Am 18.12.2023 von 15-17 Uhr finden die nächsten BANDAS-Lectures im SR 15.37 (RESOWI F3) statt.
The power of (and in) bias tests. A Twitter case study.
Paola Lopez vom Institut für Rechtsphilosophie an der Universität Wien
In 2020, the machine learning algorithm that was deployed by Twitter to generate cropped image previews was accused of carrying a racial bias. Twitter users complained that Black people were systematically cropped out and, thus, made invisible by the cropping tool. As a response, Twitter conducted bias analyses and removed the cropping tool. Soon after, the company hosted an “algorithmic bias bounty challenge” inviting the general public to detect algorithmic harm and be rewarded for their findings. This paper examines in Foucauldian terms the push-and-pull dynamics of the power relations that are at play: Firstly, it studies the algorithmic knowledge production around the cropping tool, secondly, the bias analyses and their epistemic limitations, as well as the bias discourse as a vehicle for resistance, and, thirdly, the way in which Twitter as a company effectively stabilized its position – rendering the bias discourse a vehicle for counter-resistance, too.
Network Inequalities in Academia.
Fariba Karimi, Professorin für Data Science an der TU Graz
Gender inequality in science remains prevalent, and top-down, quick-fix approaches have proven ineffective. While societal biases against women in academia are often implicated in this inequality, the influence of peer dynamics and social networks in perpetuating or supporting these biases remains largely unexplored. We use large-scale scholarly publications, collaborations, and citations over the decades to shed light on structural drivers of gender inequalities in academia. We will discuss how we can use computational modeling approaches to examine the effectiveness of interventions in overcoming those inequalities.