Exchange between science and practice at the INDI workshop

As part of the INDI project funded by the Federal Ministry of Research, Technology, and Space (BMFTR), scientists from the DIID and experts from the field gathered at the University House in Düsseldorf on October 6 and 7 to exchange ideas on current developments and experiences regarding the role of AI in online discussions. The two-day workshop served both as a forum for mutual exchange and as an opportunity to test the AI-based discussion assistants developed in the INDI project.

The first day focused on the transfer of experience. Representatives from DIID's partners Liquid Democracy e.V. and Make.org, as well as conversario, ZDF Digital Medienproduktion GmbH, and deliberAIde, presented their own projects in which AI is used to create a transparent, inclusive, and more constructive discussion process.

Liquid Democracy e.V. presented several exciting projects: First, the KOSMO project, realized in cooperation with DIID, in which an AI assistant was developed to support moderators in online discussions. The DeFaktS project showed how AI can help detect disinformation. 

Make.org Germany presented, among other things, the use of AI in its WahlCheck 2025 for the 2025 federal election, which clearly summarized the positions of the nine most important parties on various topics. In addition, users were able to ask the AI specific questions about the party programs, which generated an answer with corresponding evidence.

Conversario offered practical insight into the application of AI for community management and moderation. The tools presented help moderators keep track of extensive comment sections, keep dialogue spaces open, and at the same time integrate protective mechanisms against hate speech or destructive behavior.

This was followed by a presentation by the young start-up deliberAIde, which is developing an AI-based tool that records audio at face-to-face events and transcribes it at the same time. It anonymizes personal data, recognizes different languages, and summarizes the results in both text and graphic form. This makes the communication that takes place sustainable and usable. In collaboration with the Siegen-Wittgenstein district, the tool has already been tested in a participation process, with promising feedback from the district.

At the end of the first day, ZDF Digital Medienproduktion GmbH reported on an internal AI tool that was developed in connection with the Public Spaces Incubator project. The tool automatically checks comments for rule violations and is intended to contribute to a more respectful discussion climate. Feedback from the tests showed that users take the AI's comments seriously and in many cases adjust their posts accordingly.

The second day of the workshop was devoted entirely to user testing: participants were able to try out the discussion assistants developed in the INDI project for the first time. These assistants evaluate online comments along the dimensions of civility, reasoning, and empathy. In small groups, the participants tested the various functions, discussed possible application scenarios, developed suggestions for improvement, and reflected on ethical and normative aspects. It became clear that AI interventions in discussion forums raise not only technical but also social questions, such as fairness, diversity of opinion, and the responsibility of platforms for digital discourse.

In the concluding exchange, it became clear that the combination of scientific research and practical application provides valuable insights and impetus for responsible AI development. Participants emphasized the importance of transparency, user orientation, and interdisciplinary collaboration in developing trustworthy systems. 

We would like to thank all participants for their participation in the workshop and their constructive input. With the aim of making online debates more constructive and inclusive in the future, the INDI team will integrate the diverse ideas from the two days into the further development of its discussion assistance system.