Integrating FOIA Data and Generative AI for Comparative Government Research
Dr. Bastián González-Bustamante, Principal Investigator
Diego Aguilar and Nando Verelst, Researchers
Research Background
The Executive Triangle
Policy capacity depends on relationships between cabinet members, civil servants, and political advisers. Most studies focus on the Global North and top positions, often overlooking deeper bureaucratic politicisation
2024 Baseline
In the framework of the COST project, our team filed more than 400 FOIA requests under Chile's Transparency Law 20.285, collecting data on ministers, undersecretaries, advisers and also senior civil servants
The Innovation: RAG Technology
Document Analysis Challenge
Manual FOIA document review proved extremely time-consuming, limiting research scope and depth
RAG Solution
Retrieval-Augmented Generation links Large Language Models (LLMs) to trusted document stores, enabling source-cited answers
Vector Database
State-of-the-art LLMs using Chilean public sector terminology, indexing 2024 and 2025 FOIA releases
Example query: "List all advisers in the Ministry of Finance who previously served in party leadership roles" returns exact FOIA paragraphs as evidence
Project Objectives
1
Dataset Expansion
Incorporate 2025 observations for cabinet members and systematically include top bureaucrats and advisers
2
Automated Analysis
Deploy the RAG system to extract biographical variables, including education, party links and prior public and private roles, with human-in-the-loop validation
3
Dissemination
Produce updated COST annual report, journal articles, and RAG open-source toolkit
International Research Network
The Principal Investigator is an active member of COST Action CA22150, serving on Working Group 2. This formal affiliation provides recognised European peer review and open access to scientific exchanges
This project strengthens Universidad Diego Portales' profile in a recognised European research network