Machine Learning: What it Can, Cannot, and Should Not Do

While the statistical models for machine learning have mostly been developed in the 20^th century, with the wide availability of cheap computing and virtually endless data space, data mining and machine learning have become an accessible and mighty tool for researchers outside highly specialized groups. Beyond research, they have become a ubiquitous part of our daily life, as predictive text suggestions on our emails or face recognition to unlock our phones. This presentation will provide an overview of history, methods, and increasing usage of machine learning, from high-end research over daily life improvements to preclinical and clinical opportunities for pain research. A special focus will be set on common misconceptions, inherent bias, and the dangers of (mis)use of computational decision making.