{ "id": "plate-xxii", "number": "XXII", "title": "The Higher Court", "act": { "number": "VI", "title": "The Higher Court" }, "corpus_scope": "56,736 meta-reviewer units (90% from 2024–26) · 6,547 borderline 2026 panels · association, not cause", "question": "What does the area chair weigh differently from the panel it summarizes, and what does the bench forgive at the margin?", "page_url": "https://atlas-of-judgment.pages.dev/", "figures": [ { "id": "fig-22a", "title": "Two grammars: what the panel inspects, what the chair weighs", "deck": "The chair does not re-litigate evidence — it weighs verdicts: statistics, positioning, novelty.", "dom_host": "#court-slope", "claims": [ { "id": "22a-meta-units-count", "statement": "56,736 meta-reviewer units, nine tenths of them from 2024–26, joined at the margin to 6,547 borderline 2026 panels.", "value": { "quantity": 56736, "unit": "count", "share_2024_2026": 0.9, "n_borderline_panels": 6547 }, "source_island": "panel-data.json", "source_path": "meta[]", "derivation": "raw count of extracted meta-review logic units, n = 56,736 (PANEL.meta.meta_reviewer.n)", "recompute": "scripts/build_panel_data.py", "dom_ref": null, "verified": "checked against shipped JSON, 2026-08-25", "caveat_refs": [] }, { "id": "22a-rising", "statement": "Statistics and metrics grow from 10% of reviewer attention to 26% of the chair's, related work from 5% to 16%, novelty doubles — the objects a chair can rule on from the reviews alone.", "value": { "quantity": [0.10, 0.26, 0.05, 0.16], "unit": "share_of_units_reviewer_to_ac", "stats_metrics": [0.10, 0.26], "related_work": [0.05, 0.16] }, "source_island": "court-data.json", "source_path": "grammars[]", "derivation": "COURT.grammars.reviewer.objects[o].share vs. COURT.grammars.ac.objects[o].share, for o in {stats_metrics, related_work, novelty}, across 952,856 official-reviewer units and 56,736 meta-reviewer units", "recompute": "scripts/build_court_data.py", "dom_ref": "data-court22a=\"rising\"", "verified": "claim recomputed from source island, 2026-08-25", "caveat_refs": [] }, { "id": "22a-falling", "statement": "Empirical scope, theory, and method design — the objects that demand re-reading the paper — fall to less than half their reviewer share.", "value": { "quantity": 0.5, "unit": "share_ratio_ceiling", "objects": ["empirical_scope", "theory", "method_design"] }, "source_island": "court-data.json", "source_path": "grammars[]", "derivation": "COURT.grammars.ac.objects[o].share ÷ COURT.grammars.reviewer.objects[o].share < 0.5 for o in {empirical_scope, theory, method_design}", "recompute": "scripts/build_court_data.py", "dom_ref": "data-court22a=\"falling\"", "verified": "claim recomputed from source island, 2026-08-25", "caveat_refs": [] } ] }, { "id": "fig-22b", "title": "What the bench forgives at the margin", "deck": "At the margin, only a clarity charge reliably blocks the pardon.", "dom_host": "#court-forgive", "claims": [ { "id": "22b-clarity-clear", "statement": "Clarity's interval clears zero — a paper the panel found hard to read is 4.4 points less likely to be lifted.", "value": { "quantity": -4.4, "unit": "pp_difference", "object": "clarity" }, "source_island": "court-data.json", "source_path": "forgive[]", "derivation": "COURT.forgive.clarity.adj_delta, matched inside 0.25-point mean-rating bins among the 6,547 borderline 2026 panels, 95% interval from resampling whole papers excludes zero", "recompute": "scripts/build_court_data.py", "dom_ref": "data-court22b=\"clarity-clear\"", "verified": "claim recomputed from source island, 2026-08-25", "caveat_refs": ["borderline-margin-association"] }, { "id": "22b-identity-lean", "statement": "Identity charges (framing, novelty) lean against the lift.", "value": { "quantity": null, "unit": "directional_sign_negative", "objects": ["problem_framing", "novelty"] }, "source_island": "court-data.json", "source_path": "forgive[]", "derivation": "sign of COURT.forgive[o].adj_delta is negative for o in {problem_framing, novelty}", "recompute": "scripts/build_court_data.py", "dom_ref": "data-court22b=\"identity-lean\"", "verified": "claim recomputed from source island, 2026-08-25", "caveat_refs": ["borderline-margin-association"] }, { "id": "22b-overlooked", "statement": "Robustness and related-work charges are the ones the bench most readily overlooks.", "value": { "quantity": null, "unit": "directional_sign_positive", "objects": ["robustness_sensitivity", "related_work"] }, "source_island": "court-data.json", "source_path": "forgive[]", "derivation": "COURT.forgive[o].adj_delta most positive (paper still lifted despite the charge) for o in {robustness_sensitivity, related_work}", "recompute": "scripts/build_court_data.py", "dom_ref": "data-court22b=\"overlooked\"", "verified": "claim recomputed from source island, 2026-08-25", "caveat_refs": ["borderline-margin-association"] } ] }, { "id": "fig-22c", "title": "The redemption: what the chair praises when it lifts a paper", "deck": "75% of a lifted paper's meta-review is praise — of its numbers and its positioning; on rejected borderline papers, praise falls to 8%.", "dom_host": "#court-redeem", "claims": [ { "id": "22c-praise-share", "statement": "Roughly three-quarters of a lifted paper's meta-review is praise, favoring its numbers and its positioning; on a rejected borderline paper, praise of any kind nearly vanishes.", "value": { "quantity": [0.75, 0.08], "unit": "share_of_meta_review_units_that_praise", "lifted": 0.75, "rejected_borderline": 0.08 }, "source_island": "court-data.json", "source_path": "redemption[]", "derivation": "COURT.redemption praise-share of meta-review units on lifted vs. rejected borderline papers", "recompute": "scripts/build_court_data.py", "dom_ref": null, "verified": "claim recomputed from source island, 2026-08-25", "caveat_refs": ["justification-not-evidence"] } ] } ], "caveats": [ { "id": "borderline-margin-association", "text": "These comparisons are association, not cause: the accept line and the borderline band are inferred from outcomes, and a paper's fate may hinge on information the panel mean does not carry — a chair's own reading, a scope policy, a discounted reviewer.", "scope": "Fig 22b forgiveness dot-and-interval comparisons", "verified_by": null }, { "id": "justification-not-evidence", "text": "The meta-review is written by the area chair after the decision is already made, so its praise reflects the language of justification, not independent evidence of paper quality.", "scope": "Fig 22c redemption praise shares", "verified_by": null } ], "links": { "recompute_scripts": [ "scripts/build_court_data.py", "scripts/build_panel_data.py" ], "data": [ "/api/v1/data/court-data.json", "/api/v1/data/panel-data.json" ], "related_plates": [ "plate-xiv" ] } }