Uncovering biases in AI: a study on facial recognition and feminist futures

Abstract

Technology, particularly Artificial Intelligence (AI), often proves its potential to revolutionize the future. However, it is crucial to recognize that AI is not without its biases. AI learns from data, and data is a reflection of our history – structurally organized around racist and misogynist ideologies. In this study, we want to explore the issue of bias in AI, with a focus on facial recognition. To do this, we identified that the structure of the algorithm itself is not inherently racist or sexist. What happens is that, as pointed out by O'Neil (2017), the data that feeds these algorithms incorporates our past, including dark aspects of discrimination and prejudice, characterizing them as a new expression of racism (LIPPOLD; FAUSTINO, 2022) and the racialization of subalternized bodies. Therefore, one of the biggest challenges is the problem of mitigating biases and prejudices related to the auditing and transparency of these systems (VIEIRA, 2023). Although technological biases – based on a supposed white, male and Western supremacy – existed before the development of algorithms, we work with the following question: how could algorithms, from their applications, rid humanity of these biases? Is there a future possibility, or is AI doomed to perpetuate these biases under the guise of technological neutrality? The aim of this paper is to shed light on these questions and contribute to a future in which AI can be used in an equitable way, promoting futures based on social justice, feminist agendas and freedom from racial discrimination.