Learning, Behavior, and Pedagogy: A Systematic Review of Generative AI Use in Programming Education

With the rapid development of Generative Artificial Intelligence (GAI) technology, programming education has emerged as a core application domain. Through a systematic literature review of 45 relevant studies from the Semantic Scholar database from 2023-2025, this study examined the current applications of GAI as an auxiliary learning tool in programming education, and its impact on learning outcomes. The findings reveal that GAI-assisted instruction demonstrates significant effectiveness across seven learning indicators: programming knowledge and skills, computational thinking and logical reasoning, problem-solving ability, programming self-efficacy, learning achievement, code quality, and learning behaviors and engagement. While the majority of studies confirm that GAI enhances student performance in various areas such as task completion, test performance, code structure and quality, and promoting self-directed learning, some studies indicate that GAI use may reduce learning depth and lead to over-dependence in specific tasks or complex reasoning contexts. From a pedagogical perspective, GAI prompts a transformation in teachers’ roles from knowledge transmitters to learning facilitators and guides, necessitating corresponding adjustments in curriculum design and assessment approaches. Based on the empirical findings, this study constructs an integrated conceptual model for GAI-assisted programming education integrating four core dimensions: implementation context factors, core influencing factors, learning performance indicators, and learning outcomes. The study identifies AI tool selection, students’ foundational abilities, and task complexity as key variables affecting learning effectiveness, and synthesizes seven patterns of student learning behavior changes under GAI assistance, providing concrete theoretical foundations and implementation guidelines for educational practice.