On Friday, June 27, 2025, the DIID hosted its annual summer retreat at Schloss Mickeln in Düsseldorf. Around 30 members and guests exchanged views on current developments in science and practice. Even before the official start at 1 p.m., the Executive Board met for a regular meeting.
After all participants had arrived and enjoyed a brief snack, the program began with a report by DIID spokesperson Prof. Dr. Marc Ziegele and DIID coordinator Dr. Dennis Frieß on current DIID activities and developments. In the first thematic block “(Potential) New Members”, Prof. Dr. Katrin Weller (GESIS) introduced herself with a presentation on research data management and the challenges of handling social media data in research practice. The focus was on current projects for the systematic collection of volatile data (“data decay”) and the protection of researchers when acquiring data. In the second presentation, Leonard Tröder, Managing Director of Birds on Mars GmbH, gave an insight into the responsible use of AI in practice. He emphasized the needs-based implementation of AI applications in companies and the importance of “AI fluency”. The “Responsible AI Framework” he presented is intended to promote a reflective approach to AI and support the examination of one's own requirements for these technologies through a systematic assessment.
After a short coffee break, the second block followed with the topic “AI and disinformation”. Dr. Simon Kruschinski (JGU Mainz) presented initial empirical findings from the 2025 federal election on the use, acceptance and impact of generative AI in political campaigns. In general, parties very rarely used AI-generated content for social media communication in the 2025 federal election. Nevertheless, there are differences between the parties: While established parties used AI rather cautiously, the AfD used AI-generated content much more aggressively. Experiments on the impact of this content showed little influence on its evaluation. However, negative messages in the campaigns intensified feelings of anger and fear and worsened the image of parties overall. Carolin Klingsporn (Liquid Democracy) and Isabel Bezzaoui (FZI) then reported on “DeFaktS”, a project to detect disinformation in online discussions using Explainable AI (XAI). Several user tests demonstrated the importance of transparent and comprehensible AI in order to increase trustworthiness on the part of users.
After a short coffee and cake break, the focus was on two DIID projects. First, Dr. Nadja Wilker presented the “ERLE” project, which records the range and framework conditions of eParticipation throughout Germany and identifies good practices. As part of the presentation, some preliminary results from the ongoing field phase were presented. Finally, Prof. Dr. Marc Ziegele presented the results of the “UPEKI” project. In this project, two AI systems were developed to improve the user experience of users in online discussions. Overall, all AI applications were rated optimistically by the study participants; in particular, the so-called “stance detection”, i.e. the identification of viewpoints in posts, showed a high level of effectiveness by increasing the number of comments.
The retreat ended with a barbecue at the summer party. We would like to thank all the speakers for their presentations and everyone who attended for sharing their experiences. We are already looking forward to the next retreat on 28.11.2025!