span class="mdoc-id">M18</span>
+    <h2 class="mdoc-title">When a program failure could spread to the whole institution. Proposed § 668.16(t) lets program-level failures escalate into an institution-level Title IV finding. Five independent legal frames each give the Department's chosen design a separate problem to answer.</h2>
+  </header>
+
+  <section class="mdoc-summary">
+    <p>Proposed § 668.16(t) treats sustained program-level failure as evidence of an institution-level capacity problem. The proposed mechanism — share thresholds applied across two of three measurement years — is a procedural choice the Department made on top of the program-level statute. Phase 2 of the audit walked the same proposal through five independent legal registers (statutory authority, OIRA Circular A-4 problem statement, A-4 alternatives review, A-4 uncertainty analysis, and the State Farm reasoned-decisionmaking standard). Each register raises a separate, independently sufficient question about the proposal as written.</p>
+  </section>
+
+  <section class="mdoc-mechanism">
+    <h3>The five frames in plain terms</h3>
+    <ol class="mdoc-steps">
+      <li><strong>Statutory authority (Loper Bright + major-questions).</strong> OBBBA § 84001 sets the program-level test. The Department points to § 498 of the Higher Education Act for the institution-level escalation pathway. Loper Bright instructs courts to read the statute themselves rather than defer; the major-questions canon adds skepticism whenever an agency claims an unusual or large new power inside a smaller statutory authority. Whether the program-level statute authorizes an institution-level Title IV consequence is a question the rule's text alone does not settle.</li>
+      <li><strong>Problem statement (Circular A-4 § 1).</strong> The preamble describes the institution-level mechanism with words like "would suggest" and "could represent." A-4 § 1 asks for a demonstrated problem, not an asserted one — empirical evidence that the institution-level escalation pathway addresses a real and otherwise-unaddressed harm. The record as published does not show that demonstration.</li>
+      <li><strong>Alternatives review (Circular A-4 § 6).</strong> The preamble considers and discards alternatives — disclosure-only, threshold variations, phase-in — but does not provide comparable benefit-cost tables for any of them. A-4 § 6 expects each alternative considered to be analyzed at the same depth as the chosen design.</li>
+      <li><strong>Uncertainty (Circular A-4 § E).</strong> The 50% threshold and the 2-of-3-year window have no published sensitivity bounds. The published-flag-publication gap (cells whose pool-level fail flag the Department does not publish) is not analyzed. A-4 § E asks for sensitivity analysis around the choices that swing the rule's outcome.</li>
+      <li><strong>Reasoned decisionmaking (Motor Vehicle Mfrs. Ass'n v. State Farm, 463 U.S. 29).</strong> State Farm's four prongs — relevant data considered, satisfactory explanation given, judgment supported by the record, alternatives addressed — form the floor for a court's arbitrary-and-capricious review. The institution-level escalation pathway design as published does not, on its face, meet all four prongs.</li>
+    </ol>
+    <p>None of these five frames is by itself a verdict. Each is a separate legal hook for stakeholders, public commenters, or reviewing courts to test the proposed § 668.16(t) design.</p>
+  </section>
+
+  <section class="mdoc-summary">
+    <h3>How this might sound in plain language</h3>
+    <p>If a program at this institution fails the earnings-premium test in the first measurement year, the rule's program-level consequences attach to that program (M05 / M07). The institution-level § 668.16(t) institution-level escalation pathway is a separate machinery: if a sustained share of the institution's recipients or aid dollars are in failing programs across multiple years, the Department could find the institution itself out of compliance. Whether the rule as written can support that escalation — given that the underlying earnings-premium statute is written program-by-program — is the load-bearing legal question this panel exists to surface.</p>
