The increased use of algorithmic decision making systems (ADM-systems) in various areas of politics and society raises many intricate questions, such as
- “How are ADM-systems implemented into democratic governance?”
- “How and to what extent do they affect everyday decision-making?”, or
- “How and to what extend does society influence the implementation processes?”
Funded by the German Federal Ministry of Education and Research (BMBF), our team, consisting of researchers from the fields of informatics, philosophy and political science, addressed those questions from the different angles of our expertise during the last 1 ½ years in the project “FairAndGood ADM”. The results we obtained so far, were presented in a scientific workshop hosted by us – partly in presence, partly online. Over one and a half days, taking place on the 21st and 22nd of September, 2020 the team discussed their research findings gathered with participants from University of Düsseldorf, Saarland University and University of Twente as well as colleagues working on similar subjects from the TUK. The workshop featured four paper presentations – three by the political science team and an “ethical reflection paper” by the philosophy group. In their papers, the political science team (Kathrin Hartmann and Georg Wenzelburger) present a “deep dive” into three empirical cases that teach us how ADM systems are implemented in real life of public administration and what political processes led to the decision to introduce the systems. In the first case study, we analyzed the implementation of the ADM system ‘COMPAS’, which is used in the criminal justice system of the U.S to assess a criminal defendant’s likelihood to re-offend.
In our second case study we concentrated our research on the implementation of ‘Admission Post-Bac’ (APB), a tool that was designed to assign high school graduates to university programs. The introduction of this algorithmic matching tool was a very complicated process (as can be glanced from the figure), which we tried to theoretically explain using the Multiple Streams Framework.
Finally, the third case study focuses on ‘AMAS’, and ADM-system which shall be implemented in the Austrian unemployment service to assess the reintegration chances of a job-seeker into the labou
r market. Not only studying already available data and documents but also going into the field, conducting semi structured interviews using qualitative empirical methods, we were able to gain good insights into the political decision making and the political regulation surrounding the implementation of ADMs.
The results of the workshop were very helpful to us to further polish the papers. As each research paper was discussed by one of our participants, looking at the strengths of the paper but also discussing weaknesses to overcome, we had insightful and thorough discussions about the contributions. Besides papers, we also discussed further steps we have to take in order to provide solid information about the implementation of ADM systems for scientific researchers and what we, as researchers and citizens, can learn from the cases.