In the 21st episode of the "All Day Research" podcast, we welcome Dr.-Ing. Philipp Terhörst. Together with our host Marvin Beckmann, they talk about social bias in biometric face recognition.
Since the winter semester 2022/23, Dr.-Ing. Philipp Terhörst is the new junior research group leader in the area of "Responsible Artificial Intelligence for Biometrics" at the Department of Computer Science at Paderborn University. The topic of facial recognition, on which Dr. Terhörst has focused, is part of the research area of biometrics, in which biological data are measured and analyzed. This generally involves attempts to "verify" and "identify" identities with the help of machine learning.
In his research, Dr. Terhörst has specialized primarily in four areas, fairness, privacy, reliability, and explainability. For example, similar to algorithms in terms of "fairness," there is also a certain error rate in biometric facial recognition that can discriminate against certain groups of people through racist and sexist motives. "The consequences of such errors can be very far-reaching. I don't know if we will ever have perfectly fair algorithms in which we can weed out all errors. But we can at least further reduce the error rates even in biometric face recognition and thus counteract social bias," Dr. Terhörst said.
The implementation of the podcast was done in collaboration with the Department of Computer Science and the Student Council for Mathematics/Computer Science. The aim is to give more insights into science and thereby bring students - especially those in the first semester - closer to the study, to arouse interest and to show how much fun research and the study of computer science at Paderborn University can be.
Listen now to the twentieth episode of the "All Day Research" podcast. Find it everywhere podcast is available or here on the institute page.
Do you like the All Day Research podcast? Then feel free to leave a review.
Missed the last episodes? Don't worry, all episodes can also be found on the institute page.