+  </section>
+
+  <section class="mdoc-citations">
+    <h3>Sources</h3>
+    <ul>
+      <li><a class="cite" data-cite="prop:668.16(t)" href="#">Proposed 34 CFR § 668.16(t)</a> — institution-level escalation pathway text.</li>
+      <li><a class="cite" data-cite="obbba:84001" href="#">OBBBA § 84001</a> — program-level earnings-premium test (Public Law 119-21).</li>
+      <li><a class="cite" data-cite="hea:498" href="#">HEA § 498</a> — administrative-capability authority cited by the Department for the institution-level escalation pathway.</li>
+      <li><a class="cite" data-cite="oira:a-4-2003" href="#">OIRA Circular A-4 (2003)</a> — § 1 problem statement, § 6 alternatives review, § E uncertainty analysis.</li>
+      <li><a class="cite" data-cite="case:loper-bright" href="#">Loper Bright Enterprises v. Raimondo, 603 U.S. ___ (2024)</a> — judicial reading of statutes; agency interpretation no longer receives Chevron deference.</li>
+      <li><a class="cite" data-cite="case:state-farm" href="#">Motor Vehicle Mfrs. Ass'n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29 (1983)</a> — four-prong reasoned-decisionmaking standard.</li>
+      <li><a class="cite" data-cite="m18#five-pass-convergence" href="#">M18 — five-pass convergence walk-through</a> — Phase 2 audit detail (analyses/phase-2-nprm-audit/REPORT.md § 2).</li>
+      <li><a class="cite" data-cite="m05#program-level-consequence" href="#">M05 — program-level Title IV consequence</a> — the program-level mechanism this panel sits alongside.</li>
+    </ul>
+  </section>
+
+  <footer class="mdoc-footer">
+    <p class="mdoc-disclaimer">
+      <strong>About this panel.</strong>
+      Written by educators, not attorneys or legislative analysts. Each of the
+      five frames is a working hypothesis about a published proposed rule, not
+      a legal opinion. Verify every statutory citation, case citation, and
+      OIRA-Circular reference against the primary source before any external
+      submission, public comment, or institutional decision.
+    </p>
+  </footer>
+</article>

docs/cp-j0gw-ext-changelog.md

@@ -0,0 +1,154 @@
+# cp-j0gw — extension to cp-0on (CSU Deans EP Tool)
+
+Successor epic to closed `cp-0on`. Extends the public web tool for two new
+audiences (legislator/staff and public commenter), folds in post-2026-04-30
+analytical updates, and adds the instructive surface (Mechanism Library +
+reading-depth toggle + Public Sources page) the chair specified mid-flight.
+
+Branch: `cp-0on-ext-senate-plain-lang`. Acceptance gate (CSULB UNITID 110583
+reproduces dean memo v10 §1 EXACTLY) preserved throughout.
+
+## Stacked PR record (2026-05-05, refreshed end-of-day)
+
+**Six PRs open** on `github.com/AJBcoding/csu-deans-ep-tool`. They stack cleanly when merged in order with one trivial conflict resolution on `web/index.html` (3 persona-cards merge in sequence — see "Smoke-stack record" below).
+
+| PR | Branch | Bead | Author | Tests after merge |
+|---|---|---|---|---|
+| **#1** | `cp-0on-ext-senate-plain-lang` | cp-j0gw (epic main) | crew/anthonybyrnes | 203 |
+| **#3** | `cp-j0gw-8-systemwide-context` | cp-j0gw.8 + .13 | crew/anthonybyrnes | 212 |
+| **#2** | `polecat/obsidian/cp-j0gw.11` (+ cp-j0gw.17 lint follow-up) | cp-j0gw.11 + .17 | polecats/obsidian | 224 |
+| **#4** | `cp-j0gw.14-r19-suppressed-earnings-annotation` | cp-j0gw.14 | polecats/topaz | 228 |
+| **#5** | `cp-j0gw.15-plain-lang-explainers` | cp-j0gw.15 (M01/M03/M18) | polecats/quartz | 249 |
+| **#6** | `cp-j0gw.18-plain-lang-explainers-batch2` | cp-j0gw.18 (M04/M07/M14) | polecats/quartz | **270** |
+
+Suggested chair merge order: **#1 → #3 → #2 → #4 → #5 → #6** (matches dependency stack; tests pass at every step).
