Unlearning Gender - (Re)using Computer Vision for Dissolving the Social Construct of Gender through Artistic Practices

Masterthesis, 2024, CC BY-NC-ND 4.0 International Supervision: Manuela Naveau, Penesta Dika, Sebastian Fitz-Klausner Abstract This thesis analyses how computer vision manifests and reproduces gender binaries and how artistic practices are able to deconstruct gender binaries by reusing this technology. In particular, it focuses on automatic gender recognition systems because of their clear and pronounced reinforcement of the gender binary. Therefore, exploring how these systems function provides an insight into the social ideas, norms and assumptions that underpin the gender binary. Thus, by understanding their algorithmic categorisation, we have the opportunity to better understand ourselves as a society. This thesis argues that AGR-systems follow a biologistic and essentialist logic, and that their supposed recognition is only partly based on physical characteristics and may rely more on gender stereotypes for their categorisation than studies have suggested. This indicates the social imprint of the systems, and a parallel can be drawn to the binary gendered perception and categorisation of people by people that is common in society, as both human and machine categorisation may be traced back to gender performance and gender stereotypical behaviour rather than physiological factors. Artistic reuse, in turn, demonstrates the potential to be used as a political tool to stimulate social change and rethinking, thereby helping to denormalise harmful gender norms. This is explored and illustrated through the artwork "Unlearning Gender", an interactive installation that reuses OpenCV's AGR-system to deconstruct the binary gendered gaze of the machine, and to provide an example of what non-gendered personal perception might look like. Moreover, it also analyses the artistic approaches of other artworks in the context of challenging heteronomous social constructs and their reproduction through technology and compares them to "Unlearning Gender" to identify their strengths and weaknesses in achieving artistic and political goals.