Learning with AI: A systematic review of novice programmer use of large language model-based coding assistants, benefits, downsides, and challenges
Year: 2025
Type: review
Study design: Systematic Review
Abstract: The recent advances in generative AI (genAI) and LLM-based coding assistants have introduced new possibilities for computing education. While many focus on potential applications of these programming assistants, others express concern about their effect on novice programmers and strive to establish the best way to integrate these tools into the classroom. This thesis performs a systematic literature review of 22 studies to examine how novice programmers use LLM-based coding assistants and what educational benefits and downsides these bring. It further investigates the limitations of the tools and challenges in using them. This thesis found that novice programmers use coding assistants mainly for asking immediate help instead of improving their skills and knowledge. This usage trend led to the main downsides of overreliance and impeded skill development. Moreover, coding assistants provide some benefits, including efficiency and an improved learning experience. Additionally, novice programmers report the challenge of coding assistants often "misunderstanding" them and the problem of inaccuracy responses by the model. From these findings, this thesis highlights several implications for both learning programming and programming educators. Finally, it suggests recommendation practices on how to best adapt and integrate LLM-based coding assistants to support novice programmers' study.
Source: Aaltodoc (Aalto University)
Author: Bui, Duc Manh
Institution: Aalto University
Language: English
Subfield: Computer Science Applications
Field: Computer Science
Domain: Physical Sciences
SDG: Quality education
Open Access status: green
OpenAlex ID: W6982972069
