teleo/teleo-codex
8
0
Fork 0
Code Issues 22 Pull requests 18 Projects Releases Packages Wiki Activity Actions 1

theseus: extract claims from 2023-10-00-anthropic-collective-constitutional-ai (#425)

Browse source 
Co-authored-by: Theseus <theseus@agents.livingip.xyz>
Co-committed-by: Theseus <theseus@agents.livingip.xyz>
This commit is contained in:
Theseus 2026-03-11 07:12:05 +00:00 committed by Leo
parent ea754c52b1
commit a5bac52470
1 changed files with 14 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

15
inbox/archive/2023-10-00-anthropic-collective-constitutional-ai.md
View file

@ -7,9 +7,14 @@ date: 2023-10-01
domain: ai-alignment domain: ai-alignment
secondary_domains: [collective-intelligence] secondary_domains: [collective-intelligence]
format: paper format: paper
status: unprocessed status: null-result
priority: medium priority: medium
tags: [collective-constitutional-ai, polis, democratic-alignment, public-input, constitution-design] tags: [collective-constitutional-ai, polis, democratic-alignment, public-input, constitution-design]
processed_by: theseus
processed_date: 2026-03-11
enrichments_applied: ["democratic alignment assemblies produce constitutions as effective as expert-designed ones while better representing diverse populations.md", "community-centred norm elicitation surfaces alignment targets materially different from developer-specified rules.md"]
extraction_model: "anthropic/claude-sonnet-4.5"
extraction_notes: "Curator correctly identified the 'desired behavior vs harm avoidance' asymmetry as novel claim material. The experiment provides strong empirical evidence for existing democratic alignment claims. No follow-up performance data available—Anthropic ran the experiment but did not publish outcome evaluation comparing publicly-constituted vs expert-constituted model behavior. This is the first frontier lab deployment of democratic alignment (2023), setting precedent for CIP's subsequent work."
--- ---
## Content ## Content
@ -50,3 +55,11 @@ Anthropic and CIP collaborated on one of the first instances where members of th
PRIMARY CONNECTION: [[democratic alignment assemblies produce constitutions as effective as expert-designed ones while better representing diverse populations]] PRIMARY CONNECTION: [[democratic alignment assemblies produce constitutions as effective as expert-designed ones while better representing diverse populations]]
WHY ARCHIVED: Foundational empirical evidence for democratic alignment — supports existing claims with Anthropic deployment data WHY ARCHIVED: Foundational empirical evidence for democratic alignment — supports existing claims with Anthropic deployment data
EXTRACTION HINT: The "desired behavior vs harm avoidance" asymmetry between public and expert constitutions could be a novel claim EXTRACTION HINT: The "desired behavior vs harm avoidance" asymmetry between public and expert constitutions could be a novel claim
## Key Facts
- ~1,000 U.S. adults participated (representative sample across age, gender, income, geography)
- 1,127 statements contributed to Polis platform
- 38,252 votes cast (average 34 votes/person)
- ~50% overlap between expert and public constitutions in concepts/values
- Polis identified two separate opinion groups despite high consensus on most statements