Meta-Scientific Study of BRACIS Reveals AI Research Trends in Brazil
A recent meta-scientific investigation reviews eleven years of the Brazilian Conference on Intelligent Systems (BRACIS), the premier platform for Artificial Intelligence research in Brazil, organized by the Brazilian Computer Society since 2012. This study, accessible on arXiv (2608.09964), compiles a detailed record of all 1,066 accepted papers from 2015 to 2025, utilizing DBLP metadata, 6,765 citations from Google Scholar, and full texts. It explores publication trends, author participation, and citation patterns. Notable insights reveal a surge in Large Language Model research, rising from zero before 2020 to 19% of papers by 2024, alongside foundational work in Machine Learning, Computer Vision, and Optimization. The author community appears hourglass-shaped, with 80.5% of 2,623 authors contributing to only one edition, while institutions show a nearly threefold return rate. Citation concentration is evident, with the top 1% of papers accounting for 27% of all citations. Additionally, practices of openness have expanded, particularly in artifact sharing. This study offers a thorough overview of the Brazilian AI research environment, emphasizing trends in topics, authorship, and citation influence.
Key facts
- BRACIS is the main national venue for AI research in Brazil, hosted by the Brazilian Computer Society since 2012.
- The study covers 1,066 accepted papers from 2015 to 2025.
- Data sources include DBLP metadata, 6,765 Google Scholar citations, and full texts.
- Large Language Model research grew from zero before 2020 to 19% of papers in 2024.
- 80.5% of 2,623 authors appear in only one edition of the conference.
- Institutions return at nearly three times the author rate.
- The top 1% of papers account for 27% of total citations.
- Openness practices, including artifact sharing, have increased over the period.
Entities
Institutions
- Brazilian Computer Society
- arXiv
Locations
- Brazil