A drone-based tool developed by the University of São Paulo offers a more accurate crowd estimate at recent demonstrations in support of former President Jair Bolsonaro, challenging official figures and highlighting the importance of reliable measurements in political contexts.
In late February, a significant gathering of demonstrators supported former President Jair Bolsonaro on Paulista Avenue, São Paulo. Bolsonaro’s team claimed over half a million attendees, while the São Paulo state’s public security department estimated 600,000 people. However, researchers at the University of São Paulo (USP) utilized a drone-based tool with artificial intelligence to suggest the actual number was closer to 185,000.
This newly developed tool, created by political scientist Pablo Ortellado and his team, aims to provide more accurate crowd estimates and was adapted from a Chinese framework initially used for counting cattle. Ortellado stressed the unreliability of traditional methods of crowd estimation, which led to the development of the tool to offer a more objective measurement.
Accurate crowd size estimation is crucial in political environments, as highlighted by Steven Feldstein from the Carnegie Endowment for International Peace. Discrepancies in numbers can significantly affect perceptions of political movements. Such issues have been particularly evident in Brazil, where political unrest has grown following Bolsonaro’s ban from running for office until 2030 and his involvement in an attempted coup.
The tool has sparked interest among Brazilian analysts and media outlets for its potential to reduce political bias in crowd size reporting. Fabio Wajngarten, Bolsonaro’s lawyer, has argued that the tool underestimates the crowds at Bolsonaro’s rallies. The São Paulo Public Security department confirmed using multiple methods for crowd estimation but did not address discrepancies between their figures and those of the USP tool.
