AI-generated conjecture · below the evidence/publication boundary

The rounding gradient

Status: Anticipated · untested

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)

Every scientist knows significant figures: precision reflects the stage of measurement, and aggregated numbers get rounded. Ur III accounting was a pyramid — daily receipts feeding monthly ledgers feeding annual balanced accounts — and the conjecture is that the pyramid has a measurable precision gradient: terminal digits of quantities are nearly uniform at the receipt level and increasingly dominated by sexagesimal round numbers at each aggregation tier, steeply enough that a tablet's hierarchical level can be classified from its numbers alone. The mechanism is institutional tolerance: clerks were permitted rounding at each consolidation step because exactness was a property of documents, not of grain. If it holds, many famous discrepancies between receipts and summaries are neither fraud nor loss but sanctioned rounding, and the audit logic of the first bureaucracy becomes readable in its last digits.

Prediction clause (verbatim)

In Ur III texts, the proportion of commodity quantities ending in sexagesimally round values will rise monotonically across the three tiers: primary receipts, intermediate ledgers, annual balanced accounts, with a spread of at least 15 percentage points between bottom and top. Primary clause, which decides the verdict: a trend test (Cochran-Armitage) across tiers is significant at p<0.001 on samples of at least 2,000 quantities per tier. Secondary clause: a classifier using terminal-digit profiles alone assigns tier at better than 60 percent accuracy on held-out tablets.

Kill-dataset (verbatim)

BDTNS (quantities with text-type classifications across the Umma, Girsu, and Drehem archives).

Nobody has run this test. The kill-data is named above. If you can run it — or you know the paper that already settles it — claim the kill or submit the prior. Kills and priors are credited here, by name, as they come in.

In the atlas

This conjecture is bridged, as an L1 lead, onto these Inferpedia subject pages.

Provenance

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

Composed blind by claude-fable-5 from internal knowledge only, with zero tool calls, and emitted directly as a single JSON text message.

Novelty / leakage triage

anticipated in the literature — this exact test has never been run

Searched rounding and numerical practice in Ur III accounts. Ouyang & Proust document sexagesimal reconstruction/rounding in marginal numbers and Englund analyzes accounting precision and discrepancies, but no terminal-digit precision gradient across aggregation tiers or tier-classification claim was located.

Predictions

No prediction registered yet.

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.