Answerability, accountability, and the demands of responsibility

Knowing who or what should be held morally responsible when something goes wrong (or right) is an important part of human social relations. Some authors have suggested that certain AI-based systems pose a threat to our responsibility practices, which might lead to a ‘responsibility gap’. Such ‘gaps’ occur when we have no fitting candidate who can be held responsible for some event that was caused by an AI-system. If such gaps do in fact exist, then it might make sense to have an outright ban on such systems. Conversely, if these gaps do not exist, then perhaps there is nothing to worry about. In this paper I do not aim to resolve this debate. Rather, I wish to make a modest contribution to this literature. I will argue that in two specific senses of responsibility, namely, answerability and accountability, there might be no such responsibility gaps. While AI-systems might make it harder to know who we should hold responsible, they do not make such ascriptions impossible. Responsibility gaps then, on my view, are simply epistemic, and thus do not call for special attention any more than our usual practices of holding one another morally responsible.