Amplifying bias, automating racism
Ethically framing the issue of algorithmic discrimination
Abstract
Riding the wave of the current #BlackLivesMatter movement, in this work I present an example of racial-based discrimination resulting from a particular kind of algorithmic bias, known in literature as amplification bias. Then, in order to stimulate debate on how to effectively address this phenomenon in automated decision-making systems, I explore a case study reporting it, I analyse why it is so problematic, and I stress the need to reflect on ethical solutions to fix it. In this paper, I argue that we cannot solve harmful consequences of algorithmic bias to the root without adopting an ethically-based approach. Hence my discussion emphasizes the need to urgently address these issues, proposes actions to do that and underlines the importance of rules and norms to keep human beings accountable.
