Generative AI and Multimodal Profiling in Programming Education: A Systematic Review of Tools, Pedagogical Impacts, and Future Competencies

Abstract

This systematic review investigates the integration of generative artificial intelligence (GenAI) and multimodal learning analytics (MMLA) in programming education, with a focus on empirical implementations and their pedagogical implications. Drawing from 59 peer-reviewed studies published between 2015 and 2025, this review categorizes how GenAI tools such as ChatGPT, GitHub Copilot, and domain-specific tutors are being used to support real-time feedback, code generation, and learner scaffolding. While these tools enhance productivity and interaction, they also raise new challenges around critical thinking, overreliance, and the redefinition of assessment practices. The findings highlight a lack of longitudinal and cross-cultural studies and limited integration with multimodal sensing, such as eye tracking and emotion detection. We propose future directions for designing adaptive, ethically grounded learning environments that leverage GenAI alongside real-time multimodal feedback. This includes fostering AI literacy, ensuring learner agency, and equipping educators with tools for reflective and inclusive AI pedagogy.