Evidence First
Published scores require documented sources, evidence levels, confidence ratings, and limitations.
Assessment Program v1.0
Assessments begin with evidence, not scores.
A reusable program for evidence-based, reviewable assessments of AI models, agents, organizations, schools, cities, government systems, and products.
Purpose
Make ecosystem impacts visible and reviewable.
Foundation assessments translate the ten Framework dimensions into a structured public-interest review. They are not endorsements, rankings, or marketing claims.
Governance
Assessors must disclose conflicts, separate evidence from interpretation, and complete review before publication.
Equal Weighting
Each Framework dimension is scored from 1 to 5 and contributes 10% of the assessment.
Reassessment
Assessments are revisited after material changes, incidents, new evidence, or scheduled review dates.
Workflow
From preparation to reassessment.
The program allows preparation materials to be public before scores are ready, so scope and evidence requirements can be inspected early.
01
Preparation Phase
02
Scope Approval
03
Evidence Collection
04
Evidence Review
05
Dimension Scoring
06
Passport Analysis
07
Review
08
Publication
09
Reassessment
Evidence
Four evidence levels.
Level A and Level B evidence can support scoring. Level C informs questions and risk flags. Level D cannot be a primary scoring basis.
Level A
Highest Confidence
Official documentation, technical reports, regulatory filings, audit reports, and published standards.
Level B
Medium Confidence
Academic papers, independent research, and reputable third-party assessments.
Level C
Lower Confidence
News articles, public reporting, user reports, and incident context.
Level D
Speculation
Cannot be a primary basis for scoring; useful only as a research gap or monitoring note.
Scoring
Ten dimensions, equal weight.
Each dimension receives a 1-5 evidence-backed score. The maximum raw assessment score is 50.
Purpose
Human Agency
Transparency
Accountability
Trustworthiness
Human Rights
Ecological Impact
Intelligence
Resilience
Ecosystem Contribution
Assessment #001
ChatGPT Assessment
Status: Preparation Phase. Scope, evidence requirements, and assessment questions are being defined. No scores are published.
Program Documents
Reusable assessment rails are now in place.
The public website explains the program. The repository contains the full governance guide, evidence standards, scoring workbook, matrix template, publication template, and ChatGPT preparation record.