EPFL AI Center and CoALA Lab

Alexandra Olteanu examines how often unsupported beliefs and assumptions about what AI systems can, will or should do shape how those systems are built, evaluated, described and deployed, and therefore when they fail. Addressing the consequences of insufficiently rigorous AI artefacts such as models, systems, applications, outputs and data is often treated as the separate concern of responsible AI. She argues instead that responsible AI asks researchers and practitioners to uphold scientific integrity, so rigour is within scope for every AI scientist and practitioner, and offers a framework for broadening what rigorous AI research and practice should entail and for scrutinising current development and evaluation practices.