Generative AI Based Learning In Computer Scince For Beginners

Abstract

The integration of Generative Artificial Intelligence (GenAI), specifically Large Language Models (LLMs) such as ChatGPT and GitHub Copilot, into Computer Science (CS) education presents a paradigm shift in how introductory programming is taught and learned. While these tools offer unprecedented support for syntax correction and code generation, concerns persist regarding their long-term impact on the cognitive development of beginners. This study investigates the effects of GenAI-assisted learning on novice programmers' performance, self-efficacy, and conceptual understanding. Adopting a mixed-methods approach, sixty first-year CS students were divided into an experimental group (AI-assisted) and a control group (traditional resources) to complete a series of Python programming tasks. Quantitative analysis revealed that while the AI-assisted group demonstrated significantly faster completion times and higher code correctness during the intervention, they scored lower on a subsequent non-assisted conceptual assessment compared to the control group. Qualitative data from post-experiment surveys indicated that while GenAI reduced anxiety related to syntax errors, it inadvertently fostered an "illusion of competence," where students believed they understood logic that they had merely generated.