Abstract
This paper investigates the role of gender bias in AI-driven analyses of citizen participation, using data from the 2023 Latinobarómetro Survey. We propose that gender bias—whether societal, data-driven, or algorithmic—significantly affects civic engagement. Using machine learning, particularly decision trees, we explore how self-reported societal bias (i.e., machismo norms) interacts with personal characteristics and circumstances to shape civic participation. Our findings show that individuals with reportedly low levels of gender bias, who express political interest, have high levels of education, and align with left-wing views, are more likely to participate. We also explore different strategies to mitigate gender bias in both the data and the algorithms, demonstrating that gender bias remains a persistent factor even after applying corrective measures. Notably, lower machismo thresholds are required for participation in more egalitarian societies, with men needing to exhibit especially low machismo levels. Ultimately, our research emphasizes the importance of integrated strategies to tackle gender bias and increase participation, offering a framework for future studies to expand on nonlinear and complex social dynamics.
Presenters
Jose CuestaLead Economist and Global Lead for Data and Analytics, Social Development, World Bank, District of Columbia, United States
Details
Presentation Type
Paper Presentation in a Themed Session
Theme
KEYWORDS
Gender bias, Machine Learning, Participation, Latin America
