Award for strong science communication at TU Dortmund University

Success for TU Dortmund University's science communication: the university has come second in the idw prize for the best science press releases of 2025. The award went to a press release on the error-proneness of AI voting aids - a topic of high social relevance in which researchers from the Department of Computer Science were significantly involved.

The award-winning press release entitled "Before the 2025 Bundestag election: TU Dortmund University team investigates the susceptibility to errors of AI voting aids" reports on a study by researchers from TU Dortmund University and the Research Center Trustworthy Data Science and Security of the University Alliance Ruhr. Shortly before the early federal elections in 2025, they analyzed two new AI election assistance tools - with a clear result: Anyone who asks these systems about the positions of political parties often receives incorrect answers.

The Informationsdienst Wissenschaft (idw) honors particularly successful science communication with its annual award. From around 20,000 press releases from more than 1,000 scientific institutions, a jury selects articles that stand out for their journalistic quality, high news value and convincing presentation of complex topics. A total of 98 press offices took part in this year's competition.

For jury member and NDR information journalist Korinna Hennig, the award-winning text was particularly convincing due to its clear structure and comprehensible language. The fact that the press release makes the social relevance of the research clear in the very first sentences and clearly explains complex relationships was particularly emphasized. The additional background information and examples also facilitated further journalistic processing.

The award also underlines the importance of good science communication at TU Dortmund University. In the field of computer science in particular, research results are produced that provide an important basis for social debates - for example on artificial intelligence and its role in democratic processes. Communicating them in an understandable and transparent way is a central component of scientific work.

The fundamental problem with AI assistants is that the answers depend not only on the parties’ platforms but are also heavily influenced by user input,” says Prof. Emmanuel Müller. One advantage of the AI approach is that it allows for a deeper exploration of issues than the schematic approach used by Wahl-O-Mat. According to the researchers, this offers significant potential for transparency in election platforms and party statements. “But due to the contradictory answers, such applications do not yet seem sufficiently reliable to us. AI applications require certification and validation,” demands Emmanuel Müller. Research in this area not only highlights the weaknesses of current AI systems but also develops secure and trustworthy AI systems.

The topic of AI security is being intensively researched at TU Dortmund University. Development is taking place at several chairs within the School of Computer Science. Most recently, the team led by Prof. Daniel Neider, in a collaborative project (DFG FOR 5359: AI-FOR Deep Learning on Sparse Chemical Process Data), secured safety-critical AI systems with provable guarantees.

Translated with DeepL.com (free version)