+
+PR #2 now carries the cp-j0gw.17 lint fix (commit `38e236c`) so the broader lint scan is clean across all 6 PRs once stacked. **No remaining `OPEID6` or `cascade` literals** on dean-facing surface; the only informational lint findings are formal cite names (`STATS NPRM`, `RIA Table 3.19`) which are preserved per existing M-doc convention.
+
+## Sub-bead ledger
+
+| Bead | Type | Title | Status |
+|---|---|---|---|
+| cp-j0gw.1 | analytical | English MA (CIP 23.01, n=29, −4.97%) — REV-2 third CSULB direct-fail | ✅ done (PR #1) |
+| cp-j0gw.2 | engine | derivation_basis field — explicit AHEAD-published-flag-override (cp-wssr) | ✅ done (PR #1) |
+| cp-j0gw.3 | framing | simulation_framing field — "FORWARD SIMULATION; first STATS release 2027-07-01" | ✅ done (PR #1) |
+| cp-j0gw.4 | persona | Persona D (legislator/staff) | folded into Mechanism Library + Brief depth |
+| cp-j0gw.5 | persona | Persona E — Generate Comment Outline (DRAFT) | ✅ done (PR #1) |
+| cp-j0gw.6 | safety | "not attorneys / not legislative analysts" + per-finding verification recipe | ✅ done (PR #1) |
+| cp-j0gw.7 | content | M-doc M18 — § 668.16(t) five-pass convergence | ✅ done (PR #1) |
+| cp-j0gw.8 | context | 3-cut systemwide ("this CIP fails at N campuses") | ✅ done (PR #3) |
+| cp-j0gw.9 | content | Mechanism Library /learn/ standalone instructive surface | ✅ done (PR #1) |
+| cp-j0gw.10 | ux | Reading-depth toggle (Brief / Standard / Detail) | ✅ done (PR #1) |
+| cp-j0gw.11 | interactive | Interactive M01 cohort-floor demo widget | ✅ done (PR #2 — obsidian polecat) |
+| cp-j0gw.12 | safety | Public-data badge + dedicated /sources page | ✅ done (PR #1) |
+| cp-j0gw.13 | content | Tool README v1.1 status block + routes table | ✅ done (PR #3) |
+| cp-j0gw.14 | engine | R19 suppressed-earnings annotation — 4th provenance value | ✅ done (PR #4 — topaz polecat) |
+| cp-j0gw.15 | content | Plain-language standalone explainers M01/M03/M18 | ✅ done (PR #5 — quartz polecat) |
+| cp-j0gw.16 | qa | Lighthouse + axe-core audit | deferred — runs against preview URL after PR #1 merges |
+| cp-j0gw.17 | lint | Replace `cascade` literal in widget link text | ✅ done (PR #2 commit `38e236c` — obsidian polecat) |
+| cp-j0gw.18 | content | Plain-language standalone explainers M04/M07/M14 (batch 2) | ✅ done (PR #6 — quartz polecat) |
+
+## Chair-review track applied (this evening)
+
+Before the polecat PR returns landed, the chair-review track was iterated inline:
+
+- **Review #1 — disclaimer wording** (envelope + top-aside): approved as-is.
+- **Review #2 — systemwide context one-liner**: refined — empty-title bug fixed (whitespace collapse + omit parenthetical when missing), single-campus suppress gate added (`n_unique_campuses >= 2`), source line hyperlinked to `/sources.html` instead of monospace path. Landed in PR #3 commit `6f8d526`.
+- **Review #3 — Mechanism Library topic groupings**: applied — M07 duplication fixed (was in 2 groups), M02 promoted to "How the rule works", group 3 renamed "Edge cases, timing, and who's hit hardest", group 4 renamed "Process, appeals, and authority". Landed in PR #1 commit `7f6dfe7`.
