11May2026
14:00 Master's Defense IC3 Auditorium
Topic on
From Face to Vector: Computational Epistemicide in Facial Recognition
Student
Ana Carolina das Neves Silva da Hora
Advisor / Teacher
Sandra Eliza Fontes de Avila - Co-supervisor: Marisol Marini
Brief summary
This dissertation investigates the application of Artificial Intelligence techniques in facial recognition, analyzing them not only as measurement tools but also as sociotechnical devices that reproduce structural asymmetries. The work extends Sueli Carneiro's theory, specifically the concept of epistemicide and the production of the other as non-being, to propose the concept of "computational epistemicide." This extension demonstrates how the algorithmic conversion of the face operates the erasure of the subject and the annulment of its epistemic legitimacy. The theoretical framework articulates this basis with Denise Ferreira da Silva's critique of racial transparency, using the concepts of dispositif (Foucault) and faciality (Deleuze and Guattari) to examine the technical machinery. Methodologically, this critical lens is challenged through empirical experiments in the Casual Conversations v2 database, auditing detection architectures (Haar Cascades, MTCNN, and YOLOv11). The results indicate that the preprocessing and geometric normalization stages are central to defining readability, operating exclusion even before identity recognition. Finally, the study discusses guidelines for computational repair that integrates ethical responsibility into the design of architectures.
Examination Board
Headlines:
| Sandra Eliza Fontes de Avila | IC / UNICAMP |
| Angela Lucia Silva Figueiredo | CAHL/UFRB |
| Virgílio Augusto Fernandes Almeida | DCC / UFMG |
Substitutes:
| Mariza Ferro | IC / UFF |
| Rafael de Almeida Evangelista | Labjor/UNICAMP |