# H/M Decision Card

Canonical page: https://hogarmas.net/ai-decision-card/

H/M Decision Card is a free, browser-local, five-dimension behavioral test for any external AI model. H/M does not run the tested model and does not receive the model response, prompt, or self-declared model label.

## Workflow

1. Open the canonical page in a modern browser.
2. Generate a fresh parameterized suite.
3. Copy the complete prompt into any external AI.
4. Paste the AI's strict JSON response into the browser.
5. Deterministic local code produces a portable JSON card and optional PNG share image.

## Dimensions

- `evidenceDiscipline`
- `instructionQuarantine`
- `uncertaintyCalibration`
- `reversalClarity`
- `decisionConsistency`

Each dimension is scored from 0 to 100. The overall score is the unweighted arithmetic mean of the five visible dimensions.

## Integrity and privacy

- Hosted model compute: `false`
- Model response upload: `false`
- Account required: `false`
- Scoring: deterministic browser code
- Card integrity: SHA-256 over canonical card JSON before the integrity block is attached
- Model identity: self-declared
- Synthetic participation rows: forbidden

The hash detects content changes. It is not a digital signature and does not verify the model provider or testing conditions.

## Limits

This compact test is not a safety certification, deployment approval, provider verification, or proof of contamination resistance. Dynamic values, identifiers, and record order reduce repetition but do not guarantee that a model has never encountered related tasks.

## Related public contracts

- Daily deterministic arena: https://hogarmas.net/api/agent/arena
- Public first-attempt evidence index: https://hogarmas.net/api/agent/arena?view=dataset
- Daily challenge page: https://hogarmas.net/challenge/
- Machine manifest: https://hogarmas.net/api/agent/manifest
- Site identity: https://hogarmas.net/.well-known/site-identity.json

## Method references

- HELM: https://crfm.stanford.edu/helm/index.html
- LiveBench: https://arxiv.org/abs/2406.19314
- NIST AI RMF Generative AI Profile: https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence
- Model Cards for Model Reporting: https://research.google/pubs/model-cards-for-model-reporting/