Stats count every review on record: the published copies of the reviews from before the public flip, the new score-table heading and the old dimension label; a record in both places counts once

crossref_deposit.py

@@ -911,7 +911,8 @@ def _push_publications(sub_dir: Path, submission: dict, doi: str, *, log) -> dic
     try:
         import stats as _stats
         out = os.path.join(publications.WEBSITE_REPO, "src", "data", "stats.json")
-        fresh = _stats.compute_stats(config.REVIEWS_DIR)
+        pub = os.path.join(publications.WEBSITE_REPO, "src", "data", "public-reviews")
+        fresh = _stats.compute_stats(config.REVIEWS_DIR, pub)
         try:
             published = json.loads(Path(out).read_text()).get("total_reviewed") or 0
         except Exception:
@@ -920,7 +921,7 @@ def _push_publications(sub_dir: Path, submission: dict, doi: str, *, log) -> dic
             log(f"  crossref: stats snapshot kept: a fresh count of {fresh.get(total_reviewed)} "
                 f"would replace the published {published}")
         else:
-            stats_path = _stats.write_stats(config.REVIEWS_DIR, out)
+            stats_path = _stats.write_stats(config.REVIEWS_DIR, out, pub)
             if stats_path:
                 extra.append(stats_path)
     except Exception as exc:   # the stats page must never block a registration

intake/scripts/test_audit_20260928b.py

@@ -113,8 +113,21 @@ try:
     check(stats.compute_stats(str(rdir)).get("total_reviewed") == 0, "an empty reviews dir counts zero")
     cd = (ROOT / "crossref_deposit.py").read_text()
     g = cd.find("if (fresh.get(\"total_reviewed\") or 0) < published:")
-    w = cd.find("stats_path = _stats.write_stats(config.REVIEWS_DIR, out)")
+    w = cd.find("stats_path = _stats.write_stats(config.REVIEWS_DIR, out, pub)")
     check(0 < g < w, "a fresh count below the published total is never written")
+    pub = tmp / "public-reviews"; pub.mkdir()
+    (pub / "111.md").write_text("---\nrecord_id: 111\nreview_date: 2026-04-28T18:15:05Z\nrecommendation: RECOMMEND\n"
+                                "consensus: divided\n---\n\n### Aggregate scores\n\n| Dimension | Mean | Scores |\n|---|---|---|\n"
+                                "| Domain Fit | 4.6 | 4, 5 |\n| AI Slop Detection | 4.0 | 4, 4 |\n\n### Individual\n")
+    (pub / "111_review_quality_control.md").write_text("---\nrecord_id: 111\nreview_quality_control_flag: false\n---\n")
+    (pub / "ICSAC-SUB-00001.md").write_text("---\nrecord_id: ICSAC-SUB-00001\nrecommendation: REJECT\n---\n")
+    (rdir / "ICSAC-SUB-00001_x.md").write_text("---\nrecord_id: ICSAC-SUB-00001\nrecommendation: RECOMMEND\ndisagreement: False\n---\n")
+    st = stats.compute_stats(str(rdir), str(pub))
+    check(st["total_reviewed"] == 2, "published records count, and a record in both places counts once")
+    check(st["recommendation_mix"]["RECOMMEND"] == 2, "the raw record wins over its published copy")
+    check(st["dimension_means_overall"].get("Domain Fit") == 4.6, "the published ### Aggregate scores table is read")
+    check(st["dimension_means_overall"].get("AI Provenance Signal") == 4.0, "the old AI Slop Detection label counts as AI Provenance Signal")
+    check(st["rqc_audited_count"] == 1, "published RQC flags are read")
 finally:
     shutil.rmtree(tmp, ignore_errors=True)
 

stats.py

@@ -49,16 +49,23 @@ def _parse_frontmatter(text: str) -> dict:
     return out
 
 
+# The raw review writes "## Aggregate Scores"; the published copy writes
+# "### Aggregate scores". Reviews before 2026-05-17 name the last dimension
+# by its old label.
+_AGG_HEADING = re.compile(r"^#{2,3}\s+aggregate\s+scores\b", re.I)
+_LABEL_ALIASES = {"AI Slop Detection": "AI Provenance Signal"}
+
+
 def _parse_aggregate_means(text: str) -> dict[str, float]:
     """Pull dimension → mean from the aggregate markdown table."""
     means: dict[str, float] = {}
     in_table = False
     for line in text.splitlines():
         stripped = line.strip()
-        if stripped.startswith("## Aggregate Scores"):
+        if _AGG_HEADING.match(stripped):
             in_table = True
             continue
-        if in_table and stripped.startswith("## "):
+        if in_table and stripped.startswith("#"):
             break
         if not in_table or not stripped.startswith("|"):
             continue
@@ -67,7 +74,7 @@ def _parse_aggregate_means(text: str) -> dict[str, float]:
             continue
         if set("".join(cells)) <= set("- "):
             continue
-        label = cells[0]
+        label = _LABEL_ALIASES.get(cells[0], cells[0])
         try:
             means[label] = float(cells[1])
         except ValueError:
@@ -144,7 +151,9 @@ def _load_reviews(reviews_dir: str) -> list[dict]:
             {
                 "record_id": rid,
                 "recommendation": fm.get("recommendation", "REVIEW_FURTHER"),
-                "disagreement": fm.get("disagreement", "False").lower() == "true",
+                # Published copies carry "consensus: divided" instead.
+                "disagreement": (fm["disagreement"].lower() == "true") if "disagreement" in fm
+                                else fm.get("consensus", "").lower() == "divided",
                 "review_date": _parse_review_date(fm.get("review_date", "")),
                 "dimension_means": means,
                 "rqc_flag": rqc_flags.get(rid),
@@ -170,8 +179,15 @@ def _histogram(values: list[float]) -> dict[str, int]:
     return bins
 
 
-def compute_stats(reviews_dir: str) -> dict:
+def compute_stats(reviews_dir: str, public_dir: str | None = None) -> dict:
+    """public_dir: the website's published review records. The raw files behind
+    the papers reviewed before the public flip (2026-05-16) are gone from
+    reviews/, so the published copies are the only record of those reviews;
+    a record in both places counts once (raw wins)."""
     reviews = _load_reviews(reviews_dir)
+    if public_dir:
+        seen = {r["record_id"] for r in reviews}
+        reviews += [r for r in _load_reviews(public_dir) if r["record_id"] not in seen]
     now = _dt.datetime.now(_dt.timezone.utc)
     cutoff = now - _dt.timedelta(days=30)
 
@@ -226,8 +242,8 @@ def compute_stats(reviews_dir: str) -> dict:
     }
 
 
-def write_stats(reviews_dir: str, out_path: str) -> str:
-    stats = compute_stats(reviews_dir)
+def write_stats(reviews_dir: str, out_path: str, public_dir: str | None = None) -> str:
+    stats = compute_stats(reviews_dir, public_dir)
     os.makedirs(os.path.dirname(out_path), exist_ok=True)
     with open(out_path, "w", encoding="utf-8") as f:
         json.dump(stats, f, indent=2, ensure_ascii=False)
@@ -250,5 +266,10 @@ if __name__ == "__main__":
             "~/Desktop/icsac/icsacinstitute.org/src/data/stats.json"
         )
     )
-    written = write_stats(rdir, out)
+    pub = (
+        sys.argv[3]
+        if len(sys.argv) > 3
+        else os.path.join(os.path.dirname(os.path.dirname(out)), "data", "public-reviews")
+    )
+    written = write_stats(rdir, out, pub if os.path.isdir(pub) else None)
     print(f"wrote {written}")