Beyond Biases Towards Attention Ethics

Algorithmic fairness has become such a central topic in AI ethics that many experts tend to limit the entire field to this topic alone. As a downside of this success story, discussions around algorithmic fairness and “debiasing” methods have often lost grasp of concrete applications and the sight of the ethical reflections they emerged from. This often results in meaningless applications and some confusion, for instance, between cognitive and algorithmic biases. This thought-provoking talk recalls the fundamental questions behind algorithmic fairness, the questions that researchers and practitioners tried to address when the concerns emerged, and throws light on how the research slowly deviated from its original ambitions to instead focus on the technicalities of debiasing methods alone. Adopting the perspective of an industry practitioner, it presents concrete examples from real use cases of Generative AI-based products, presenting tensions that require moral arbitrations. Finally, Attention Ethics sheds a whole new light on the literature on algorithmic biases from an anthropological viewpoint, thereby helping us overcome the limitations of our current ethical approaches to fairness in AI.