Civic Blueprint logoCIVIC BLUEPRINT

View this file on GitHub

docs/CONTENT_PROVENANCE.md

Provenance: collaborative. How Civic Blueprint labels human and AI collaboration.

On this page
  1. Content Provenance Standard
  2. Purpose
  3. Provenance labels
  4. Application rules
  5. Minimum disclosure template
  6. Scope
  7. Versioning

Content Provenance Standard

Purpose

This document defines how Project 2028 labels content provenance across all public and internal artifacts. The goal is straightforward: make authorship and generation pathways legible so readers can evaluate claims with appropriate context.

This standard is part of the project's trust posture and directly relates to:

The project is intentionally dog-fooding its own provenance policy position.

Provenance labels

Use exactly one primary label per artifact.

human
Definition
Written directly by the steward without AI drafting.
Typical examples
Outreach messages, direct steward reflections, personal notes
collaborative
Definition
Human-directed, AI-drafted, steward-edited. Steward sets intent/constraints, AI helps draft, steward revises and approves final form.
Typical examples
Core synthesis docs and public memos with steward sign-off
ai-generated, steward-curated
Definition
AI generates candidate content under defined protocols; steward curates, selects, and contextualizes outputs.
Typical examples
Proposal catalog entries, exchange outputs synthesized from multi-round generation
ai-generated
Definition
AI-generated content not yet reviewed or adopted by the steward.
Typical examples
Raw drafts, intermediate experimental output

Application rules

  1. Public-facing pages and documents should show provenance clearly and early (header, frontmatter-derived metadata, or an explicit callout near the top).
  2. Collaborative and curated artifacts should briefly describe process (for example: "drafted with AI and revised/approved by steward").
  3. Raw AI output should not be presented as endorsed project position.
  4. When provenance changes (for example, a raw AI draft becomes steward-curated), update the label.
  5. Do not over-claim human authorship for AI-assisted synthesis.
  6. Do not over-claim AI autonomy where substantial human direction and editorial judgment were applied.

Minimum disclosure template

Use this short form where space is limited:

Provenance: label — one-sentence process note.

Examples:

  • Provenance: collaborative — Drafted through human-directed AI iteration, then revised and approved by the steward.
  • Provenance: ai-generated, steward-curated — Generated through structured agent rounds; entries were selected and contextualized by the steward.

Scope

This standard applies to:

  • Core documents in project-2028
  • Memos and proposal catalog artifacts
  • Exchange documents where provenance is not already obvious from process descriptions
  • Public presentation layers in civicblueprint.org that render or summarize project content

Versioning

This is Version 1 of the provenance standard. It is expected to evolve as practitioner feedback arrives and as the project learns which labels are both legible and accurate in practice.