Guided AI integration model (GAIIM): Effects of verification-oriented generative AI-supported learning on undergraduate students’ algebraic problem-solving

Abstract

Artificial intelligence (AI) is increasingly being integrated into higher education; however, most instructional applications emphasizeanswer generation rather than mathematical reasoning and critical verification of AI-generated mathematical solutions. The current study examined the effectiveness of the Guided AI Integration Model (GAIIM), a verification-oriented instructional framework that utilizes AI for verification of students’ algebraic problem-solving within a problem-solving context. An explanatory sequential mixed-methods design was employed. Quantitative data were collected from 86 undergraduate mathematics education students using a randomized pretest-posttest control group classroom design. During the 12-week intervention, students in the experimental group engaged in GAIIM-based instruction involving independent problem-solving, AI-generated solution comparison, error detection and justification, error resolution, and reflective verification, while the control group received conventional algebra instruction. Qualitative data were subsequently collected from 10 purposively selected students through semi-structured interviews and analyzed thematically to explain and elaborate on the quantitative findings. Students in the GAIIM group demonstrated significantly higher posttest performance in algebraic problem-solving, error detection, and error resolution after controlling for pretest scores. Qualitative findings further revealed stepwise verification, iterative engagement with AI feedback, conditional trust in AI-generated responses, and difficulties interpreting and correcting AI-generated mathematical reasoning. The findings suggest that structured AI-supported verification can support students’ algebraic problem-solving and error-verification performance.