Abstract

The fairness of AI algorithms is much debated. In this project, we use administrative data and ML methods to compare AI and human decisions in terms of fairness. On the one hand, we use profiling data from caseworkers, where unemployed people are classified at the beginning of unemployment according to their probability of becoming long-term unemployed. On the other hand, we use ML methods to predict this probability. Both predictions are compared with respect to fairness aspects, e.g. gender or nationality.