Our talk is framed within the research project "Towards non-binary face recognition: a critical study on gender and identity to build a more equitable AI," which is conducted at the Department of Computer Sciences at VU Amsterdam. It addresses the rapid proliferation of automatic face analysis in our society. While this technology, such as facial recognition, is primarily utilized for security and law enforcement purposes, it is increasingly being operative in other domains such as recruitment, education, and facial expression analysis. However, facial recognition systems consistently adhere to a gender binary framework and rarely account for individuals who identify as non-binary. Consequently, these types of human-machine interfaces reinforce existing prejudices against these communities.
By examining crucial questions about the circumstances under which digitalization shapes knowledge and identities, our hypothesis is that non-binary databases are absent in facial recognition systems. In light of these issues, the objective of our talk is to create a focused platform for discussing research challenges, concerns, and solutions associated with the development of inclusive facial recognition systems. This will be achieved through a critical analysis of the interplay between gender, identity, and face recognition technologies. Our aim is twofold: i) to encourage an interdisciplinary discourse on non-binary identities and face recognition technologies to promote the development of inclusive, diverse, and trustworthy AI, and ii) to highlight the dichotomy between self-perceived gender and machine-classified gender.
In the second part of our talk, we will analyze an audiovisual artwork titled "Zizi Queering the Dataset" by Jake Elwes. This piece showcases different facial portraits in a morphing loop, visualizing what a Generative Adversarial Network has learned from re-training a dataset that includes portraits alongside facial images of drag and non-binary individuals. This artistic endeavour will raise a series of epistemic issues regarding big data and their situated and ideological significance.