fix(constructs): verify MFQ against the source the review sheet omitted

Checking whether the follow-up questions for the reviewer were fair
turned up that two of them were not.

The review sheet shipped with an EMPTY source_url for mfq_care and
mfq_fairness, so the reviewer had nothing to check those against and
correctly flagged it. That was our omission, not a missing document:
the MFQ30 is public, all four items in each construct match it
verbatim, and both are now verified with the source recorded. This also
settles the "differently than/from others" disagreement in favour of
the wording already stored.

Also confirmed independently, so the reviewer does not need to be asked:
the Subjective Happiness Scale PDF at the source_url we gave her does
read "Compared to most of my peers", exactly as she corrected it.

And for context on the two item-order findings: the APA measures-package
printing of the RSES numbers its items exactly as we store them, while
the reviewer's renumbering matches another widely-used printing. Same
version-mismatch pattern as Edmondson and the MFQ, so it is a question
about which printing to standardise on rather than an error either way.

83 verified (was 81), 11 unverified (was 13).

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

CHANGELOG.md

@@ -7,17 +7,17 @@ Format: [Keep a Changelog](https://keepachangelog.com/). User-visible changes on
 ### Changed
 - Construct library verified against the source publications. All 525 items across the
   94 constructs were reviewed (wording, reverse-scoring keys, subscale grouping,
-  citations); 81 constructs are now marked verified and show who reviewed them and
+  citations); 83 constructs are now marked verified and show who reviewed them and
   when, instead of the blanket "needs verification" flag every construct carried. The
   review corrected 61 reverse-scoring flags across 14 constructs (the library had 35
   flagged items, it now has 96), 7 item wordings, one subscale grouping in the SCS-SF,
   and a batch of dead or incomplete citation links. Corrected constructs ship as a new
   version: the picker offers the corrected one, and runs that used the earlier version
-  still open, export, and reproduce exactly as before. 13 constructs stay flagged, each
-  now saying why: two open wording questions (the IPIP scales and the K10), two where
-  the review's proposed wording disagreed with the cited publication and the published
-  wording was kept, two with unresolved item-ordering questions, and three with no
-  reachable source document on file. (spec 0007)
+  still open, export, and reproduce exactly as before. 11 constructs stay flagged, each
+  now saying why: two open wording questions (the IPIP scales and the K10), one where the
+  review's proposed wording disagreed with the cited publication and the published wording
+  was kept, two with unresolved item-ordering questions, and two whose recorded source
+  document no longer resolves. (spec 0007)
 
 ### Added
 - Runs now warn when every item in a construct is reverse-scored. Scores are raw

README.md

@@ -54,8 +54,8 @@ Configuration: copy `.env.example` to `.env` and fill what you need.
   storage is ephemeral and the instance may reset.
 - The construct library ships the lab's full collection of 94 constructs. Item wording,
   reverse-scoring keys, subscale grouping, and citations were reviewed against the source
-  publications in August 2026: 81 constructs are marked verified and carry the reviewer
-  and review date. The other 13 stay flagged in the interface and each records why - an
+  publications in August 2026: 83 constructs are marked verified and carry the reviewer
+  and review date. The other 11 stay flagged in the interface and each records why - an
   open wording question, a proposed correction that disagreed with the cited paper, an
   unresolved item-ordering question, or no reachable source document. See
   `docs/specs/0007-construct-library-verification-pass.md`.

backend/tests/test_construct_library_versions.py

@@ -175,8 +175,8 @@ def test_review_applied_expected_shape():
         by_status.setdefault(c["verification_status"], []).append(c)
 
     assert len(by_status["archived"]) == 21, "superseded v1 files"
-    assert len(by_status["verified"]) == 81
-    assert len(by_status["needs_verification"]) == 13
+    assert len(by_status["verified"]) == 83
+    assert len(by_status["needs_verification"]) == 11
 
     live = [c for c in constructs if c["verification_status"] != "archived"]
     assert len({c["construct_id"] for c in live}) == 94, "one live version per construct"
@@ -199,7 +199,6 @@ def test_review_applied_expected_shape():
             # PI decision: restoring the shared K10 stem onto each item
             "k10",
             # reviewer's correction contradicts the cited publication
-            "mfq_fairness",
             "team_psychological_safety_scale",
             # reviewer reported an item-ORDER problem that is not applied
             "cbi_work_related_burnout",
@@ -207,7 +206,6 @@ def test_review_applied_expected_shape():
             # no reachable source on record, so "verified" cannot be claimed
             "bas_2",
             "cage_questionnaire",
-            "mfq_care",
         ]
     )
     for c in by_status["needs_verification"]:

docs/specs/0007-construct-library-verification-pass.md

@@ -16,8 +16,8 @@ loadings computed from mis-flagged items are wrong in a way nothing surfaces.
 Noor Skhiri reviewed all 525 items (returned 2026-08-25), checking wording, reverse
 keying, subscale grouping, and citation for each. This spec applies her review.
 
