On Monday (3rd of May), we had an exciting inauguration event for the Dataninja research training group. It was attended by about 150 curious people, scientists, and researchers from Germany and around the world.

The Dataninja Research Training Group creates an important impetus to make AI decisions transparent and to develop new, robust systems in such a way that they can adapt flexibly to changing requirements.

— Prof. Dr. Barbara Hammer

For the event opening the minister for culture and science of North Rhine-Westphalia, Isabel Pfeiffer-Poensgen, held a speech in which she was pointing out the general view on current developments in the research of AI and the importance of research in this area. This was followed by remarks from the Rector of Bielefeld University, Gerhard Sagerer, who further emphasized the need for education and research in (trustworthy) Artificial Intelligence. The final words of the introduction had Barbara Hammer giving an overview of what Dataninja is all about, introducing our projects and also the involved Ph.D. students in a short video.

AI should be honest about what is knows and what it doesn’t know.

— Prof. Dr. Peter Flach

The inauguration speech was held by Peter Flach, current President of the European Association for Data Science. He was giving a very intriguing presentation focused on quantifying the uncertainty of AI systems and in which situation an algorithm should actually say “I don’t know“, instead of giving an uncertain answer, e.g. classifications close to a decision boundary. Flach then talked about his 2016 paper called Background Check​1​, where a method is presented to restrict a high decision certainty of a classifier to its comfort zone. Furthermore, the technique also allows to do outlier detection and express the confidence of a prediction.

The event ended in a final panel discussion on the question How far is it towards a trustworthy AI? It was great to get the diverse views of a couple of renowned researchers from industry and academia (Dr. Cédric Archambeau (Amazon Web Services, Berlin), Prof. Dr. Peter Flach (Univ. Bristol), Prof. Dr. Eyke Hüllermeier (LMU Munich), Prof. Dr. Kristian Kersting (TU Darmstadt, hessian.ai), Prof. Dr. Ute Schmid (University of Bamberg, bidt), Prof. Dr. Heiko Wersing (HRI Europe)). The event ended with a gathering with attendants — and it was nice to see so many friendly faces of scientists from all over Germany, showing already the interest in the program and providing a good start for further fruitful connections and exchange.


  1. 1.
    Perello-Nieto M, Menezes Silva Filho T, Kull M, Flach P. Background Check: A general technique to build more reliable and versatile classifiers. In: 2016 IEEE 16th International Conference on Data Mining (ICDM 2016). Institute of Electrical and Electronics Engineers (IEEE); 2017:1143-1148. doi:10.1109/ICDM.2016.0150