Examining the potential and pitfalls of AI in problem solving

The integration of artificial intelligence (AI) into mathematical problem-solving has shown significant potential to enhance student learning and performance. However, while AI tools offer numerous benefits, they are prone to occasional conceptual and arithmetic errors that can mislead users and obscure understanding. This research examines such errors to improve the role of AI in solving mathematical problems. The study particularly assesses the abilities of AI tools—ChatGPT-4, Gemini, and CoPilot—in addressing proportional reasoning errors commonly made by students. ChatGPT-4 achieved the highest accuracy rate among the tested tools, correctly answering 10 out of 14 questions. Additionally, ChatGPT-4 provided more detailed explanations in its responses, with a higher word count compared to the other tools. However, all tools replicated certain errors commonly made by students on specific reasoning questions. In conclusion, while AI tools hold promise for enhancing mathematics education, they still have limitations. Improving AI’s contextual understanding and problem-solving adaptability could lead to the development of more robust educational tools.