-Two of her findings raise questions that are the PI's to answer, so they are
-deliberately NOT applied here (see Non-goals).
+Some of her findings are not applied here: two need a PI decision, and several turned
+out to be version differences or gaps in the review sheet we sent her (see Non-goals).
 
 ## Contract
 
@@ -44,7 +44,7 @@ append-only means old versions remain resolvable for runs that already used them
 Metadata-level changes do NOT feed `item_hash` and are therefore applied in place on
 the existing version:
 
-- `verification_status` promoted `needs_verification -> verified` for 81 constructs.
+- `verification_status` promoted `needs_verification -> verified` for 83 constructs.
 - Citation and `source_url` repairs (dead links replaced with the URLs the reviewer
   supplied, DOIs she supplied added).
 - A `review:` provenance block on every construct recording reviewer, review date, and
@@ -91,7 +91,7 @@ construct reproduce exactly as before, because v1 rows are never mutated.
 
 ## Non-goals
 
-13 constructs keep `verification_status: needs_verification`, each with the reason
+11 constructs keep `verification_status: needs_verification`, each with the reason
 recorded in its `review.notes`.
 
 **Two of the reviewer's wording corrections are NOT applied, because each disagrees with
@@ -99,7 +99,9 @@ the publication its own construct cites.** Both were checked against the source:
 
 - `mfq_fairness` item 3: she proposed "treated differently from others"; the published
   MFQ30 reads "treated differently **than** others", which is what the library already
-  had.
+  had. This one is resolved: the MFQ30 is now recorded as the construct's `source_url`
+  (the review sheet shipped without one), all four items check out verbatim against it,
+  and the construct is verified.
 - `team_psychological_safety_scale` item 4: she proposed "It is safe to take a risk **in**
   this team", sourced from a third-party questionnaire (novopsych TPS-7); Edmondson
   (1999), the cited paper and the recorded `source_url`, reads "on this team", which is
@@ -117,10 +119,12 @@ feeds `item_hash` and decides what each `sim_item_N` export column refers to, so
 reordering is a scoring-visible change that needs a decision rather than a silent fix.
 Their reverse-scoring corrections ARE applied.
 
-**Three constructs have no reachable source on record** (`bas_2`, `cage_questionnaire`,
-`mfq_care`; `mfq_fairness` too). The reviewer checked what she could and flagged the
-missing links; "verified verbatim against the original publication" cannot honestly be
-claimed without one, so they stay unverified until a working source is on file.
+**Two constructs have no reachable source on record** (`bas_2`, whose recorded URL 404s,
+and `cage_questionnaire`, whose URL returns 403). "Verified verbatim against the original
+publication" cannot honestly be claimed without one, so they stay unverified until a
+working source is on file. `mfq_care` and `mfq_fairness` were in this group because the
+review sheet shipped with an EMPTY `source_url` for both - our omission, not a missing
+document. The MFQ30 is now recorded for each and their items confirmed verbatim.
 
 **Two findings are left pending an explicit PI decision:**
 
@@ -151,7 +155,7 @@ similarities plus flags; `adjustment_strategy` remains a recorded parameter).
   updates the existing row.
 - `test_sync_still_refuses_item_change_under_same_version` - the append-only guard is
   intact.
-- `test_review_applied_expected_shape` - 21 v2 files, 81 verified, 13
+- `test_review_applied_expected_shape` - 21 v2 files, 83 verified, 11
   needs_verification, 96 reverse flags, and every unverified construct carries a recorded
   reason.
 - `test_superseded_files_keep_their_original_items` - compares every tracked construct

