At the core of TWON is the development of a digital twin of an online social network. This model replicates the structural and interaction dynamics of real platforms. The virtual user population is created using generative AI agents that produce, share, and interact with content, mimicking realistic user behavior. This approach enables researchers to analyze complex online discourse dynamics without relying on data access or cooperation from platform providers.

The simulations are complemented by empirical case studies examining real-world debates and platform dynamics, for example around political conflicts or societal crises. Based on these analyses, the project develops metrics for assessing the quality of online debates, including indicators such as exposure to political content, diversity of information sources, and user engagement in political discourse.

TWON also emphasizes societal engagement and knowledge transfer. Through formats such as Citizen Labs, researchers discuss project findings with citizens and stakeholders to reflect on the implications for democratic societies. The project ultimately aims to develop evidence-based recommendations for policymakers, platform providers, and civil society to support healthier digital public spheres.