Making sense together: Human-AI communication through a Gricean lens

With the introduction of ChatGPT in 2022, Large Language Models (LLMs) have been increasingly used in classrooms to support teaching and learning. However, the nature of communication between students and LLMs remains under-examined. The current study explored this communication through the lens of Paul Grice’s cooperative principle, examining the extent to which students and ChatGPT adhered to Gricean maxims of quantity, quality, relation and manner in their one-on-one communication during class. The study was implemented in 2023 in a Year 10 English class in an all-boys independent high school in Western Australia. 10 students participated in a poetry-focused lesson, where they interacted with ChatGPT to explore poetic texts. The lesson was designed not only to support understanding of the poetry but also to foster students’ capacity for critical questioning and engagement with AI-generated output by positioning the GenAI as a dialogic partner. The findings identified specific user violations of Gricean maxims in students’ communication with ChatGPT and the impact such violations had on ChatGPT’s output. These findings affirm the relevance of Grice’s cooperative principle and maxims for analysing conversations with LLMs and the potential for identifying interaction patterns with GenAI that differ from human-to-human conversations, underscoring the importance of examining human input as well as LLMs’ output. In place of a recent trend towards technocentric approaches to researching human-GenAI communication, the paper advocates a sociotechnical approach as a means to examine such interactions holistically.