On the Identification and Suppression of Hate Speech in Online Contexts

Abstract

This paper focuses the issue of identifying indirect hate speech on digital platforms. In previous studies, the authors have addressed automated flagging and suppression of hate speech in YouTube material. We found that such operations are characterized both by problems pertaining to the vagueness of the hate speech concept, as well as a problem compounding tendency of content creators towards circumventing identification and suppression efforts by mainly making use of indirect and tacit references. The indirect references chiefly function through various means of indicating levels of meaning above the immediate sentence, or the immediate signification, often by referring commonly held worldviews or ideological structures. This implies that automated suppression that relies on the flagging of keywords faces certain structural reliability issues in relation to these indirect communications, and important ethical and rights-related problems are also embedded here. For this reason, this paper will explore methods for the reliable identification of indirect hate speech. We will explore two methods: speech act theory and Grice's theory of incorporated cooperativity, and ascertain whether they separately or in combination can provide a framework for the reliable identification of indirect acts of hate speech. The paper also indirectly emphasizes the importance of worldviews, and the critical analysis of faith and worldviews, in regard to addressing contemporary political issues.