The recent rapid development in machine learning (ML) has opened up unprecedented possibilities in analyzing and predicting diverse phenomena. From the humanities, to the social and cognitive sciences, to the natural sciences, fields previously closed off to predictive analysis are now the subject of machine learning investigations. This workshop aims to facilitate discussions among experts from different fields as to the philosophical issues, ethical concerns, and technical benefits of the use of ML in the various sciences. Do ML methods really differ from traditional statistical methods, and, if so, are they complementary? What is the epistemic status of the conclusions drawn using ML methods, compared with other more familiar types of learning, such as experimentation? Finally, are today’s machine learning methods amenable to causal inference, and (if not) what needs to be done to advance these methods into the causal realm?
Philosophical interest in ML is driven by its remarkable intermingling of disparate theoretical frameworks. Learning is said to be accomplished when an information theoretic quantity is minimized, an empirical distribution is matched, or a Bayesian posterior is computed. One central goal of this workshop is to facilitate an interdisciplinary dialogue on machine learning. We aim to bring together philosophers of science, statisticians, and data science practitioners to share insights from different communities. The purpose is to introduce workshop participants to a diverse collection of perspectives and methodologies in the hope of engendering further interdisciplinary thinking.