14.10.2025-27.01.2026: Exter­nal Speaker Series

14/10/2025

Partisan Leanings in Algorithmic Advice: An Audit of Political Advice from Large Language Models

Jan Zilinsky

Technische Universität München | Technical University of Munich

Large Language Model (LLM) chatbots are poised to become key intermediaries between citizens and their sources of political information. Yet, we know little about the accuracy of their political content (and advice) and the conditions under which these models exhibit political biases. As users increasingly turn to AI for information and guidance on topics ranging from shopping to politics, understanding the nature and quality of AI-generated output is critical. This paper introduces an approach for auditing LLMs and investigating their partisan leanings in the political advice they generate. The study systematically evaluates the voting recommendations of several leading models, including versions of Claude, Gemini, GPT, Grok, DeepSeek, and Llama, by prompting them with queries from distinct types of voters. Evidence from a large set of queries shows that while most models provide ideologically consistent and “correct” advice to clearly aligned users, their recommendations for cross-pressured voters are highly variable and often reveal distinct partisan biases.

18/11/2025

Government-opposition relations, media coverage, and citizens’ attitudes

Or Tuttnauer

MZES Mannheim

Previous research has repeatedly suggested that the actions of parties in parliament, and specifically the interactions between the government and the opposition, are at least partially aimed at attracting voters. Indeed, recent works show that government-opposition relations in parliament do indeed correlate with consequent citizens’ attitudes. Of course, ordinary citizens – and most political experts – do not observe parliamentary activity directly but rather through news outlets, social media (which mostly echoes mainstream media), or the parties themselves. However, early It is presumed that the media serves as the link between what goes on in parliament and the public perceptions of it, this linkage has not yet been tested, certainly not in a comparative manner.

In this ongoing research project, I aim to show, first, that parliamentary government-opposition relations are consistently reflected by their media coverage, and second, that the effect of parliamentary behaviour on citizens’ attitudes is mediated by the the tone and “volume” or salience of news regarding government-opposition interactions. I do so by linking parliamentary activity data from 12 democracies to news data from the Integrated Crisis Early Warning System (ICEWS) and survey data from the CSES and the ESS projects. The findings of this project will be of special interest to scholars in the fields of legislative studies, political communication, and public opinion.