Value-priority measurement of persons & organizations
We measure value priorities revealed in public records (decisions, actions, statements) as rankings on a domain × 45-cell grid, and report the distance between professed (words) and revealed (actions).
AIO 10003 — analyze an individual’s value priorities from public records
The standard number is the goal statement. The digits 1·00·0·3 say whose values, in which domain, and which action this document covers.
Individual — persons (and organizations with public records)
Common — domain-independent
Base document
Analyze — measure revealed priorities
What this standard defines — event collection (pre-registered sampling frame, source URLs, confidence grades), pair coding (genuine conflicts only, i≻j), ranking estimation (Bradley–Terry / Plackett–Luce with confidence intervals), professed–revealed gap reporting, and verified/unverified · observed/inherited/estimated labeling. What it does not define — whether a person, organization, or value is good or bad. These measurements describe priority structure only — they never judge (non-normative). All evidence comes from already-public records, and subjects may contest any finding; corrections are published with sources.
Step 3 of the integrity loop — Analyze (individual axis)
AIO 10003 is the first published standard on the individual (1) subject axis. The individual-axis setting (10001) and recording (10002) standards are planned stages; the same analysis cell on the organization (2) axis holds AIO 20003 · Benchmark.
How we measure
Public records (decisions, actions) are collected within a pre-registered sampling frame. Every event carries source URLs and a confidence grade.
We conservatively code only genuine conflicts — which value was sacrificed for which (i≻j). Mere mentions are not evidence.
Bradley–Terry / Plackett–Luce models estimate latent rankings per cell with confidence intervals, suited to sparse, unbalanced observations.
We report the distance between professed rankings (words) and revealed rankings (actions) — the core output, a measurement, not a verdict.
- Public records never cover all value pairs — covered pairs are "observed"; the rest are honestly labeled inherited, estimated, or no-data.
- Only cells meeting verification conditions (coder reliability, transitivity, minimum observations) are marked "verified" — we never call an entire subject "verified."
- A professed–revealed gap is not a disqualifier — it is a reported finding.
Measurement status — targets in progress
Contest and respond
Subjects and their organizations may contest any finding at info@aioq.org; corrections are published with sources. Suggestions for measurement targets are welcome at the same address.