Artificial intelligence (AI) is driving profound changes that can be seen in many areas—economics, healthcare, finance, the military, and more. Alongside the vast opportunities it presents come numerous risks that we still struggle to predict or even perceive. Because AI now permeates every corner of human activity, we must build a framework of trust that ensures its outputs respect fundamental rights and freedoms. In 2023, an open letter signed by global experts and major industry leaders called for AI research and development to focus more on “making today’s powerful, state-of-the-art systems more accurate, safe, interpretable, transparent, robust, aligned, trustworthy, and loyal.” The authors also stressed the urgent need to work with policymakers to speed the creation of a solid governance framework for AI—one that includes new specialized regulatory authorities as well as audit and certification measures. The European AI Act marks a first step in this direction.
Regulating, auditing, and improving AI systems poses an unprecedented scientific challenge. Rising to it demands collaboration across many disciplines—computer science, mathematics, law, economics, and more. Our team is part of this research movement: we bring together expertise in machine learning, law, mathematics, optimization, computer science, and economics to create a fertile environment for reflection that addresses today’s grand challenges of trust and regulation. Our goal is to help lay the groundwork for a sustainable, ethical, and responsible future for AI technology, transforming the way these systems are managed and governed.
Research objectives
- From global to local measures of biases
- Understanding the multi aspects of bias in the data and in the optimization
- New robust bias mitigation methods
- Value alignment of AI algorithms
- Regulations and Self-Regulation of AI
- Trust through Legality
Industrial applications
- Loréal & Artefact, Fairness for Image Recognition , CIFRE Phd J-M. Loubes
- Supported by INRIA’s Project REGALIA
- CONTINENTAL, Anomaly detection in evolving environments – Application to production lines, PhD thesis supervised by J.-M. Loubes and L. Travé-Massuyès
- Supported by US Air Force grants (J. Bolte)
- EFELIA grants for teaching (J. Eynard, J-M. Loubes)
- IUF grant fellow (Pauwels co-chair)
Main scientific goals
- New methods to detect and certify disloyal behaviour of AI algorithms
- New methods to mitigate disloyal behaviors
- Legal proofs of compliance of the algorithm
- Guidelines for best practice in economy