evidoria

← Back to browse

Good practice Imported

CitizenLab's AI Text-Clustering Turns 1,700 Youth Climate Ideas into 15 Policy Priorities in Belgium

Belgium · Brussels · See the Belgium profile · See the Brussels profile

Evidence: Descriptive / self-reported Top 32% 67/100 · Ask Evidence Copilot about this practice

Belgium's Youth for Climate movement used CitizenLab's NLP platform to collect 1,700 ideas and 32,000 votes, auto-clustering them into 15 priorities handed to policymakers after 1,400 untagged ideas were matched by similarity to manually tagged ones.

1,700
Ideas submitted (2019)
~32,000
Votes cast (2019)
15
Policy priority themes generated (2019)

Details

Promoter
CitizenLab (now Go Vocal) with Youth for Climate Belgium
Period
2019
Keywords
civic tech, citizen engagement, climate policy, NLP

Context

During the 2019 Belgian school climate strikes, the Youth for Climate movement partnered with Brussels civic-tech firm CitizenLab (now Go Vocal) to turn a flood of youth proposals into a report policymakers could use.

Activities

Over roughly three months, 1,700 ideas were submitted and drew about 32,000 votes. CitizenLab manually tagged a working subset and used natural-language similarity matching to auto-tag the remaining ~1,400 ideas.

Results

The clustering distilled submissions into 15 priority themes shared with the youth community and used to open dialogue with elected officials. CitizenLab's manual spot-checks found some mis-tags, though usually topically adjacent.

Conclusions

By 2024 the underlying platform was used by roughly 500 governments and organisations worldwide, though this specific youth campaign was a one-off, time-boxed exercise rather than a standing government service.

Implementation
Indicative cost
Low (< €50k) — Not disclosed; a time-boxed civic-tech platform engagement rather than a capital project.
Time to results
Short (< 1 year) — Ran over roughly three months in 2019.
Staffing & skills
CitizenLab/Go Vocal civic-tech team

Conditions for success

  • A pre-existing NLP clustering platform rather than bespoke development
  • Manual tagging of a working subset to seed the similarity-matching model

Common failure modes

  • Automated tagging produced some mis-tags, requiring manual spot-checking

Commonly funded by

National / regional programmes

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

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