Limits to ChatGPT’s Pragmatics in the domains of Commonsense and Culture.

Abstract

Does ChatGPT deliver its explicit claim to be culturally sensitive and its implicit claimto be a friendly digital person? These claims are investigated from the perspective oflinguistic pragmatics, particularly Grice's cooperative principle in communication.Following the pattern of real-life communication, turn-taking conversations reveallimitations in the LLM's grasp of the entire contextual setting described in the prompt.The prompts included ethical issues, a hiking adventure, geographical orientation andbodily movement. For cultural sensitivity the prompts came from a Pakistani Muslim inEnglish language, from a Hindu in English, and from a Chinese in Chinese language.The issues were deeply cultural issues involving feelings and affects. Qualitativeanalysis of the conversation pragmatics showed that ChatGPT is often unable toconduct conversations according to the pragmatic principles of quantity, reliablequality, remaining in focus, and being clear in expression. We conclude that ChatGPTshould not be presented as a global LLM but be subdivided into several culture-specificmodules.

Ähnliche Arbeiten