2022
Territories of Exception
Country/area: Brazil
Organisation: Documental.xyz
Organisation size: Small
Publication date: 19/4/2021

Credit: Coordination: Fernanda Bruno, Adriano Belisario e Paulo Tavares
Data analysis and visualisation: Adriano Belisario
Research assistant: Ingra Maciel e Debora Pio
Web developer: Marlus Araújo
Biography: Carried out by MediaLab.UFRJ and Agência Autônoma, in partnership with Redes da Maré, Fogo Cruzado, Pista News, Witness e Rede LAVITS. This project was conceived in 2017 by MediaLab.UFRJ and Agência Autônoma as part of the investigative collaborations driven by data and methodologies of forensic architecture. The work relied on the support of Fundação de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ) and the Ford Foundation.
Project description:
From 2019 to 2021, our team studied human rights violations and the use of helicopters as vantage points from which to shoot at civilians during police raids in Rio de Janeiro. We have created an original database with over 700 events, 3d models, and new analysis using machine learning techniques on social media posts to reveal the impacts of such practice.
MediaLab.UFRJ and Agência Autônoma have developed the project in partnership with grassroots organizations and local NGOs. The story was published by Documental.xyz, a platform for media and data-driven narratives and stories on human, territorial and environmental rights.
Impact reached:
The research was used by legal claims in the local justice system. These lawsuits seek to implement harm reduction measures for law enforcement raids. The project was presented internationally at events such as RightsCon. Also, our story has been shortlisted by Prêmio Gabo and won a local journalistic award.
Techniques/technologies used:
We applied several techniques such as open-source intelligence (OSINT), fieldwork, investigative journalism methods, 3d modeling, or natural language processing (NLP) to reveal territorial patterns and psychological impacts. There is a description below of essential techniques and tools:
1) OSINT: we used visual investigation methodologies, geolocation, and advanced web search strategies to contextualize evidence captured by local citizens. Video and photos collected in social media were fundamental to understanding our research topic better.
2) 3d modeling: Inspired by forensic architecture techniques and based on the visual materials gathered, we created 3d models to spatialize those videos and photos.
3) NLP: after collecting thousands of Twitter posts with specific keywords, we applied topic modeling algorithms to identify testimonials about shootings from police helicopters. This technique allowed us to clean our dataset and revealed hidden patterns such as sleeping disorders and psychological impacts caused by those raids.
4) Fieldwork: thanks to a partnership with a local NGO (Redes da Maré), we developed fieldwork strategies to geolocate critical evidence and acquire new materials about human rights violations at a local level.
5) Data analysis: we used traditional techniques and tools for tabular data, like spreadsheets or Python/R codes, to explore and describe the information we gathered, including unpublished governmental data about aircraft employment as shooting platforms.
What was the hardest part of this project?
Like other brazilian cities, the city of Rio de Janeiro has been going through a strong conservative wave, led mainly by government agencies. These institutions have been working to withdraw basic rights from the population, especially the poorest. Therefore, the most hardest part of this project was not to put at risk the residents of these territories, who helped in a significant way to carry out the research. When a slum dweller accuses the security forces of some violation of human rights, he or she can be threatened and suffer serious retaliation.
What can others learn from this project?
We obtained and structured critical governmental datasets to understand police operations in Rio de Janeiro. Now, it is organized in our Github repository [1] among dozens of scripts, other tables, geodata, and resources for data journalists. Exploring our materials can bring insights for data-driven journalists interested in human rights worldwide. For instance, we have created open Python notebooks for fact-checking our data-driven conclusions.
The platform we used (documental.xyz) is innovative and allows data visualization to be easily understood by anyone despite any previous knowledge about the territory studied. Therefore, our methodology to contextualize visual evidence can set an important precedent for other local organizations. Our methodology also can inspire innovative uses of geodata, photos, videos, and 3d modeling to contextualize human rights violations.[1] https://github.com/medialabufrj/territorios-de-excecao