2025-02-09-世界银行-性别偏见_公民参与和人工智能（英）_30页_982kb

The study analyzes the role of gender bias in citizen participation using AI-driven methods, specifically decision trees, on data from the 2023 Latinobarómetro Survey. It identifies three forms of gender bias—societal (measured by machismo norms), data, and algorithmic—as significant factors hindering fair analysis of civic engagement. Key findings show that individuals with low gender bias, high education, political interest, and left-wing views are more likely to participate, with a specific machismo threshold of below 0.214 for higher likelihood. Gender bias persists even after applying corrective measures, suggesting that both societal and institutional factors must be addressed simultaneously. The research emphasizes the need for integrated strategies, including technical AI mitigation and policy interventions, to promote inclusive participation and reduce the gender gap. Overall, AI can enhance understanding of complex social dynamics but requires careful design to avoid reinforcing biases.