The Manchester Centre for AI Fundamentals and Manchester's ELLIS Unit are co-hosting a series of seminars featuring expert researchers working in the fundamentals of AI.
Title: Fairness in online systems
Abstract: As online systems increasingly shape our digital experiences, from job recommendations to advertising and content moderation, ensuring fairness in these systems has become a critical challenge. Biases in data, algorithms, and user interactions can lead to unintended disparities, reinforcing societal inequalities. Traditional approaches, such as fairness through unawareness that ignore sensitive attributes often fail to address underlying biases, while real-world constraints, such as data limitations and competing optimization goals, add further complexity. In this talk, we explore fairness in online systems, focusing on bias detection, mitigation strategies, and trade-off between fairness and business objectives. Using real-world data from an online advertising platform, we highlight key challenges, recent advancements, and open questions. We aim to bridge the gap between theoretical fairness frameworks and their practical implementation in large-scale, real-world systems.
Bio: I am a senior research scientist at Criteo AI Lab in the Fundamental Deep Learning team. I contribute to both academic and industrial research, focusing on explaining and improving machine learning methods through probabilistic and causal statistical approaches. At Criteo, I lead research initiatives aimed at ensuring algorithmic fairness in online systems and provide consulting to policymakers on AI regulations.
In addition to my professional endeavors, I am an advocate for gender diversity and inclusivity in the tech industry, actively promoting and organizing events to foster equitable opportunities and representation.
Before joining Criteo, I did my postdoc at Inria Grenoble Rhone-Alpes in the Statify team. The goal was to continue my previous research on the exploration distributional properties of Bayesian neural networks. More specifically, I was interested in explaining the difference between deep learning models of wide and shallow regimes in order to improve the interpretability and efficiency of the models.
I obtained my PhD degree in applied mathematics in 2022 at the University Grenoble Aples and Inria reseach center. I was a part of Statify and Thoth teams, under supervision of Julyan Arbel and Jakob Verbeek. During November 2019-January 2020, I was visiting Duke University and working on prior predictive distributions in Bayesian neural networks under supervision of David Dunson. Prior to that, I obtained my Bachelor degree at Moscow Institute of Physics and Technology (MIPT) and did the second year of Master program at Grenoble Institute of Technology (Grenoble - INP, Ensimag).
