Date
Publisher
arXiv
This review examines the role of artificial intelligence (AI) agents in programming education, focusing on how these tools are being integrated into educational practice and their impact on student learning outcomes. An analysis of fifty-eight peer-reviewed studies published between 2022 and 2025 identified three primary categories of AI agents: chatbots, generative AI (GenAI), and intelligent tutoring systems (ITS), with GenAI being the most frequently studied. The primary instructional objectives reported include enhanced programming support in 94.83% of studies, motivational and emotional benefits in 18.96%, and increased efficiency for educators in 6.90%. Reported benefits include personalized feedback, improved learning outcomes, and time savings. The review also highlights challenges, such as setup barriers documented in 93.10% of studies, overreliance resulting in superficial learning in 65.52%, and concerns regarding AI errors and academic integrity. These findings suggest the need for instructional frameworks that prioritize the development of prompt engineering skills and human oversight to address these issues. This review provides educators and curriculum designers with an evidence-based foundation for the practical and ethical integration of AI in programming education.
Study at a glance
- Question: How are AI agents being integrated into programming education, and what are their benefits and challenges?
- Setting and data: 58 peer-reviewed studies on AI in programming education published between 2022 and 2025 across five academic databases.
- Method: Systematic literature review following PRISMA 2020 guidelines with MMAT and SANRA quality appraisals.
Why it matters
High dropout rates challenge introductory programming courses, but whether artificial intelligence tools effectively align with educational goals across diverse learning contexts remains uncertain.
Key findings
- Across 58 reviewed studies, 94.83% reported programming support as a primary instructional objective, followed by motivational gains (18.96%) and educator efficiency (6.90%).
- Reported pedagogical benefits included personalized feedback (44.83% of studies), learning gains (32.76%), time efficiency (29.31%), debugging support (20.69%), and 24/7 availability (15.52%).
- Key reported challenges included institutional implementation barriers (93.10% of studies), student overreliance leading to surface learning (65.52%), factual inaccuracy and hallucinations (32.76%), and academic integrity risks (15.52%).
Summary drafted with AI from the full paper and may contain errors. See the full paper for details.
What is the application?
Who age?
Why use AI?
Study design
