Datasetpublic

chrishuberreitz/calibrated-authority-index

The Calibrated Authority Index 59 knowledge institutions, coded on how they construct trust in AI. Version 2026-07-31 · mean Calibrated Authority 9.7/12 · CC-BY-4.0 Nature, JAMA, the BBC, Oxford, UNESCO and dozens more wrote public rules for generative AI. Read together they reveal one pattern none of them named: they permit AI where its work can be cheaply checked, and reserve for a human the work that can't be. This dataset is that pattern, made measurable — each policy scored… See the full description on the dataset page: https://huggingface.co/datasets/chrishuberreitz/calibrated-authority-index.

sourceHugging Facecc-by-4.0updated 2mo agoView on Hugging Face

0likes26downloads

Dataset Card

The Calibrated Authority Index

59 knowledge institutions, coded on how they construct trust in AI. Version 2026-07-31 · mean Calibrated Authority 9.7/12 · CC-BY-4.0

Nature, JAMA, the BBC, Oxford, UNESCO and dozens more wrote public rules for generative AI. Read together they reveal one pattern none of them named: they permit AI where its work can be cheaply checked, and reserve for a human the work that can't be. This dataset is that pattern, made measurable — each policy scored on six dimensions into a single 0–12 number, anchored to the verbatim quote each score rests on. It is re-scored weekly as policies change.

—Live site: https://calibrated-authority.chrishuberreitz.com
—Methodology & how-to-cite: https://calibrated-authority.chrishuberreitz.com/methodology
—Source repo (PRs / corrections): https://github.com/chrishuberreitz/calibrated-authority-index

Columns

column	meaning
`id`, `name`, `segment`	institution identity + category
`D1`	Traceability & inspectability (0–2)
`D2`	Human authorship & accountability (0–2)
`D3`	Disclosure & labeling (0–2)
`D4`	Synthetic-identity / fabrication prohibition (0–2)
`D5`	Human validation in loop (0–2)
`D6`	Evidential-trust emphasis (0–2)
`ca`	composite Calibrated Authority score, sum of D1–D6 (0–12)
`posture`	Prohibitive / Balanced / Enabling
`c2_fit`	does the policy land where the thesis predicts (fits / deviates)
`c3`	trust-logic: Evidential / Relational / Both-split / Neither
`twilight`	uses precedent-collapse / "twilight" framing
`url`	the source policy document

Load it

python

from datasets import load_dataset
ds = load_dataset("chrishuberreitz/calibrated-authority-index")
print(ds["train"][0])

Cite it

Reitz, C.H. (2026). The Calibrated Authority Index (version 2026-07-31). https://calibrated-authority.chrishuberreitz.com

License

CC-BY-4.0 — use it, cite it.