Decoding the Silence: Using AI to Map 'Epistemic Silences' in State Policy

Traditional M&E frameworks are overwhelmingly biased toward measuring what is visible and actively reported such as bulk infrastructure budgets or the number of communal taps installed. However, the greatest threats to vulnerable populations often hide in what the state actively chooses not to measure, such as the dangerous physical distance a child must walk to access that water. This masterclass demonstrates how AI can transcend basic quantitative number-crunching to serve as a high-level tool for Critical Discourse Analysis. Drawing on recent longitudinal research into South African political manifestos, this session will teach evaluators how to train AI models to detect "epistemic silences" and structural vulnerabilities in government policy. By leveraging frameworks like Joan Tronto's Ethics of Care, we will show how AI can expose the discursive disconnect between state promises and material reality.