On Friday 24 January 2020 the CPDP (Conference on Privacy and Data Protection) I had the pleasure of moderating a panel on Regulatory requirements for algorithms used for governmental decisions.
The use of algorithms by government goes way back, e.g. in the 1990s sentencing systems were developed as well as decision systems for unemployment benefits. Most systems were not Artificial Intelligence (AI), but straightforward executing decision trees. Since 2013 through advances in machine learning, Artificial Intelligence has entered practice on a wide scale, including government. In this panel we look at (self-)learning algorithms, and what legal constraints should apply to governmental decisions based on use of these algorithms. Questions addressed are:
- What algorithms we do allow and what not, and in case we allow algorithms, under what conditions?
- What conditions should be met when government uses algorithms developed by private parties?
- Under what circumstances can the human in the loop be meaningful?
Panelists were:
- Jennifer Cobbe, Cambridge University (UK),
- Marlies van Eck, eLaw Institute Leiden University (NL)
- Heleen Janssen, Ministry of the Interior and Kingdom Relations, Cluster New Technologies, Public Values and fundamental Rights (NL)
- Tom Korman, European Parliament (EU)
There is a video of the panel.
For videos of previous panels see e.g.
Obscured by algorithms: how to prove “yes” if computer says “no” (2016)
‘Purpose limitation is dead, get over it’. On the undermining of data protection law (2019)
