BETWEEN MONOLINGUALISM AND MULTILINGUAL ACCOMMODATION:

Abstract

Within multilingual speech communities, the use of more than one code in a single interaction is a completely predictable behavior. Human beings tend to fully use their linguistic force in daily interactions. As artificial intelligence (AI) has settled among the community, serving people for the smallest tasks, the inquiry on whether such tools can handle code switching (CS) as naturally as humans do raises. The current study analyses human-AI interactions all containing instances of CS through a sociolinguistic lens. The intended goal, revealing whether AI tools such as ChatGPT are primarily monolingual tools, gravitating towards English as a dominant language or if they do accommodate to the use of CS when the users trigger a mixture of linguistic varieties in a conversation. Moreover, the research seeks to identify any noticeable patters used by the AI tool when it comes to CS. To achieve this end, a number of corpora featuring human-AI CS interactions were meticulously analyzed. The used corpora included human-AI interactions involving prompts including switches between English, French and Arabic. The results of the study show that ChatGPT switches to another code, attempting to match the patterns used by the human party of the interaction but only to a certain extent. The AI tool may preserve a monolingual approach by reducing the multilingual switches and offering translations. This proves that the tool is primarily monolingual but would occasionally CS solely if the switch was triggered by the interlocutor.