+- **Review #4 — M18 panel prose**: skipped (chair direction); BUG FIX shipped regardless — 5 doubled fragments `"institution-level institution-level escalation pathway"` stripped (glossary-substitution artifact). Landed in PR #1 commit `9c68b91`.
+- **Review #5 — sample comment outline**: not yet run.
+
+## Smoke-stack record (2026-05-05 evening)
+
+All 5 PRs were stacked on top of `origin/main` in temporary branch `smoke-stack` to verify mergeability:
+
+```
+main → +PR#1 (203 tests)
+     → +PR#3 (212 tests)
+     → +PR#2 (212 tests, 1 trivial conflict on web/index.html — both branches added persona-cards at same insertion point; resolution = keep both, in sequence)
+     → +PR#4 (228 tests)
+     → +PR#5 (249 tests)
+```
+
+Production lint scan (`npx tsx scripts/lint-glossary.ts content web`) caught 3 dean-facing literal violations on the cp-j0gw extension surface (separate from the regression test, which only lints `content/mdocs/`):
+
+- `src/engine.ts` verification_recipe: 4 uses of forbidden literal "OPEID6" — fixed in PR #1 commit `0993045`
+- `web/sources.html` PPD source-card: 2 uses of "OPEID6" — fixed in PR #1 commit `0993045`
+- `web/js/comment-draft.js` outline scaffold: 1 use of "cascade" — fixed in PR #1 commit `0993045`
+
+PR #2 (obsidian widget) carries one similar lint violation (link text "M01 — Cohort visibility cascade" uses forbidden literal "cascade"). Surfacing as a follow-up note rather than amending obsidian's PR unilaterally; can be patched in a 1-line follow-up commit if the chair wants the broader-scope lint clean before merge.
+
+Formal NPRM/RIA/citation names like "STATS NPRM 2026-07666" and "RIA Table 3.19" are retained on the Sources page even though they trip the broader lint — these are primary-source titles, not prose; the existing M-doc panels follow the same convention.
+
+## What changed in this branch
+
+### New pages
+- `web/learn.html` — Mechanism Library (all 17 panels organized into 4 topic groups; depth toggle)
+- `web/sources.html` — every primary source the tool reads or cites
+- `web/comment-draft.html` — DRAFT comment outline generator
+
+### New panels
+- `content/mdocs/M18.html` — § 668.16(t) institution-level pathway, five legal frames
+
+### Engine extensions
+- `src/types.ts` — `DerivationBasis`, `VerificationRecipe` types; envelope adds `simulation_framing` + `expertise_disclaimer`
+- `src/engine.ts` — `deriveDerivationBasis()` + `buildVerificationRecipe()`
+- `src/citations.ts` — `SIMULATION_FRAMING`, `EXPERTISE_DISCLAIMER` constants; `M_DOCS.M18`
+- `src/api/contract.ts` — `PANEL_IDS` (17), `AUTO_TRIGGERED_PANELS` (9, adds M18)
+- `src/api/schema.ts` — required fields for new envelope and verdict additions
+
+### UI affordances
+- Top-disclaimer aside on every page: "not attorneys, not legislative analysts; forward simulation"
+- Public-data badge in every page header
+- Header nav: Mechanism Library / Sources / Comment outline (DRAFT)
+- Reading-depth toggle (Brief / Standard / Detail) — localStorage-persisted
+- Per-verdict collapsible verification recipe with engine reference and step list
+
+### Tests
+- 11 → 16 acceptance test cases (English MA, derivation_basis, recipes, M18 auto-trigger, simulation_framing, expertise_disclaimer)
+- All other test sites updated for 17-panel manifest
+- 203 / 203 passing (was 197 baseline)
+
+## 5-pass Ralph review (this branch)
