Docs
This is the reference surface for AIM (Adaptive Intelligent Measurement). It is written for the engineer who needs to verify. Every page answers in this order: what it does, what you send, what you get back, what can go wrong.
What AIM measures#
AIM measures performers, human or AI, on calibrated interval scales. The flagship ruler is the Model of Hierarchical Complexity (MHC) reasoning-complexity order: a Rasch-family conjoint measurement scale on which humans, AI agents, and robots take locations with stated standard errors. Additional scales (hallucination propensity, persuasion severity and effectiveness, and others) are in build or in calibration on the same measurement discipline.
A measurement is a location on a construct, plus the standard error at that location, plus the calibration vintage of the instrument that produced it. Nothing ships without the uncertainty stated.
The measurement-and-evaluation discipline#
The discipline behind every page on this surface:
- Every published value carries a stated standard error. No bare scalars.
- Rows that fail the floor or lack evidence are shown and flagged, never removed.
- Certification tiers (Canonical, In calibration, Watch) are stated on every scale and every claim. Tier changes are events, dated and recorded.
- The API is planned, not live. Where an endpoint is shown, it is a contract under design, labeled PLANNED, consistent with aim.json (programmatic_access = planned, not yet available, see F-API-MCP-1).
The four doors#
Concepts. The measurement model, the stated-SE-per-band discipline, and the certification tiers. Start here if you need to know what a number from AIM means.
API (planned). The programmatic surface under design: endpoints, authentication model, request and response shapes, error taxonomy. Nothing on that page is callable today.
Scientific references. The published method and the registrations behind every number: NIST AI 800-3 alignment, the De Gruyter metrology chapter (open access) and the Cambridge Handbook chapter on unobtrusive measurement, OSF registrations, and the 16-year research program.
Glossary. Plain definitions of the terms used across the site: Rasch measurement, logit, standard error, many-facet model, the Model of Hierarchical Complexity, Kaleidoscope, Hydra, and Echo Wall.
Status of this surface#
These pages are the engineering and scientific reference for AIM. The API is planned, not live. Where a number appears it is the verified figure, and where a scale is not yet certified it is labeled with its tier. Links on this surface are stable.