Generative AI Meets Participatory Budgeting: Will It Truly Put Citizens in Charge?
Why This Is Trending — Right Now
Cities worldwide are experimenting with generative AI to make participatory budgeting more accessible and engaging. These new systems allow residents to describe community problems in plain language and instantly generate polished, cost-estimated proposals for review.
Meanwhile, civic tech groups are publishing guidance for AI in public participation, emphasizing the importance of transparency, explainability, and inclusivity. Governments are also introducing policies that define how public servants should ethically use AI in decision-making.
It’s clear that AI is moving from city hall dashboards into the heart of democratic participation — and the implications are massive.
The Promise (and Why People Are Excited)
Lowering barriers to entry: Not everyone is a policy writer. AI can help citizens turn ideas into actionable proposals, empowering those who’ve never participated before.
Faster sense-making: When thousands of ideas pour in, AI can summarize, group, and highlight common themes — ensuring community voices aren’t lost in volume.
Better trade-off understanding: Citizens can instantly see how funding one idea might impact another — making deliberation more data-informed and productive.
Language inclusion: Generative AI tools can translate ideas and simplify technical jargon, allowing residents from multilingual or low-literacy backgrounds to contribute meaningfully.
The Peril (and Why Advocates Must Lean In)
Algorithmic bias: If models learn from data biased toward affluent or majority populations, they may unconsciously prioritize certain ideas or styles of expression.
Opaque scoring: When proposals are edited or filtered by AI without clear explanations, public trust erodes.
Token participation: Cities may adopt flashy AI platforms without giving citizens real influence in budget outcomes.
Privacy concerns: Participant data — including writing style, opinions, or location — could be misused without robust safeguards.
Digital divides: Communities without internet access or digital literacy risk further marginalization.
What “Good” Looks Like: A Civic Checklist UPA Can Champion
1️⃣ Open the black box Publicly release prompt templates, AI models used, and audit logs showing how proposals were clustered, edited, or scored.
2️⃣ Design for equity Test and monitor outcomes by neighborhood, language, and income group to ensure fairness in how ideas surface and advance.
3️⃣ Guarantee a human review path Every AI-edited or filtered proposal must have a simple appeal process for human review.
4️⃣ Ensure plain-language outputs All summaries, dashboards, and recommendations must have accessible versions written in everyday language.
5️⃣ Co-design with communities Bring residents — especially underrepresented groups — into workshops that shape prompts, filters, and interface design.
6️⃣ Mandate open procurement When cities buy or license these tools, contracts should require transparency, interoperability, and citizen oversight rights.
Global Lessons for Local Leaders
Cities such as São Paulo, Barcelona, and Seoul are already piloting AI-assisted civic budgeting. Early results show faster proposal processing and improved engagement — but also underscore the risks of overreliance on vendors.
The lesson: public institutions must build their own capacity to govern civic AI systems, not outsource democracy to private algorithms.
UPA’s 90-Day Action Plan
Phase 1: Map the Landscape (Weeks 1–3) Identify which cities in UPA’s network are experimenting with AI in citizen engagement. Publish a short report rating platforms on transparency and fairness.
Phase 2: Co-Design & Educate (Weeks 4–6) Host civic hackathons and workshops bringing together residents, coders, and policymakers to design fair AI participation workflows.
Phase 3: Monitor & Measure (Weeks 7–9) During live budgeting cycles, conduct “shadow audits” of AI edits and proposal filtering, then publish equity scorecards.
Phase 4: Policy Push (Weeks 10–12) Introduce a model “Civic AI Transparency Ordinance” requiring explainable algorithms, human appeals, and public access to audit data.
Conclusion
Generative AI could be the most powerful tool for citizen empowerment since the public hearing — or the most subtle form of digital disenfranchisement.
With the right advocacy, transparency standards, and community oversight, AI can help cities transform data into democracy — and ensure every voice truly counts.
Unified Public Advocacy stands ready to lead this movement — where technology serves people, not the other way around.