+
+The chair instruction was to "pass the existing design through several 5-pass
+ralph loops to help iterate the integration." This branch was iterated through
+one full 5-pass cycle. Snapshots were committed per substantive change rather
+than per critique pass, on the rationale that pass-per-commit would clutter
+the ledger. The five passes:
+
+| Pass | Dimension | Critique → action |
+|---|---|---|
+| 1 | analytical accuracy | English MA missing from acceptance test (D1); AHEAD-flag-override behavior not documented (D2); cp-dw65 simulation framing absent (D3). Fixed in commits aa3f850, 6d9a478, f8b9767. |
+| 2 | plain-language clarity | Tool was dean-only; Mechanism Library + Brief reading depth open the surface to senate staff and the public. Fixed in commit 79cf924 + Mechanism Library commit. |
+| 3 | disclaimers & verification | Single-string footer reminder insufficient for new audiences. Added 3-layer disclaimer (top aside + envelope expertise_disclaimer + per-finding verification recipe). Fixed in commits 6d9a478 + f8b9767. |
+| 4 | comment-basis extension | No outline generator; commenters had no scaffold. Added /comment-draft/ that emits structured markdown with bracketed [you-fill-in] prompts, never pre-written legal arguments. Commit 79cf924. |
+| 5 | integration / regression | All 17 panels in manifest; M18 auto-triggers on FAIL; 203/203 tests; vitest + lint-glossary clean; smoke test confirms API includes new envelope fields and routes resolve. |
+
+## Deferred items
+
+- `cp-j0gw.4` — Persona D as a separate page. The Mechanism Library + Brief
+  depth covers most of the use case; revisit after stakeholder review.
+- `cp-j0gw.8` — 3-cut systemwide context ("this CIP fails at N campuses").
+  Requires the CSU systemwide-exposure CSVs (`analyses/csu-systemwide-exposure/output/3cut-2026-05-04/`)
+  to be ingested by the Python build pipeline as a static `systemwide-cip-context.json`
+  bundle. New sub-bead recommended.
+- `cp-j0gw.11` — Interactive widget for one mechanism. Bonus item; defer.
+
+## Acceptance gate verification
+
+The CSULB UNITID 110583 acceptance test (dean memo v10 §1) PRESERVED:
+
+```
+✓ Music MM (5009/Master's)        FAIL @ -32.65%, n=28
+✓ Art MFA (5007/Master's)         FAIL @ -20.40%, n=23
+✓ English MA (2301/Master's)      FAIL @ -4.97%, n=29   [NEW — REV-2 cp-j0gw.1]
+✓ Theatre BA (5005/Bachelor)      PASS @ +6.06%, n=81 — within R19 noise band
+✓ Cinematic Arts BA (5006/Bach)   NOT MEASURED  (B16 invisibility)
+```
+
+## Deploy gate
+
+This branch DOES NOT trigger Cloudflare Pages production deploy. Branch was
+pushed and PR opened for chair review. After chair greenlight, merging to
+`main` triggers the Pages production build via `.github/workflows/deploy.yml`.

src/api/contract.ts

@@ -140,13 +140,13 @@ export type AnalysisPostResponse = AnalysisGetResponse;
 
 export const PANEL_IDS = [
   'M01', 'M02', 'M03', 'M04', 'M05', 'M06', 'M07', 'M08', 'M09',
-  'M10', 'M11', 'M12', 'M13', 'M14', 'M15', 'M17',
+  'M10', 'M11', 'M12', 'M13', 'M14', 'M15', 'M17', 'M18',
 ] as const;
 
 export type PanelId = (typeof PANEL_IDS)[number];
 
 export const AUTO_TRIGGERED_PANELS = [
-  'M01', 'M03', 'M04', 'M05', 'M07', 'M12', 'M13', 'M14',
+  'M01', 'M03', 'M04', 'M05', 'M07', 'M12', 'M13', 'M14', 'M18',
 ] as const satisfies readonly PanelId[];
 
 export const LEARN_MORE_PANELS = [