## Fostering More Inclusive Democracy with AI

## Source details

**Canonical URL:** [Fostering More Inclusive Democracy with AI](https://www.imf.org/en/publications/fandd/issues/2023/12/pov-fostering-more-inclusive-democracy-with-ai-landemore)

## Other formats

- [Markdown version](/en/publications/fandd/issues/2023/12/pov-fostering-more-inclusive-democracy-with-ai-landemore/index.md)
- [Structured JSON version](/en/publications/fandd/issues/2023/12/pov-fostering-more-inclusive-democracy-with-ai-landemore/index.json)
- [Bundle manifest](/en/publications/fandd/issues/2023/12/pov-fostering-more-inclusive-democracy-with-ai-landemore/bundle-manifest.json)

## Bibliographic details
- Authors: HELENE LANDEMORE
- Published: December 1, 2023

---

### Thesis and framing
- Core argument: AI can enhance democratic institutions by ensuring citizens’ voices are truly heard, strengthening collective governance rather than simply enabling regulation.
- Key concerns addressed:
  - Fear that AI could undermine democracy, concentrate power in tech companies, and replace human decision-making with “algocracy.”
  - Counterpoint: current human societies retain political capacity and responsibility to shape AI development and harness AI to strengthen democracy.

### Participatory experiments: history and scale
- Longstanding efforts:
  - For 40 years many governments have run experiments to include ordinary citizens in policymaking beyond voting.
  - A 2020 Organisation for Economic Co-operation and Development report found close to 600 cases where a random sample of citizens engages deeply and formulates informed policy recommendations.
- Mass participation examples:
  - Participatory constitutional processes in Brazil, Kenya, Nicaragua, South Africa, Uganda (1980s–1990s); more recently Chile, Egypt, Iceland.
  - 2019 Great National Debate (France): 1.5 million participants.
  - EU Conference on the Future of Europe: 5 million website visitors and 700,000 people engaging in debate.
- Limitations of prior efforts:
  - Mostly low-tech, analog; often no AI.
  - Politicians overwhelmed by raw data, leading to citizen input being ignored.
  - Deliberation quality often superficial.

### Enhanced deliberation through AI: examples and mechanisms
- Taiwan Province of China / pol.is:
  - pol.is allows elaborated opinions, voting on others’ opinions, and maps opinion landscapes to identify consensus and minority views.
  - Recent integration of machine learning: contributors can engage with an LLM that speaks for different opinion clusters, helping users locate allies and opponents and reducing polarization.
  - Reach: platform used to consult 12 million people, or nearly half the population.
- Corporate experiments:
  - Meta Community Forums (2022): randomly selected users deliberated on climate content regulation.
  - December 2022 Meta experiment: 6,000 users from 32 countries in 19 languages discussed cyberbullying in the metaverse over several days; AI facilitated speaking times, topic decisions, and pacing.
  - Current evidence: no proof AI facilitators outperform humans yet, but AI facilitators will be considerably cheaper and could enable scaling deliberative processes from thousands to millions.

### Translation, summarization, and analysis capabilities
- Functional advances:
  - Instantaneous translation in multilinguistic groups as a near-term frontier.
  - Summarization of collective deliberations to distill and surface citizen input.
- Comparative performance:
  - Recent research: AI is 50 percent more accurate than human beings for summarization (as evaluated by trained undergraduates comparing AI summaries and human coders’ summaries of deliberation transcripts).
- Role of human judgment:
  - Some human judgment likely remains necessary; AI can aid human analysts, facilitators, and translators.
- Funding and development:
  - OpenAI launched a grant program, Democratic inputs to AI, subsidizing the 10 most promising teams working on algorithms serving human deliberation.

### Risks, limitations, and mitigation approaches
- Enumerated risks:
  - Data bias, privacy concerns, potential for surveillance, legal challenges.
  - Digital divide and potential exclusion of illiterate and techno‑skeptical groups.
- Non-technological remedies:
  - Many issues require political, economic, legal, and social solutions first, not technology alone.
- Technological mitigations:
  - Zero-knowledge protocols (zero-knowledge proofs, or ZKP) suggested to verify identity without collecting participant data (examples: text messaging authentication, blockchain); applicable for online voting, deliberative contexts, whistleblowing.
  - Generative AI can expand access to scarce knowledge and tutoring, adapt explanations to individuals’ cognitive styles, produce images, and convert oral input to written input.

### Policy implications and recommendations
- Democratic engagement in AI governance:
  - Greater citizen voice and input are needed in AI governance domestically and at the international institutional level.
  - AI-empowered democracies could enforce AI regulation more effectively and legitimize oversight capacities.
- Investment priorities:
  - Urgent investment in AI tools that safely augment participatory and deliberative potential of governments.
  - Support for algorithmic tools and platforms that scale deliberation and ensure inclusivity (including translation, summarization, and privacy-preserving identity verification).
- Caution:
  - Balance enthusiasm with safeguards to prevent democracy itself from becoming a casualty of the AI revolution.

### Key statistics and figures (verbatim)
- 40 years
- close to 600 cases
- 1.5 million participants
- 5 million people to visit the website and 700,000 to engage in debate
- pol.is engaged 12 million people, or nearly half the population
- 2022 (Meta experiments)
- December 2022
- 6,000 users from 32 countries in 19 languages
- AI is 50 percent more accurate than human beings for summarization (per cited research)
- Democratic inputs to AI grants subsidized the 10 most promising teams

*Fostering More Inclusive Democracy with AI by HÉLÈNE LANDEMORE, F&D Magazine, December 2023*

---

## Content in this bundle

- **Point of View: Fostering More Inclusive Democracy with AI**
  - [Point of View: Fostering More Inclusive Democracy with AI (Markdown version)](/-/media/files/publications/fandd/article/2023/december/pov-landemore.pdf.md){rel="alternate" type="text/markdown"}
  - [Point of View: Fostering More Inclusive Democracy with AI (PDF)](/-/media/files/publications/fandd/article/2023/december/pov-landemore.pdf){rel="external" type="application/pdf"}

---

_Source: https://www.imf.org/en/publications/fandd/issues/2023/12/pov-fostering-more-inclusive-democracy-with-ai-landemore_