Purpose

The Algorism Index is a public behavioural accountability framework that evaluates the documented, verifiable actions of institutions and public figures against the Six Principles of Algorism.

It exists because the powerful are rarely held to the same behavioural standard they impose on others. Credit scoring evaluates ordinary people. Hiring algorithms filter applicants. Content moderation polices speech. But the institutions and individuals who design, deploy, and profit from these systems face no equivalent evaluation. The Index applies the same framework upward that is already being applied downward.

The Six Principles

Every evaluation scores the subject against these six principles, applied consistently regardless of political orientation, industry, or geography.

#PrincipleDefinitionInstitutional Application
1TruthfulnessTell the truth, even when it costs you.Are public statements consistent with internal documents and observed actions?
2ResponsibilityOwn your actions and the results.Does the institution acknowledge harm, or deflect to abstract forces?
3RepairFix the harm you cause.When harm is documented, does the institution take corrective action or issue statements without structural change?
4ContributionCreate value for others.Does the institution create net value for stakeholders, or extract value while externalising costs?
5DisciplineKeep your standards when tested.Does the institution maintain standards under competitive pressure or public criticism?
6IntegrityThink for yourself. Act coherently.Does the institution act consistently with its stated values, or abandon them when inconvenient?

Each principle is scored 0–5 for the specific action under evaluation, with documented reasoning. There is no single composite score. Multi-axis evaluation prevents oversimplification and resists gaming.

Evaluation Process

1. Subject Selection

Subjects are selected based on public salience, documentation quality, and relevance to AI governance, institutional power, or systemic behavioural patterns. Initial evaluations focus on institutions. Public figure evaluations begin only after the methodology has been tested through institutional evaluations.

2. Evidence Collection

All evidence must be publicly verifiable: official statements, regulatory filings, court documents, investigative journalism, public financial records, published research. Each claim cites at least one primary source. Where sources conflict, conflicting accounts are noted.

3. Scoring

ScoreMeaning
0Active violation. Action directly contradicts the principle.
1Significant failure. Pattern of neglect or disregard.
2Below standard. Some effort but significant gaps remain.
3Adequate. Meets baseline expectations.
4Strong. Consistent alignment. Proactive effort.
5Exemplary. Demonstrates the principle under pressure.

4. Repair Path

Every evaluation includes specific, concrete actions the subject could take to improve. The Index is not just “here is what you did wrong” but “here is what right looks like.”

5. Right of Reply

Before publication, the subject is notified and given the opportunity to respond. Responses are published alongside the evaluation, unedited.

Output: The Pattern Brief

Each evaluation is published as a structured Pattern Brief designed for journalists, investors, employees, procurement officers, and the public.

Headline: One sentence. The core finding.
Subject: Institution or public figure
Action Evaluated: Specific, documented action
Six Principles Scorecard: Six scores (0–5) with reasoning
Evidence Dossier: Facts / Inferences / Normative Judgments (clearly separated)
Steelman Defence: Strongest plausible defence of the action
Repair Path: Concrete improvement actions
Right of Reply: Subject’s response, unedited
Pattern Risk Flags: Matching coercion patterns from the library

Pattern Risk Categories

For quick reference, evaluations include simplified A/B/C grades for relevant pattern risks.

Pattern RiskDefinitionRelated Principles
Truth DecaySystematically misrepresents reality through false statements, manufactured consensus, or censorship.Truthfulness, Integrity
Accountability VoidDeflects responsibility to abstract forces, market conditions, or subordinates.Responsibility, Repair
Extraction PatternExtracts value while externalising costs to workers, communities, or ecosystems.Contribution, Repair
Standards ErosionAbandons stated standards under competitive pressure or political convenience.Discipline, Integrity
Capture RiskGovernance captured by narrow interest group overriding broader stakeholder interests.Integrity, Contribution

Credibility Safeguards

Against Partisan Capture

The Six Principles are politically neutral. Every evaluation praises where warranted and criticises where warranted, regardless of political alignment.

Against Gaming

No single composite score. Multi-axis evaluation prevents optimising one number while neglecting others. No paid ratings, no sponsored evaluations.

Against Authoritarian Misuse

The Index evaluates actions, not beliefs. The Repair Path makes it constructive. All assessments are inspectable, contestable, and subject to abstention under uncertainty.

Against Scope Creep

Only publicly documented and verifiable actions. If evidence is insufficient, the evaluation states “insufficient evidence” rather than forcing a verdict.

What the Index is not: Not a social credit system (applies upward, not downward). Not a court (no penalties or legal process). Not journalism (adds structured evaluation to documented facts). Not surveillance (uses only public information). Not a moral authority (publishes reasoning; readers decide).

Publication Protocol

Source verification: Every claim cites at least one primary source.

Steelman clause: Every evaluation includes the strongest plausible defence.

Fact/inference/judgment separation: The evidence dossier clearly labels each.

Right of reply: Subjects are contacted before publication where feasible.

Corrections SLA: Factual errors corrected within 72 hours, visibly timestamped.

Open methodology: This document is published in full. Anyone can apply the framework independently. The Index is reproducible by design.

Version 1.0 • March 2026 • This methodology is free to reproduce in its entirety with credit to Algorism.org.