May Different AI Tools Address Misconceptions in Physics?

Abstract

This study aims to comparatively examine the potential of four widely used free AI chatbots (ChatGPT, DeepSeek, Gemini, Grok) to spread misconceptions about high school-level physics. Free versions of ChatGPT, DeepSeek, Gemini, and Grok's chatbots were selected as AI tools. As fundamental physics concepts, the concepts of mass, weight, heat, temperature, and capacitors, which are the most common misconceptions among students, were selected. For this purpose, the responses of the AI tools to the questions posed to them were compared in terms of scientific accuracy. The responses of various AI tools are presented, along with links and screenshots. The responses were analyzed by three researchers who are experts in physics education. The results show that although AIs provide scientifically accurate answers to some questions, all AI systems contain statements consistent with the known misconceptions held by students. Particularly on topics with differing scientific interpretations, such as mass and weight, chatbots have been observed to generally provide definitive answers from a single perspective, disregarding different scientific viewpoints.