evidoria

← Back to browse

Good practice Imported

Brasil Participativo — AI-Assisted Semantic Clustering of Citizen Proposals for the Federal Multi-Year Plan

Brazil · Brasília · See the Brazil profile · See the Brasília profile

Evidence: Observational / pre–post Top 51% 60/100 · Ask Evidence Copilot about this practice

Brazil's federal platform, built on Decidim, drew 8,000+ citizen proposals and 1.5 million votes for the 2024–2027 Multi-Year Plan; researchers then applied BERTopic-plus-LLM clustering to 10,000+ of its proposals to make citizen input usable for policy work at scale.

8,000+
Formal proposals submitted for the 2024-2027 PPA
1.5 million
Votes cast during the PPA consultation
1.4 million
Platform accesses during the PPA consultation
10,000+
Proposals analysed via BERTopic + LLM clustering study (2025 study)

Details

Maturity
Scaling
Promoter
Secretaria-Geral da Presidência da República & Ministério do Planejamento e Orçamento
Period
2023–present
Keywords
citizen participation, federal budget planning, NLP, semantic clustering

Context

Brasil Participativo is the Brazilian federal government's official digital participation platform, launched in 2023 by the Secretaria-Geral da Presidência da República (SGPR) together with the Ministério do Planejamento e Orçamento (MPO), built on the open-source Decidim framework. Its first major use was collecting citizen contributions to the 2024-2027 Plano Plurianual (PPA), Brazil's federal multi-year budget plan.

Objectives

Because manual classification of proposals at this scale is infeasible, the government sought a way to turn large volumes of unstructured citizen input into data that public servants can act on, applying AI-based semantic clustering to make participation processes usable for policy work.

Activities

The PPA process registered roughly 1.4 million platform accesses, more than 8,000 formal proposals and about 1.5 million votes; the platform was subsequently used for Brazil's National Climate Plan consultation, which drew around 50,000 contributors, and for the National Youth Congress process. A 2025 study applied BERTopic combined with seed words and automated large-language-model validation to more than 10,000 proposals drawn from these three processes.

Results

The study reports that the resulting topic clusters were coherent and aligned with the government's official policy taxonomies while requiring minimal manual expert effort. The government has paired this technical use of AI with an explicit governance step, publishing a 'Guia de Uso Ético de Inteligência Artificial' (Ethical AI Use Guide) directly on the platform.

Implementation
Indicative cost
Medium (€50k–€500k)
Time to results
Medium (1–3 years)
Staffing & skills
Secretaria-Geral da Presidência da República (SGPR), Ministério do Planejamento e Orçamento (MPO)

Conditions for success

  • Open-source Decidim participation framework
  • Independent 2025 academic case study validating the clustering approach
  • Published governance guardrail (Ethical AI Use Guide) setting expectations for AI use on citizen contributions

Common failure modes

  • The clustering study reports coherence and taxonomy alignment but no independent accuracy benchmark or comparison group

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Replication kit

Reusable artefacts from this practice — as published by their sources.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Similar practices you may find useful