Uploaded November 2025 | Updated September 2026, 2 hours ago
In this presentation, Marco Germanò discusses a study that examines how AI systems are reshaping the production of truth in contemporary societies through predictive and classificatory outputs that increasingly serve as epistemic anchors in public decision-making. While often framed as neutral tools, these systems are embedded in digital infrastructures and sociotechnical imaginaries predominantly shaped by institutions in the Global North. In Global South contexts, their adoption thus introduces not only material asymmetries, but also epistemic impositions in which situated ways of knowing and acting are typically encoded and institutionalized without effective local input. Building on postcolonial theories of epistemicide and accounts of epistemic injustice, Germanò introduces the concept of algorithmic epistemicide to describe how such AI systems displace or overwrite situated knowledges through probabilistic logics that claim objectivity while eluding avenues for contestation. Rather than framing such harms as merely distributive or representational, as part of the existing literature has done, Germanò seeks to foreground the ontological effects of algorithmic systems that restructure what counts as evidence, risk, or eligibility in public life through largely proprietary, opaque processes.
Germanò develops his argument through two case studies in Brazil, drawing on desk research and fieldwork conducted with local stakeholders. The first examines predictive policing tools piloted in Rio de Janeiro through public-private initiatives, in which AI-generated crime maps inform public security strategies and legitimize intensified policing in favelas, often at odds with local understandings of safety and violence. The second analyzes the use of machine learning models in welfare fraud detection, particularly within the Cadastro Único and Auxílio Brasil programs, where algorithmic classifications have led to the exclusion of low-income families based on opaque data correlations. Across both cases, the study demonstrates how algorithmic outputs become treated as authoritative while grassroots actors—including public defenders, civil society organizations, and local communities—develop counter-narratives and practices of resistance that challenge the epistemic legitimacy and legal authority of these systems. Through this work, Germanò expects to contribute a theoretical framework for understanding algorithmic systems not merely as tools of governance but as epistemic agents that participate in reorganizing everyday conditions of intelligibility. This conceptual lens enriches discussions regarding how AI is lived, contested, and possibly reimagined across the Global South.
In this presentation, Marco Germanò discusses a study that examines how AI systems are reshaping the production of truth in contemporary societies through predictive and classificatory outputs that increasingly serve as epistemic anchors in public decision-making. While often framed as neutral tools, these systems are embedded in digital infrastructures and sociotechnical imaginaries predominantly shaped by institutions in the Global North. In Global South contexts, their adoption thus introduces not only material asymmetries, but also epistemic impositions in which situated ways of knowing and acting are typically encoded and institutionalized without effective local input. Building on postcolonial theories of epistemicide and accounts of epistemic injustice, Germanò introduces the concept of algorithmic epistemicide to describe how such AI systems displace or overwrite situated knowledges through probabilistic logics that claim objectivity while eluding avenues for contestation. Rather than framing such harms as merely distributive or representational, as part of the existing literature has done, Germanò seeks to foreground the ontological effects of algorithmic systems that restructure what counts as evidence, risk, or eligibility in public life through largely proprietary, opaque processes.
Germanò develops his argument through two case studies in Brazil, drawing on desk research and fieldwork conducted with local stakeholders. The first examines predictive policing tools piloted in Rio de Janeiro through public-private initiatives, in which AI-generated crime maps inform public security strategies and legitimize intensified policing in favelas, often at odds with local understandings of safety and violence. The second analyzes the use of machine learning models in welfare fraud detection, particularly within the Cadastro Único and Auxílio Brasil programs, where algorithmic classifications have led to the exclusion of low-income families based on opaque data correlations. Across both cases, the study demonstrates how algorithmic outputs become treated as authoritative while grassroots actors—including public defenders, civil society organizations, and local communities—develop counter-narratives and practices of resistance that challenge the epistemic legitimacy and legal authority of these systems. Through this work, Germanò expects to contribute a theoretical framework for understanding algorithmic systems not merely as tools of governance but as epistemic agents that participate in reorganizing everyday conditions of intelligibility. This conceptual lens enriches discussions regarding how AI is lived, contested, and possibly reimagined across the Global South.
