What do machine learning algorithms see when they look at us? Some concerns about the transparency and discriminatory effects of profiling based on machine learning.

In this presentation I will show how machine learning, when it is used to make sense of human behavior and characteristics (‘profiling’), can lead to infringements in terms of privacy, data protection and antidiscrimination law. One major concern from the perspective from data protection law is the question how to create useful transparency about the functioning of machine learning algorithms. I illustrate some of the issues related to transparency with recent work I have done in the USEMP project (http://www.usemp-project.eu/). Another important concern is how to distinguish which machine learning categorizations should be considered ‘good’ and legitimate differentiations, and which ‘bad’ discriminations (in the sense that they are either illegitimate, or at least undesirable from an ethical perspective). Looking at current privacy and antidiscrimination law, I argue that the existing legal framework might need to be extended.  In discussing the conundrums of transparency and differentiation/discrimination in relation to machine learning algorithms, I’ll pay some specific attention to the implications of the new General Data Protection Regulation.

The slides from the presentation can be downloaded here: Gradual equality – itu 26 MAY 2016_v1.3

Sign up for the talk is not required. 


Katja de Vries is a legal researcher and philosopher of technology affiliated to the Institute for Computing and Information Sciences (iCIS) at the Radboud Universiteit Nijmegen (the Netherlands) and the Centre for Law, Science, Technology, and Society (LSTS, Vrije Universiteit Brussel, Belgium). Currently she is working on the USEMP (http://www.usemp-project.eu/) project which will result in a transparency tool that shows users of social networks which (commercially interesting) information can be derived from their data (http://databait.eu ).  In a few months from now Katja de Vries will defend her PhD thesis (‘Machine learning/Informational fundamental rights. Reconciling two Baroque practices of making sameness with a  overnmentality of proportionality’).  Her PhD research looks at how machine learning, when it is used to make sense of human behavior and characteristics, can lead to infringements in terms of privacy, data protection and antidiscrimination law.  De Vries has been a member of the European “Living in Surveillance Societies”-network, and has worked on the FIDIS (Future of Identity in the Information Society) and SIAM (Security Impact Assessment Measure – A decision support system for security technology investments) projects. She publishes on a wide range of legal and philosophical topics and has co-edited ‘Privacy, Due Process and the Computational Turn’ (Routledge, 2013). De Vries studied at Sciences Po in Paris, obtained three masters degrees with distinction at Leiden University (Civil Law, Cognitive Psychology and Philosophy) and graduated at Oxford University (Magister Juris).

Time: May 26 2016, 12:00-14:00
Auditorium 3
Join the Facebook event here.
The event will be in English. 

ITU address:
IT-University of Copenhagen
Rued Langgaardsvej 7
DK-2300 Copenhagen S