packages/construct_library/apply_review.py

@@ -52,7 +52,6 @@ PENDING = {
     "k10",
     # Reviewer's correction disagrees with the cited publication (see below).
     "team_psychological_safety_scale",
-    "mfq_fairness",
     # Reviewer reported an item-ORDER problem, which is not applied here. Order
     # feeds item_hash and decides what sim_item_N means in an export, so it is
     # not a cosmetic change to make without a decision.
@@ -62,7 +61,6 @@ PENDING = {
     # publication" cannot honestly be claimed.
     "bas_2",
     "cage_questionnaire",
-    "mfq_care",
 }
 PENDING_NOTE = {
     "k10": (
@@ -76,11 +74,6 @@ PENDING_NOTE = {
         "safe to take a risk on this team' - the wording already stored here. Kept as "
         "published; the two sources disagree and the citation should decide."
     ),
-    "mfq_fairness": (
-        "Reviewer proposed 'treated differently from others'. The published MFQ30 reads "
-        "'treated differently than others' - the wording already stored here. Kept as "
-        "published. No source_url on file yet."
-    ),
     "rses": (
         "Reviewer maps stored items 1,2,3,6,7,9,10 to source item numbers 7,10,2,9,1,3,6: "
         "the item ORDER does not match the source. Reverse-scoring corrections are applied; "
@@ -103,7 +96,6 @@ PENDING_NOTE = {
         "reviewer's wording correction ('ought to' -> 'should') is applied; a working "
         "source link is still needed before this can be called verified."
     ),
-    "mfq_care": "Citation is correct; no source_url on file yet, so wording is unconfirmed.",
 }
 IPIP_NOTE = (
     "Reviewer confirms wording matches the source apart from the leading 'I' that CCR "
@@ -176,6 +168,10 @@ SOURCE_URL_FIX = {
     "dirty_dozen_narcissism": "https://psycnet.apa.org/fulltext/2010-10892-021.pdf",
     "dirty_dozen_psychopathy": "https://psycnet.apa.org/fulltext/2010-10892-021.pdf",
     "lot_r": "https://psycnet.apa.org/fulltext/1995-07978-001.pdf",
+    # The sheet went out with no source_url for these two, so the reviewer had
+    # nothing to check against. Added afterwards and confirmed item by item.
+    "mfq_care": "https://moralfoundations.org/wp-content/uploads/files/MFQ30.doc",
+    "mfq_fairness": "https://moralfoundations.org/wp-content/uploads/files/MFQ30.doc",
     "collectivism_horizontal": "https://psycnet.apa.org/fulltext/1997-38342-009.pdf",
     "individualism_horizontal": "https://psycnet.apa.org/fulltext/1997-38342-009.pdf",
 }
@@ -211,6 +207,14 @@ QUESTIONNAIRE_FIX = {
 REVIEW_NOTE = {
     "lot_r": "The scale's 4 filler items are correctly excluded from this construct; they are "
              "not scored in the source.",
+    "mfq_care": "Reviewer flagged the missing source_url (the sheet shipped without one for this "
+                "construct). Source added and all 4 items confirmed verbatim against the "
+                "published MFQ30. The construct uses a subset of the foundation's items, which "
+                "the reviewer confirmed as correctly grouped.",
+    "mfq_fairness": "Reviewer flagged the missing source_url (the sheet shipped without one) and "
+                    "proposed 'treated differently from others'. Source added: the published "
+                    "MFQ30 reads 'treated differently than others', which is the wording already "
+                    "stored here, so it is kept and all 4 items are confirmed verbatim.",
 }
 
 

packages/construct_library/constructs/mfq_care.yaml

@@ -7,12 +7,15 @@ category: moral-foundations
 description: Concern with suffering, compassion, and protection of the vulnerable (MFQ Care/Harm).
 citation: Graham, J., Nosek, B. A., Haidt, J., Iyer, R., Koleva, S., & Ditto, P. H. (2011). Mapping the
   moral domain. JPSP, 101(2).
-verification_status: needs_verification
+verification_status: verified
 review:
   reviewer: Noor Skhiri
   date: '2026-08-25'
-  outcome: pending_pi_decision
-  notes: Citation is correct; no source_url on file yet, so wording is unconfirmed.
+  outcome: verified
+  notes: Reviewer flagged the missing source_url (the sheet shipped without one for this construct). Source
+    added and all 4 items confirmed verbatim against the published MFQ30. The construct uses a subset
+    of the foundation's items, which the reviewer confirmed as correctly grouped.
+source_url: https://moralfoundations.org/wp-content/uploads/files/MFQ30.doc
 items:
 - item_id: care_1
   text: Compassion for those who are suffering is the most crucial virtue.

packages/construct_library/constructs/mfq_fairness.yaml

@@ -7,13 +7,15 @@ category: moral-foundations
 description: Concern with justice, rights, and equal treatment (MFQ Fairness/Cheating).
 citation: Graham, J., Nosek, B. A., Haidt, J., Iyer, R., Koleva, S., & Ditto, P. H. (2011). Mapping the
   moral domain. JPSP, 101(2).
-verification_status: needs_verification
+verification_status: verified
 review:
   reviewer: Noor Skhiri
   date: '2026-08-25'
-  outcome: pending_pi_decision
-  notes: Reviewer proposed 'treated differently from others'. The published MFQ30 reads 'treated differently
-    than others' - the wording already stored here. Kept as published. No source_url on file yet.
+  outcome: verified
+  notes: 'Reviewer flagged the missing source_url (the sheet shipped without one) and proposed ''treated
+    differently from others''. Source added: the published MFQ30 reads ''treated differently than others'',
+    which is the wording already stored here, so it is kept and all 4 items are confirmed verbatim.'
+source_url: https://moralfoundations.org/wp-content/uploads/files/MFQ30.doc
 items:
 - item_id: fair_1
   text: Justice is the most important requirement for a society.