Why ‘SIM’?

sim-lab jul 19, 2025

Why are we called ‘SIM’ lab? We study social intelligence in human development and evolution, but also in story characters and non-human intelligent systems, including bots and AI-models.

SIM stands for social intelligence modelling, which is reflected in our research using agent-based and classification models.

But it also stands for stories in the mind, reflecting what social cognition in essence is: a form storytelling in which we (silently or out loud) connect visible elements of the social world to invisible ones, such as beliefs, intentions, desires, and emotions. We work with children aged 4-12 to study the link between storytelling and forms of social cognition known as Theory of Mind or Intentional Reasoning. And we train language/AI models using narrative data to study the link between stories and intelligence in a completely novel way.

We also study language/AI models that collaborate with humans on various tasks. Here, SIM refers to simulation: in order to coordinate actions successfully, both sides have to simulate the others’ perspective.

Two talks @ AAAI ToM Workshop in Philadelphia

sim-lab mrt 3, 2025

Theory of Mind (ToM) is a skill that is used in both human and AI cognition to infer the mental states of others, and to use these mental states to predict their behaviour in the world. Experts from all over the world joined an AAAI conference workshop in Philadelphia to discuss the benefits and challenges of using ToM in AI systems. The diversity in the field was represented by 4 keynote speakers, 7 spotlight speakers, and 33 poster presentations during a day full of exchanged ideas. 

Ramira van der Meulen joined this event as a spotlight speaker, to highlight the misconceptions that exist around implementing ToM in AI. In addition, she presented a poster on how different models of intelligence influence the learning process and knowledge retention of both individual and collectives of AI agents. Overall, it was an inspiring day with many insights on how ToM could be practically useful in AI applications, and we look forward to potential future sessions. 

The two presented works can be found as included extended abstracts in the Workshop Proceedings. A longer version of the work on AI-ToM Misconceptions can be found on ArXiv.