FFTAC Cognitive liberty · defiance · participation
Wiki

Participation must change a consequential lever

Participatory AI Governance

A lifecycle framework for public standing in AI problem framing, data stewardship, annotation, post-training, evaluation, deployment, monitoring, and appeal.

AI governance framework Reviewed 2 min read

Participation in AI allocates power over which problems are defined, whose language becomes data, which harms are measured, whose preferences shape post-training, who tests systems, and how affected people obtain correction or remedy.

The decisive question is not whether stakeholders were consulted. It is whether they could change objectives, datasets, evaluation criteria, release conditions, procurement rules, or remedies.

Participation Start reading
On this page
  1. Participation Across The AI Lifecycle
  2. Participation-Washing
  3. Governance Mechanisms
  4. Conscious Contribution, Not Disappearance
  5. Cognitive Liberty Requirements
  6. Related Pages

Participation Across The AI Lifecycle

Problem framing

Challenge the objective, target, success definition, and assumption that automation is necessary.

Data stewardship

Govern provenance, representation, licensing, consent or lawful basis, retention, correction, and benefit.

Annotation and curation

Define labels, edge cases, cultural context, disagreement, and documentation with affected communities.

Post-training

Make preference sampling, heterogeneity, aggregation, and default-value choices visible and contestable.

Red teaming and evaluation

Convert external findings into tests, mitigations, release criteria, and public closure records.

Deployment and remedy

Provide monitoring, incident reporting, explanations, human review, appeal, and revision after release.

Participation-Washing

A process is participation-washing when the organization predetermines the questions, extracts unpaid contribution, compresses dissent into artificial consensus, publishes attendance instead of responses, and gives participants no continuing authority.

Governance Mechanisms

Participation mechanisms and evidence of effect
MechanismDecision leverageEvidence of effect
Problem-framing workshopObjective and system necessityRevised problem statement or decision not to automate
Community-curated datasetCorpus composition and documentationDataset card, provenance, governance, version history
External red teamSafety tests and mitigationsNew test cases, closure status, release decision
Public standards or rulemakingDeployment obligationsDocket, response to comments, enforceable text
Community oversight boardMonitoring and remedyDecision rights, incident log, appeals, public reports

Conscious Contribution, Not Disappearance

FFTAC favors public-interest archives, community-curated corpora, open knowledge, multilingual documentation, participatory audits, standards, and appeals. It does not present data strikes, deletion campaigns, or platform abandonment as the solution to representation or governance.

Cognitive Liberty Requirements

  • Mental privacy and limits on behavioral or biometric inference.
  • Provenance, limitations, and meaningful explanations.
  • Plural viewpoints without forced ideological conformity.
  • Contestability, correction, human review, and appeal.
  • User-controlled export, portability, and interoperable tools.