Last week, several WHAT-IF work package team members met in Konstanz to discuss the core of our WHAT-IF machine. The two-day workshop brought together the WP4 team (modeling & simulation) of David Garcia (University of Konstanz) and Annie Waldherr (University of Vienna) as well as colleagues from WP5 (immersive experiments) and WP3 (data collection). 

At the heart of the workshop was the conceptual design of the WHAT-IF machine—an agent-based simulation that combines theoretical knowledge, experimental findings, and agents based on generative AI to explore interventions to improve democratic discourse on digital platforms.

The workshop clarified and helped align the complementary experimental and simulation approaches within the larger WHAT-IF project. Small-scale interaction experiments will inform the baseline agent-based model and large-scale simulations of platform dynamics, by combining existing knowledge and empirical findingsto map who interacts where, with LLMs to simulate what is said. Early experiments using general LLMs can guide the design of more tailored agent behaviors. Such models will focus on how users respond—by engaging, reporting, or withdrawing—depending on the civility and like-mindedness of the conversation.

Curious to see what will come!

In the photo (left to right): Wouter van Atteveldt, Aytalina Kulichkina, Constantine Boussalis, Alessandro Bellina, Rupert Kiddle, Max Pellert, Giordano De Marzo, Andri Rutschmann, Annie Waldherr, David Garcia, Taehee Kim