Artificial intelligence in education: Comparing the responses of different large language models

Matej Urbančič

Summary:  Currently, there are several advanced large language models (LLMs) freely available for testing, and their use is steadily increasing. This paper aims to compare the results produced by some selected models for use in an educational setting. A qualitative research design was employed to identify the structure of the outputs and to analyse the information and key ideas related to questions about the purpose of education. The findings raise concerns about the reliability and relevance of the results, as they are neither equally informative nor consistent across different LLMs, with variations occurring even when the same questions are repeatedly tested. At present, there is no consensus on the optimal approach to integrating AI into education, nor on the potential impact of AI on learning, teaching, work and society. While it appears that the risks associated with AI can be managed, training in the use of LLMs is crucial at present, as these models will significantly impact various educational domains.