Representations of Hate Discourses: Interactions between Human Linguistic Pragmatics and AI Multimodal Pragmatics

Abstract

Considering the intertwining relationships between the Web 2.0 (O’Reilly 2005) and hate speech proliferation as an intrinsic peculiarity of contemporary societies (Balirano and Hughes 2020), in which online communities of belonging are created through users’ interaction and engagement (Zappavigna 2014), this paper compares human linguistic conception of hate speech with its corresponding AI multimodal design. More specifically, it analyses how the concept of hate speech is linguistically realised by university students according to the social categories of age, body, gender, disability, ethnicity, and religion, as defined within the research framework. The subsequent phase of analysis focuses on the ensuing AI-generated images to understand how linguistic input provided was interpreted and visually elaborated on. The two phases respectively apply Gricean principles of conversational cooperation (Grice 1975) and Kress and van Leeuwen’s multimodal theory (2020). Through this approach, this article explores the various articulations of hate speech as shaped by individual perceptions and student-machine interaction, thereby highlighting potential biases of both socio-cultural and computational nature. Given the pervasive role of AI programs in contemporary societies, discrepancies may shed light on how the machine conceives, reprocesses, and realises the intended concept in ways that diverge from human intention, thereby potentially stripping the user of the agency that fundamentally characterises human communication. In essence, it examines the extent to which AI technologies can bolster or contrast the spread of hate discourse.