<![CDATA[Artificial Intelligence (AI) in education not only offers significant potential for facilitating access to information but also presents challenges to students’ independence and thinking processes. Problem-solving is a fundamental cognitive skill that requires abstraction and algorithmic thinking. Preliminary analysis revealed that 22 out of 25 students experienced difficulties in identifying relevant information and designing algorithms to solve realistic mathematical problems. These students made errors in simplifying problems and formulating strategic solution steps despite using AI. This qualitative study aims to describe the errors made by students in the processes of abstraction and algorithmic thinking. In addition, students’ solutions were compared with AI-generated outputs to identify their error characteristics. The study involved three students and two AI models (ChatGPT and Gemini) as subjects. Data were collected through realistic problem-solving tests and interviews. The findings indicate that students remain vulnerable to both conceptual and procedural errors, whereas AI predominantly exhibits conceptual errors in the abstraction process. These findings highlight the need to enhance students’ abstraction and algorithmic thinking skills through more effective instructional approaches. Although AI has the potential to support simulation and analysis, it cannot fully replace students’ thinking abilities; therefore, students’ confidence in solving problems should also be strengthened]]>
Copyrights © 2026
