Journal of Computational Analysis and Applications (JoCAAA)
2024
The current study explored the effect of integrating ChatGPT into mathematics instruction on the development of critical thinking skills among sixth-grade students in the Arab society in Israel.As artificial intelligence tools increasingly influence education, understanding their pedagogical value for young learners has become essential.The research employed a controlled experimental design using a pretest/posttest format with two groups: an experimental group (28 students) who engaged in guided learning activities incorporating ChatGPT while solving mathematical word problems, and a control group (31 students) who studied the same content using traditional instructional methods.Both groups included male and female students with similar socioeconomic and academic backgrounds.Critical thinking skills were assessed through two complementary instruments: (1) a culturally adapted Likert-scale questionnaire based on the framework developed by Rodríguez Rojas et al. (2024), and (2) eight open-ended questions designed according to Facione's (2015) model of critical thinking, focusing on analysis, inference, interpretation, and explanation.Quantitative data were analyzed using paired and independent t-tests to determine both within-group and between-group differences.Qualitative data were analyzed through content analysis to identify patterns of reasoning and argumentation in students' written responses.The quantitative findings demonstrated a statistically significant improvement in the experimental group's mean critical-thinking scores between the pre-and post-tests (p = 0.012), whereas no significant change was found in the control group (p = 0.432).The comparison of change between groups indicated a moderate advantage for the experimental group, approaching statistical significance (p ≈ 0.07).Qualitative analyses supported these results: students who learned with ChatGPT produced longer, more coherent, and evidence-based explanations, used precise mathematical and logical terminology, compared alternative solution paths, and reflected on the validity of information provided by the AI.In contrast, the control group tended to provide shorter and more descriptive answers, showing limited depth of reasoning.