Generative AI influences how people learn, work, think and make decisions; as well as how they socialise and relate to one another. Yet its effects on well-being remain largely unmeasured, as the evaluation of frontier AI applications still focuses on what these models can do, rather than on what they do to the people who use them.
This webinar, organised by the OECD Centre for Well-being, Inclusion, Sustainability and Equal Opportunity (WISE) in collaboration with the MIT Media Lab, will present two complementary instruments designed to close that gap.
The first is ImpactBench, an open, expert-guided benchmark launched in October 2026 by the MIT Media Lab’s Advancing Humans with AI (AHA) program. Built with more than 80 experts from over 40 institutions, including clinicians, educators, legal scholars and community advocates, ImpactBench tests frontier models in thousands of simulated multi-turn conversations, using 809 behavioural metrics across 26 benchmarks of psychological, physical and social well-being.
The second is the MIT–OECD Global Panel on AI and Well-being, an initiative under development in the OECD WISE Centre. Organised as a two-wave survey across OECD and partner countries, the Panel will aim to measure people’s well-being, their use of AI, the agency they retain over it and the impact they attribute to it, in order to estimate AI’s causal effects on people’s lives.
For further inquiries, please contact wellbeing@oecd.org.
