Improving AI-mediated Hate Speech Detection: A Genuine Ethical Dilemma

Abstract

AI-mediated hate speech detection is indispensable for contemporary communication platforms. But it has known deficiencies in terms of bias and context-awareness. I argue that improving on these known deficiencies leads into a genuine ethical dilemma: It will increase the epistemic and social utility of these platforms, while also helping bad faith political and corporate actors to suppress unwelcome speech more swiftly and efficiently.