How well can AI detect fakes?

Ahead of the upcoming election campaign, researchers will provide insights into AI-assisted fact-checking and their projects involving media organizations and citizens.
According to disinformation experts, facts had little influence on the last U.S. election campaign. This was partly due to the overwhelming amount of circulating false information, which significantly eroded trust in both established media and democratic institutions. While Germany’s situation is different, the detection of disinformation is becoming increasingly urgent as the Bundestag election campaign approaches—especially since the creation and dissemination of disinformation has become much more efficient with the help of artificial intelligence (AI).
However, AI can also assist in detecting disinformation - and everyone can take part in this effort! Several research groups at TU Berlin are working on developing AI systems that can detect misinformation and disinformation faster and more effectively. They collaborate with both major media organizations and citizens who want to promote fact-based dialogue - regardless of their journalistic background or location.

At the event researchers will offer a behind-the-scenes look at three selected projects and participate in a panel discussion. Key questions include:

•    How effective is AI at detecting fake texts, images, speech, and videos?
•    Why is collaboration with human fact-checkers always necessary, and how can this work most effectively?
•    What implications does this have for science, media, and politics?
   

    Date: Tuesday, December 10, 2024, at 5:00 PM (doors open at 4:30 PM)
    Location: Audimax (H 0105), TU Berlin Main Building, Straße des 17. Juni 135, 10623 Berlin, Ground Floor
    Event language: German
    Registration: The event is free and open to everybody. Please register here.

Program

Introduction and Welcome
Presentation of the “Strengthening Democracy” initiative at TU Berlin by Prof. Dr.-Ing. Sebastian Möller, Quality & Usability Lab, TU Berlin

Presentation of Three Selected Projects

Student Project: “Strengthening Democracy with Chatbots”
A joint seminar by the Quality & Usability Lab at TU Berlin and the Mobile Dialog Systems Department at Otto von Guericke University Magdeburg. Building on established large language models (like ChatGPT), six chatbots were developed to simulate conversations with science deniers—for example, individuals denying the results of evolutionary research or the effects of vaccinations. One chatbot will be put to the test, with false information countered live.
Speakers: Dr.-Ing. Stefan Hillmann (TU Berlin), Jun.-Prof. Dr. Ingo Siegert (Otto von Guericke University Magdeburg), David Karic (TU Berlin student)

Project: “noFake”
A collaboration between TU Berlin, Ruhr University Bochum, and TU Dortmund with the investigative platform CORRECTIV. This AI-powered assistant tool enables citizens to uncover false information themselves. It is made available to the fact-checking community through CORRECTIV.Faktenforum. Additionally, the tool “Checknewsin1.click” automatically identifies propaganda and manipulation in social media. Prof. Dr.-Ing. Dorothea Kolossa (TU Berlin) will report on the system's 99% accuracy in identifying AI-generated images and discuss the surprising fact that AI-generated texts are more varied than those created by humans.

Project: “news-polygraph”
Germany’s largest research project on AI-based disinformation detection, involving TU Berlin, the German Research Center for Artificial Intelligence (DFKI), and partners like Deutsche Welle and Rundfunk Berlin-Brandenburg. The project focuses on developing AI methods to detect “deepfakes” in images, videos, and audio data as well as perform text content analysis. Dr. Vera Schmitt (TU Berlin) will explain how AI can support journalists in analyzing disinformation. A key focus will be the concept of “transparency,” which is essential for deploying AI systems in sensitive areas such as disinformation detection.

Audience Q&A on the Projects

Panel Discussion: “Democracy and AI—An Open Relationship?”
Topics include:

Panelists:

    Prof. Dr.-Ing. Dorothea Kolossa
    Prof. Dr.-Ing. Sebastian Möller
    Dr. Vera Schmitt
    Prof. Dr. Sabine Ammon (Philosophy and Ethics of Design and Technology, TU Berlin)
    Moderator: Wolfgang Richter, Communications, Events, and Alumni Office, TU Berlin

Audience Q&A