Two MAGNET researchers presented work on gender measurement at the conference "Better Data for Better Jobs and Lives: Innovations in Survey Measurement in the Age of AI," held at World Bank Headquarters in Washington, D.C. on December 8-9, 2025.
Aletheia Donald (Senior Economist, Africa Gender Innovation Lab, World Bank) presented Whose Job Is it? Implicit Gender Bias Toward Occupations in India and Uganda on December 9. The paper adapts the Affect Misattribution Procedure (AMP) for use in low- and middle-income country field settings to measure implicit bias toward gender-nonconforming occupational choices. The study deployed the AMP with women and men across India and Uganda. Respondents viewed occupational primes (either text-based or AI-generated visual images) and then rated the pleasantness of neutral images. The research found that 44% of respondents in India and 54% in Uganda exhibited bias toward gender-incongruent occupational roles, with effects varying by country context and labor market characteristics.
Losira Nasirumbi Sanya (Makerere University) presented Measuring Women's Agency with Unsupervised Machine Learning Approaches on December 9. The study explored whether computational linguistics can provide scalable tools for quantifying agency from qualitative interview data. Using recordings of 303 respondents answering open-ended questions about personal and household decisions in Uganda, the research team applied unsupervised machine learning to identify linguistic patterns associated with agency. The frequency of the pronoun "I" emerged as the most consistent indicator of agency among the approaches tested so far, with women who positioned themselves as grammatical subjects more frequently tending to exhibit higher agency scores.
Both presentations demonstrated applications of AI and machine learning to gender measurement challenges in survey research.
International Food Policy Research Institute