Initiative file / Judicial
Filing Classifier
Evidence record v2
01 / 39
- Institution
- Judicial Branch
- Editorial classification
- Verified adoption
- Initiative type
- AI system
- Documented phase
- Operational
- Relationship to AI in the source
- Explicit relationship to AI
- Execution evidence
- Confirmed
- First evidence
- Last verified
- Next review: Dec 8, 2026
Documented scope
What it is
Judicial Branch AI system that uses machine learning and optical character recognition to read and automatically classify filings submitted to specialized debt collection courts. The institution documents its deployment across the entire debt collection jurisdiction.
Objective stated by the institution
Automatically assign a type to filings submitted to specialized debt collection courts in order to organize work and speed up case processing.
Known timeline
Only milestones with an explicit date in the sources or editorial review are shown.
- First evidence
- Pilot start
- Latest editorial verification
- Next scheduled review
Evidence reading
Each dimension is assessed separately. Undetermined marks an evidence gap; not applicable means the dimension does not correspond to the documented object.
What is confirmed
- The pilot began in Pérez Zeledón in March 2020, and the Superior Council approved its nationwide expansion in December 2022.
- The system uses machine learning and optical character recognition to extract text and identify the type of filing.
- Circular 241-2025 requires feedback to the Classifier and manual classification when the system does not resolve a filing.
- Judicial Bulletin 165-2026 places Depurador de Escritos, an enhancement associated with the Classifier, within SIGMA; it does not place the entire system there.
What could not be determined
- No public independent measurement of current accuracy after nationwide deployment was found.
- The sources do not isolate the Classifier's effect on response times or caseload from other tools and management measures.
Documented results
Results are published only when a source traceably attributes them to this initiative.
The Judicial Branch reported 223,689 filings processed and 89% effectiveness from March 2021 through August 2023, during the pilot, training and refinement stages.
According to the Judicial Branch, the Classifier processed 1,302,899 filings across the debt collection jurisdiction from January 1 through November 29, 2024.
Evidence matrix
Sources and traceability
Sources state which claims they support, their origin and the date on which they were consulted.
- [1]Judicial Branch implements artificial intelligence to reduce the debt-collection caseload
- Publisher
- Poder Judicial de Costa Rica
- Source type
- Official primary source
- Consulted
Supports: existence · stated objective · execution · AI technique · operational use · reported result
- [2]Judicial Branch advances artificial intelligence in its debt collection courts
- Publisher
- Poder Judicial de Costa Rica
- Source type
- Official primary source
- Published
- Consulted
Supports: existence · stated objective · execution · AI technique · operational use · reported result · governance
- [3]Judicial Branch receives international recognition for technological progress and gender commitment
- Publisher
- Poder Judicial de Costa Rica
- Source type
- Official primary source
- Published
- Consulted
Supports: existence · AI technique · reported result
- [4]Debt collection courts reach their lowest caseload in four years in 2024
- Publisher
- Poder Judicial de Costa Rica
- Source type
- Official primary source
- Published
- Consulted
Supports: existence · execution · operational use · reported result
- [5]Circular 241-2025: mandatory feedback for the Filing Classifier artificial intelligence system
- Publisher
- Poder Judicial de Costa Rica
- Source type
- Official primary source
- Published
- Consulted
Supports: execution · operational use · governance
- [6]Judicial Bulletin 165-2026: use of Depurador de Escritos associated with the Classifier
- Publisher
- Poder Judicial de Costa Rica
- Source type
- Official primary source
- Published
- Consulted
Supports: execution · operational use · governance
Open questions
- What are the accuracy, error rate and volume of manual corrections after nationwide deployment?
- What documented audit, traceability, notice and appeal mechanisms complement feedback and manual fallback?
Context
The pilot began in Pérez Zeledón in March 2020; the Superior Council approved expansion in December 2022, and an institutional source confirmed nationwide operation in 2024. Circular 241-2025 requires feedback to the system and manual classification when needed. In 2026, the Judicial Branch documented an enhancement called Depurador de Escritos within SIGMA; this does not imply that the entire Classifier is hosted in SIGMA.
Related projects
Verified adoptionOperational
Verified:
ML model for budget execution forecasting
Machine-learning model used by the Judicial Branch to predict budget execution from its historical behavior. The institution’s 2019 accountability report documents its use and associated savings.
- Initiative type
- AI system
- Operational start
Verified adoptionOperational
Verified:
Anonymization with Nymiz
Artificial-intelligence tool implemented by the Judicial Branch to depersonalize rulings and protect sensitive data. The institution identifies Nymiz as the platform used and documents its application in the process.
- Initiative type
- AI component
- Operational start
Ecosystem and capabilitiesOperational
Verified:
AI analysis of Constitutional Chamber rulings
Research project by UCR’s Legal Research Institute and the State of the Nation Program. Since 2020 it has developed a standardized database, thematic classification and automated analysis of Constitutional Chamber rulings for academic research.
- Initiative type
- Research
- First evidence