Making artificial intelligence more transparent

The project, "AI Forensics: Accountability through Interpretability in Visual AI Systems," focuses on the use of image systems. "The field of facial recognition in particular continues to grow. For example, in the U.S., AI with facial recognition is being used to detect potential threats. What serves general security is at the same time an intrusion into personal rights. We want to make AI decisions comprehensible," explains project participant Prof. Dr. Claude Draude from the Scientific Center for Information Technology Design at the University of Kassel.

The goal is an Internet platform that is freely accessible to the public and makes large data sets accessible. Machine learning models should thus be made analyzable and understandable. "Users can upload an image, for example, and activate an automatic examination. The system checks whether the image is already in a dataset (dataset forensics), how it was used in a model (model forensics), and where and for what purpose (application forensics). The scope of these capabilities varies from model to model, illustrating different degrees of transparency and interpretability across the spectrum of models available for research," Draude explains.

Artificial Intelligence Initiative of the Volkswagen Foundation

The project is funded as part of the Volkswagen Foundation's "Artificial Intelligence" initiative. It will run for three years and is being supported with 1.5 million euros. In addition to the University of Kassel, the Institute for Artificial Intelligence and Media Philosophy at the Staatliche Hochschule für Gestaltung in Karlsruhe, the Department of Computer Science at Durham University in the UK, and the Department of Germanic and Slavic Studies at the University of California in Santa Barbara (USA) are involved. Technical partners are the Steinbuch Centre for Computing at the Karlsruhe Institute of Technology (KIT) and the NVIDIA CUDA Research Center at Durham University.

With its initiative, the Volkswagen Foundation is strengthening cross-disciplinary and cross-national research on the responsible further development of artificial intelligence (AI) systems. It is now funding seven project consortia from the social and technical sciences with a total of 9.8 million euros within this framework.

Contact:
Prof. Dr. Claude Draude
University of Kassel
Department 16 - Electrical Engineering/Computer Science
Gender/Diversity in Computer Science Systems
Phone +49 561 804-6028
E-mail claude.draude[at]uni-kassel[dot]de

Sebastian Mense
University of Kassel
Communications, Press and Public Relations
Phone 0561 804-1961
E-mail: Press[at]uni-kassel[dot]de