Stefan Schmid: Communication Bias and Political Alignment in Large Language Models: An Empirical and a Regulatory Perspective

Large language models (LLMs) are increasingly shaping citizens’ information ecosystems. Products incorporating LLMs such as chatbots and AI Companions are increasingly used for decision support and information retrieval, including in sensitive domains, raising concerns about hidden biases and growing potential to shape individual decisions and public opinion. In this talk, we will discuss: What is new about the digital dissemination of political information when we talk about AI? Which types of biases exist and what are the dangers? And to what extent can these dangers be mitigated through regulation -- and what steps would need to be implemented, and by whom?

Stefan Schmid will also report on recent empirical studies conducted at Weizenbaum Institute (published at AIES 2025 with Jan Batzner, Volker Stocker, Stefan Schmid, and Gjergji Kasneci) and on their recent work with Weizenbaum Fellow Adrian Künzler (published in the Communications of the ACM 2026)

Stefan Schmid is a Professor at the Technical University of Berlin and Principal Investigator at Research Group: Digital Economy, Internet, Ecosystems and Internet Policy

The Colloquium will be moderated by Manfred Hauswirth (Director and Principal Investigator at Research Group: Digitalization and Opening Up Science)