Inferpedia - an encyclopedia of the missing

AI-generated conjecture · below the evidence/publication boundary

← All conjectures · Institutions, law & bureaucracy

The dealer's razor

Bronze Age Mesopotamia
Status: Prior

Status is derived only from the shepherd-authored triage/prediction data above -- community submissions and claims are a separate overlay and can never change it (see the participation panel below).

This is a proposed connection between two domains, generated by a language model. It is not an article and not evidence: it sits below the evidence/publication boundary. A quantitative prediction and a named kill-dataset are attached (when registered) so the claim stays falsifiable rather than merely evocative.

Claim (verbatim)

The dealer's razor. Selection bias from art-market economics meets Assyriology's sampling problem: the antiquities market is a high-pass filter for glamour, so unprovenienced (market-bought) tablets are systematically enriched in literary, lexical, and royal genres relative to excavated tablets of the same period, because dealers selected what would sell.

Prediction clause (verbatim)

Within each of the five largest CDLI periods, the administrative-genre share among tablets lacking a provenience is <= 0.7x the share among provenienced tablets, while the combined literary-plus-royal share is >= 1.8x, holding in at least four of the five periods; parity between market and excavated genre profiles kills it — and would certify the catalogue's sampling as unbiased.

Kill-dataset (verbatim)

Kill: CDLI genre labels crossed with provenience presence/absence and periods (in house); this turns the catalogue's own 'sampling regime unverified' caveat into the measurand.

Provenance

Run: Fresh agent generation · model: claude-fable-5

Generated by a fresh Fable-tier instance at maximum effort with generation-first blindness (no repo/web/DB access); titles-only knowledge of existing items, embedded in titles_supplied per the batch-2 lane rule; prompt pre-committed in docs/GOAL_CONJECTURES_BATCH3_20260705.md (b043140). Novelty unverified by construction. titles_supplied stripped to the committed sidecar conjecture_fresh_fablemax_batch3_titles_supplied_20260705.md at import (schema additionalProperties:false; relaxation queued).

Novelty / leakage triage

Adjacent (closely related prior work exists)

The mechanism is qualitatively established — post-1991 looting flooding the market, demand skewing to prestige/legible pieces, provenience-reconstruction bias recognized as a live methodological problem — but no quantitative genre-share comparison between provenienced and market tablets was located. In-house computable (CDLI provenience presence x genre).

Predictions

Killed registered 2026-07-05 calibration prediction (parent triage: leaked/adjacent)

Resolution: Killed

Caveats: A genuine kill - the registrant's priors said the direction was plausible and the data went the other way: within CDLI's catalogued corpus, market-flavored (unprovenienced) tablets skew MORE administrative than excavated ones in both test periods, the opposite of prestige-selection. A plausible historical reading, offered as interpretation not finding: cuneiform's unprovenienced stratum was substantially built by the great administrative-archive lootings (Drehem, Umma) that flooded the early-20th-century market with exactly the 'boring' genre - i.e. the market selected for AVAILABILITY, not legibility. Binding caveats: origin_place-empty is a proxy for market provenance (excavated pieces can lack findspots); the universe is CDLI-catalogued tablets only (prestige pieces routed to art-market channels may never enter CDLI, a selection this test cannot see); the harvest's sampling regime is unverified; the undated '?' stratum shows the OPPOSITE pattern (1.671) and deserves its own look. The conjecture as operationalized is dead; a version about NON-catalogued market material remains untestable here.

In-house test (triage: adjacent - the market-selection mechanism is qualitatively established, unquantified): among CDLI records of the same period, unprovenienced tablets (origin_place empty - the market-flavored stratum) are genre-enriched toward legible/prestige categories (non-administrative) relative to provenienced tablets.

Resolution criteria: Population: tradition=cuneiform, period_canonical_name in {Ur III, Old Babylonian}, genre = raw genres_for_ws_e (empty excluded, share reported). For each period compute the NON-administrative genre share among unprovenienced vs provenienced records. CLAUSE PRECEDENCE: (1) KILLED if the unprovenienced non-administrative share is <= 1.1x the provenienced share in BOTH periods, or lower in either. (2) SUPPORTED if the ratio is >= 1.5x in BOTH periods. (3) Otherwise INCONCLUSIVE. Narrative (non-binding): full genre breakdowns per stratum, the top enriched genres, and the '?'-period stratum reported for context. Caveat binding at registration: origin_place-empty is a PROXY for market provenance (some excavated pieces lack recorded findspots), and the local harvest's sampling regime is unverified.

Known priors disclosure: Seen at registration: stratum sizes only (Ur III 1,283 unprovenienced / ~58.4k provenienced; OB 3,826 / ~11.1k) and the period-level genre distributions from the entropy-floor resolution (Ur III ~97.9% administrative overall; OB ~43% administrative) - but NOT any genre-by-provenience cross-tab. The conjecture's direction is plausible from the qualitative literature (which is why triage said adjacent); the size of any enrichment is genuinely unknown.

As pre-registered: non-administrative genre share per provenience stratum per period, clause precedence explicit. Script preserved alongside the artifact.

Dataset: CDLI catalogue metadata (126,000 records): genre-labelled Ur III (provenienced 58,320 / unprovenienced 1,135) and Old Babylonian (10,381 / 3,327) strata by origin_place presence. Read-only.

computed 2026-07-05

Weigh in

No community feedback yet.

Add your take

Posted immediately (spam is removed). Community feedback is never an adjudicated verdict and never changes this conjecture's triage label or status above.