📍 Hybrid Event: MZES, Mannheim + Zoom
📆 October 26, 2022
Data-driven approaches for the allocation of public resources promise to make fast, reliable, cost-efficient and objective decisions. However, there are also concerns about such approaches. For example, data-driven algorithmic profiling in the context of the allocation of labor market support programs led to public outrage in Austria. Fairness concerns were raised, as gender and citizenship were found to influence allocation decisions. Thereby, they bear the risk of disparate treatment. In this workshop, we will provide an introduction to fairness notions in machine learning and discuss the possibilities and limitations of technical approaches. A data-driven profiling system for allocating support to jobseekers will be implemented in Python and provided as executable code snippets. Our aim is to discuss and evaluate fairness metrics in a realistic example. Prior knowledge of Python is not necessary and the libraries used are also available in R.
📝 Slides
👤 Eva Achterhold is a master's student at the Chair for Statistics and Data Science in Social Sciences and the Humanities at LMU Munich. Her research interests include the study of the socio-cultural impact of AI, especially with regard to discrimination and transparency, and the application of methods to mitigate negative consequences. She is currently working on the topic of fairness in algorithmic decision making in the context of allocating support programs to unemployed individuals.