Ethics and Epistemology of Machine Learning
The Ethics and Epistemology of Machine Learning (EEML) research group is part of the Cluster of Excellence Machine Learning for Science at the Universty of Tübingen. We are a mixed-background team of researchers working at the intersections of machine learning, philosophy of science, formal learning theory, causality, and law.
Featuring
Konstantin Genin / Sebastian Zezulka / Raysa Benatti / Mykhailo Bogachov
It is widely acknowledged that questions of scientific methodology depend on ethical ones. If an experiment is unethical, it ought not to be performed. If an algorithm is unfair, it ought not to be implemented. From this perspective, ethics responds to methodological advances by rushing to install new guard-rails. But ethical questions also depend on methodological ones. Whether an experiment is ethical depends on whether similarly reliable inferences could be made from non-experimental data. Whether an algorithm is fair depends on how well it manages delicate tradeoffs between competing explications of fairness. The answers to these questions typically turn on methodological ones and -- more often than not -- these are both highly technical and hotly contested. From this perspective, methodological advances lead inevitably to ethical ones. The goal of the "Ethics and Epistemology" research group is to work these problems from both sides: to approach methodological issues with an eye to their social consequences and to approach ethical issues with an eye to methodological solutions.
