Skip to content
Polity Atlas
01
Knowledge Atlas
02
Living Atlas
03
Explorer
04
Polity Lab
05
Observatory
Search or ask
/
EN
Sign in
Menu
Search or ask
01
Knowledge Atlas
Concepts
Mechanisms
Thinkers
Claims & evidence
Knowledge graph
Glossary
02
Living Atlas
World map
Network view
Catalogue
Featured case
Submit an experiment
03
Explorer
Governance Explorer
Side by side
04
Polity Lab
Start a design
Designer
Stress test
Charter
Institution repository
05
Observatory
Signals & research
Briefings
Scenarios
Research themes
Digest
Language
Dark theme
Sign in
Join
Knowledge Atlas
/
Sources
Web page
■
Primary source
Collective Constitutional AI: Aligning a Language Model with Public Input
Anthropic and the Collective Intelligence Project · 2023
Open publisher page ↗
Copy citation
BibTeX
Type
Web page
Accessed
2026-09-01
Claims citing this source
3 claims
■
Established
About 1,000 US adults recruited by PureSpectrum submitted 1,127 statements and cast 38,252 votes on Polis; after moderation 275 statements remained, overlapping about 50% in concepts with Anthropic's 58-principle constitution.
■
Established
The model trained on the public constitution performed equivalently to the standard model on MMLU, GSM8K and helpfulness and harmlessness ratings, and showed lower bias across nine social dimensions of the BBQ benchmark.
■
Established
The organisers acknowledge that many subjective judgement calls in moderating and translating public statements into training principles shaped the final public constitution.
EN
English
Русский
Español
Français
Deutsch
Português
中文
日本語
العربية
हिन्दी