Methodology · v0.1

How this site decides what AI caused.

The core discipline is to keep three separate things separate: a price moved, AI demand exists, and AI caused the move. Only the third requires an attribution grade.

The rubric

Four grades, applied to a specific claim

AI-driven

AI demand is the dominant, documented cause in the defined market and period.

AI-contributing

AI demand is a measurable factor among other causes that remain material.

Contested

Credible evidence points both ways, or the measurement cannot isolate the effect.

Scapegoated

Cited evidence contradicts the attribution or does not establish its causal mechanism.

Unit of analysis

Grade the sentence, not the topic

“Data centers increase electricity demand” and “AI raised my bill” are different claims. The first can be supported nationally while the second remains a utility-level question. Pages preserve the common wording so a narrow finding cannot be inflated in retelling.

Evidence hierarchy

Primary when possible, plural when necessary

  1. Measured public seriesBLS, EIA, FRED, queue and tariff data.
  2. Agency analysisMethods and models with definitions, ranges and dates.
  3. Filings and producer statementsUseful for capacity and order books; treated as interested evidence.
  4. Peer-reviewed researchBest for mechanisms and uncertainty, not always current market state.
  5. Trade press and anecdotesLead generation only until corroborated.

Causal checklist

Five questions on every report card

  1. What exactly moved: price, wage, allocation, lead time or capacity?
  2. What is the proposed AI-demand pathway?
  3. Which non-AI buyers and supply shocks share the market?
  4. What counterfactual or control would isolate the effect?
  5. What evidence would change the grade?

Index policy

Ranges over false precision

The Crowding Index will show basket coverage, component contributions and sensitivity to alternative weights. Missing series are not silently carried forward. Revisions retain the previous vintage and appear in the changelog. The household AI Tab stays unpublished until Tier-1 effects have a defensible pass-through model.

Governed updates

Automation drafts. A person publishes.

Watchers may fetch releases, identify changed values and draft summaries. They cannot silently flip a grade. Every public change records the old state, new state, reason, source, check date and editor approval. A failed or uncertain update remains queued.

Conflicts and corrections

A disclosed interest raises the evidence threshold.

The capital-markets ledger receives the most conservative threshold because the editor is a partner in a crypto fund that could benefit from an AI-capital-crowding narrative. Corrections are evaluated on evidence and logged whether they strengthen or weaken an AI attribution.