Alessandra Buccella, Institute for Interdisciplinary Brain and Behavioral Sciences, Chapman University

— Even though artificial intelligence has been connected to philosophy (and cognitive science) since its early days, the relatively “young” field of machine learning adds an intriguing new layer to the relationship among these disciplines. In this talk, I discuss two domains of machine learning research in which philosophical theories and concepts can help find innovative solutions to common problems. The first domain in which I believe philosophy of mind can provide machine learning developers with helpful insights is computer vision. In particular, I argue that ideas coming from the phenomenological tradition, such as Merleau-Ponty’s notion of “style”, can inspire the design of better machine learning algorithms for object recognition across a variety of contexts and modifications (e.g., changes in illumination, visual angle, distance, etc.). The second domain is that of so-called “Trustworthy AI”. Here, I argue that, in order for users without technical knowledge about machine learning to trust these systems, the latter must have the ability to construct and articulate functional explanations for their outputs that can play the (normative) role of reasons.