From Canberra to Tübingen: Robert C. Williamson accepts professorship for "Foundations of Machine Learning Systems

Prof. Dr. Robert C. Williamson is one of the pioneers of the field of machine learning, and his work and commitment have had a tremendous impact. Having worked at the National University, Canberra, Australia, for most of his life, he has now accepted the offer for a professorship in Tübingen and will move here soon.

In spring 2021, he will start his position as Professor for the “Foundations of Machine Learning Systems” at the Eberhard Karls University Tübingen. The W3 professorship has been set up in the context of our Cluster of Excellence "Machine Learning: New Perspectives for Science".

“We are absolutely excited and really proud that Bob Williamson is moving to Tübingen”, says Ulrike von Luxburg, one of the speakers of the Cluster. “He is one of the most sharp-minded discussion partners I’ve ever met. His way of asking innocent-looking questions has shaped my own way of approaching science”.

In his research Williamson focuses on understanding and designing machine learning systems as a whole. “The vast majority of research in the field of machine learning comprises the development of new algorithms for solving particular technical problems”, he explains. “I am interested in how to think about the behaviour of such systems at a higher level.”

Data quality and fairness
He also examines the challenging problems arising from the fact that machine learning is being embedded into socio-technical systems. “These problems are relatively poorly studied and understood at present”, Williamson says. Their social impact, however, can be high. He is therefore concerned with data quality, for example, and with ethical topics such as fairness.

In machine learning the use of “convenience” data – data that has possibly been collected for other purposes – is widespread. “This data is rarely perfect”, Williamson says.  At present there are still only limited theories to guide the use of such imperfect data. “Such theories seem essential to the more widespread embedding of machine learning technologies in existing socio-technical systems”, he declares.

Since machine learning systems are increasingly being used to make consequential decisions regarding people (for example in the allocation of credits or the selection of university students), this opens opportunities for ethical harm. Algorithmic decisions can be unfair, privileging one group over others. “My interest includes the development of fundamental limits to fairness”, Williamson says. He has already shown fundamental trade-offs between fairness and utility in a particular setup. He believes that it will be possible to use the machinery of information theory to develop more sophisticated results that will provide guidance as to the limits of the possible.

From Canberra to Tübingen
Robert Williamson already has strong connections to the Tübingen machine learning community. Not only has he been a member of the Scientific Advisory Board of our Cluster of Excellence and is thus familiar with its work and progress, but he has also known many of the researchers in Tübingen and Stuttgart for a long time and has co-authored papers with some of them. “Tübingen is, in my view, the best place in Europe, and in fact I think now the world, for the sort of machine learning research I wish to do”, he says. “The collection and breadth of talent there is amazing.”

For further information please visit "Foundations of Machine Learning"