University of Twente Student Theses


Automated Misgendering : an Inquiry Into the Ethics of Automatic Gender Recognition

Hoefsloot, Marit Eva (2021) Automated Misgendering : an Inquiry Into the Ethics of Automatic Gender Recognition.

[img] PDF
Abstract:This thesis investigates the ethics of the use of automatic gender recognition technologies (AGR). AGR tools are a type of facial analysis technology used to identify and verify someone’s gender identity. The research question for this thesis is: Is the use of AGR technologies in their most accurate and fair form ethical? Moreover, how can the negative impact of AGR’s use be alleviated pragmatically? To answer these questions, I will research how and for what purpose the technology is built, investigate the underlying assumptions regarding gender, and discuss the harmful consequences of the (mis-)use of AGR technologies. The first chapter will show that AGR technology is built on the assumptions that gender can be externally determined based on one’s appearance, that gender is a binary categorisation, and it is stable over time. In the second chapter, I argue that these assumptions are incorrect, as self- identification is central to gender identity and people must have first-person authority and autonomy over their gender and their gender alone (Bettcher, 2009). This claim is not an epistemic one but a moral one; while you might disagree with someone’s gender identity, you are obliged to respect their authority on the matter. Finally, in the third chapter, I argue for the immorality of the use of AGR technologies. It fundamentally disrespects the authority and agency people have over their gender. As the developers and users of AGR are set to make significant financial and time efficiency gains due to its broad implementation (O’Neill, 2021), it is not realistic to aim to abolish the technology. Thus, I argue that there is another option. Through the education of AGR’s users concerning the nature of the results, the harmful consequences of using AGR tools can be minimised. This education involves creating a new vocabulary to reflect that the results from AGR technologies are merely predictions with probabilities, which can be either accurate or inaccurate. I call the results created by AGR tools probabilistic information and the information given by the subjects themselves agential information to reflect Dembroff and Saint-Croix’s agential identity (2019). Through such vocabulary, we become more aware of the authority behind the pieces of information. As the agential information must be respected, the new vocabulary can help resolve conflicts in situations where the AGR result does not correspond to someone’s self-identified gender. Keywords: automatic gender recognition, gender identity, agential identity, first-person authority, ameliorative analysis
Item Type:Essay (Master)
Faculty:BMS: Behavioural, Management and Social Sciences
Subject:08 philosophy
Programme:Philosophy of Science, Technology and Society MSc (60024)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page