The Black Box Problem in Artifical Intellingence: Philosophical Reflections on Opacity and Explainable Al
The widespread use of Machine Learning (ML) in domains as varied as medical diagnosis, product-recommendation, and autonomous driving gives rise to the so-called Black Box Problem in AI: the problem of explaining the behavior of 'opaque' computing systems. Although the Black Box Problem has become the subject of extensive technical, legal, and public debate, several philosophical questions remain unanswered. First, it remains unclear what the concept of 'opacity' actually means in this context, and how it relates to the concept of 'explanation'. Second, extant discussions have been unable to present a systematic answer to the question of why and for whom the Black Box Problem is actually problematic. Third, although many technical solutions to the Black Box Problem have already been proposed within the Explainable AI research program, it remains unclear against which explanatory norms the quality of these solutions should be evaluated. In this talk I will raise each one of these philosophical questions, and propose tentative answers.
Referent/Referentin
Dr. Carlos Zendik, Bereich Philosophie, Otto-von-Guericke Universität Magdeburg
Veranstalter
Institut für Philosophie
Termin
29. Oktober 201916:15 Uhr - 18:00 Uhr
Ort
Geb.: 1146Raum: B313
Im Moore 21
30167 Hannover