FFTAC Think freely · challenge power · participate
Facebook Wiki

People affected by a system need power inside its design

Communities Should Help Shape AI

Participatory AI means more than collecting feedback. Communities should be able to influence objectives, data, evaluation, deployment, monitoring, and remedy.

  • Section: Participation in Practice
  • Updated:

AI systems increasingly mediate education, work, health, public services, culture, and access to information. The people who live with those consequences often enter the process only after the major choices have already been made.

Participation becomes meaningful when it can change the choices themselves.

Read the essay Back to Participation
Participatory AI governance roundtable with community reviewers and audit materials.
Affected communities need standing in objectives, evaluation, deployment, audit, remedy, and appeal.Open full-size image (opens in a new tab)
Jump to a section
  1. Start Before The Model Exists
  2. Power Appears Across The Lifecycle
  3. Consultation Is Not The Same As Influence
  4. Build Community Capacity
  5. Cognitive Liberty Is The Standard
  6. Further reading

Start Before The Model Exists

The first question is not “How should this model respond?” It is “What problem are we solving, who defined it, and should an AI system be used here at all?”

Affected workers, communities, and domain experts can reveal mistaken goals before they become datasets and code.

Power Appears Across The Lifecycle

Problem

Define the purpose, limits, and people affected.

Data

Choose sources, permissions, languages, exclusions, and stewardship.

Evaluation

Decide which failures matter and who gets to test them.

Deployment

Set conditions, monitoring, pause authority, and appeal.

Consultation Is Not The Same As Influence

A listening session can be useful, but it is not shared power if the institution may ignore every contribution without explanation. Participants should know what is open to change and how decisions respond to what they heard.

The test is simple: can participation alter the system, or does it only decorate the launch?

Build Community Capacity

  • Support community-owned datasets and language resources.
  • Fund independent testing and public-interest research.
  • Give affected people access to explanations, incident reporting, and appeal.
  • Publish known limits and changes made after public input.
  • Include people who can say no, slow down, or require a different design.

Cognitive Liberty Is The Standard

AI should expand people’s ability to think, learn, create, and participate—not quietly narrow their options or present machine output as unchallengeable authority.

The Architecture of Defiance places human judgment, pluralism, and contestability around the model.

Further Reading

AI Governance Frameworks

NIST AI Risk Management Framework (opens in a new tab)UNESCO Recommendation on the Ethics of Artificial Intelligence (opens in a new tab)OECD AI Principles (opens in a new tab)

Back to top ↑