Luc Rocher
Associate Professor, University of Oxford
Jan 22, 2026
10:00 am
EST
Jan 22, 2026
10:00 am
In-person
One Portsoken
514
Portsoken Street
London, E1 8PH, UK
London, E1 8PH, UK
The Roux Institute
Room
514
100 Fore Street
Portland, ME 04101
Portland, ME 04101
Network Science Institute
2nd floor
2nd floor
Network Science Institute
11th floor
11th floor
177 Huntington Ave
Boston, MA 02115
Boston, MA 02115
Network Science Institute
2nd floor
2nd floor
Talk recording
AI systems are increasingly deployed in social settings such as health advice, therapy, and companionship. This public adoption can seem justified given widely reported success stories of models’ capabilities, and strong performance on Machine Learning benchmarks. In this talk, we will take a more fundamental step back to investigate AI adoption: how should we test and evaluate these models, in ways that genuinely predict their performance and risks in the wild. We will compare standard practices in Machine Learning research with best practices from the social sciences, through a series of recent studies conducted at the Synthetic Society Lab.
About the speaker
Luc leads the Synthetic Society Lab, a research group working to make technology and digital power accountable to the public, and guide the development of accountable, sustainable, and safe algorithms that serve the public interest.Their research investigates the risks posed by large-scale collections of digital human traces from social media to biometrics as well as deployed AI technologies, identifying gaps in how technology is regulated and how risks are documented, and proposing better models for academic research using sensitive human data. Prior to joining Oxford, Luc received a PhD from the Université catholique de Louvain in 2019 and worked as a researcher at the Data Science Institute and Computational Privacy Group of Imperial College London, at the ENS de Lyon, and at the MIT Media Lab.
Share this page:
