Talking to AI: Metapragmatics and sociotechnical imaginaries
When people interact with “intelligent” agents, such as chatbots, they often engage in verbal provocations as a way to test the limits of the agent’s “intelligence” or apparent “humanness”. One function of such exchanges is to aid people in developing pragmatic parameters and inferential processes for conversing with machines, what I have previously called algorithmic pragmatics (Jones 2020). Recently, it has become popular to share these provocative exchanges over social media in the form of metapragmatic artifacts, such as screenshots of conversations with chatbots or videos of interactions with voice assistants. In this talk I will focus on three examples of such artifacts: 1) performances of conversations with Amazon's Alexa and Apple's Siri posted as short videos on YouTube and TikTok, 2) screenshots of people’s intimate conversations with the virtual companion Replika, and 3) outputs people have received after presenting provocative prompts to large language models. Based on my analysis I argue that these metapragmatic artefacts function to facilitate collective engagements with algorithmic pragmatics that feed into larger societal imaginaries, not just about artificial intelligence, but also about what it means to produce “appropriate” language and what it means to be human.
References
Jones, R.H. (2020). The rise of the pragmatic web: Implications for rethinking meaning and interaction. In C. Tagg and M. Evans (eds.) Message and medium: English language practices in new and old media. Amsterdam: De Gruyter Mouton. 17-37.