Social Cognition is a rich, multi-disciplinary field: from ethology to philosophy, from neurosciences to sociology, it crystallizes hard scientific questions across a broad range of fields.

In parallel, since 2022 and the first release of ChatGPT, a new range of machine learning techniques are completely transforming our understanding of AI: tasks that we thought out-of-reach to machines a few years ago (like translation, image or music generation, text summarization) are now considered as basic Generative AI capabilities.

At the SocialMinds AI Lab, we are asking ourselves the question:

What would it take to bridge Social Cognition and Generative AI?

We want to close several critical gaps in knowledge:

• how can we make the current generation of GenAI models – including Visual-Language-Action (VLA) models – aware of humans and social interactions?
• social cognition is tightly related to embodied cognition: how can Generative AI be given an ’embodiment’?
• to closer we get to the social sphere, the more ethics questions we raise: what a caring, human-first ‘Social AI’ might look like?
• developing new powerful AI for the sake of it is not sufficient: how can we build a desirable future for a society where humans can co-exist with artificial social agents?

We address these questions through three main research activities: