Eirini Ntoutsi
Full professor at the Bundeswehr University Munich
Abstract: How to make AI more fair and unbiased
AI-driven decision-making has already penetrated into almost all spheres of human life, from content recommendation and healthcare to predictive policing and autonomous driving, deeply affecting everyone, anywhere, anytime. The discriminative impact of AI-driven decision-making on certain population groups has been already observed in a variety of cases leading to an ever-increasing public concern about the impact of AI in our lives. The domain of fairness-aware machine learning focuses on methods and algorithms for understanding, mitigating, and accounting for bias in AI/ML models. Most of the work in this field focuses on limiting learning settings, typically binary classification with a binary protected attribute. The reality is however more complex, for example, discrimination can occur based on more than one protected attribute, the class distribution might be imbalanced, the population characteristics might change, and more than one learning task might need to be solved at the same time.
In this talk, I will talk about fairness in supervised learning covering the basic binary class mono-discrimination setting as well as works towards more realistic challenges like discrimination for multiple protected attributes, discrimination under class imbalance and multi-task discrimination.