Using ChatGPT errors to encourage student exploration and self-learning in mathematics

• Philip Slobodsky ^[2] ; Mariana Durcheva ^[1]
  1. [1] Sami Shamoon College of Engineering

     Sami Shamoon College of Engineering

     Israel

     
  2. [2] Halomda Educational Software, Rishon LeZion, Israel
• Localización: International journal of mathematical education in science and technology, ISSN 0020-739X, Vol. 56, Nº. 8, 2025, págs. 1401-1425
• Idioma: inglés
• DOI: 10.1080/0020739X.2025.2505200
• Enlaces
  • Texto completo
• Resumen

  • AI-based bots (ChatGPT) are capable of solving mathematics problems, and students often use them for homework preparation, self-learning, etc. This raises a number of didactical and technical questions: How can students submit assignments containing complex mathematical expressions using only a keyboard? How should mathematics errors in ChatGPT's responses be addressed? Should teachers discourage students from using these tools due to potential misuse in homework submissions? Can students gain any benefits from using AI in their self-learning? Here, we share our ideas on how to address these issues using the tools offered by the Halomda e-assessment and e-learning platform, which provides connection to ChatGPT. Given that ChatGPT may make errors in mathematics operations, the system's help module advises students to verify answers using other tools such as graph plotter and algebraic calculator, and to use subject-specific Key Prompts. We discuss two surveys conducted with pre-service teachers at two colleges in Israel, where students share experiences using ChatGPT. The findings reveal that half of them do not verify solutions by ChatGPT; majority do not use ChatGPT for self-learning. However, among students who engage with ChatGPT, its use enhances their learning performance: designing prompts can foster engagement with the material and improve comprehension.