Embracing large language models in education: insights and reflections from the Caribbean context

Purpose Given the rapid adoption of AI-assisted coding tools within both software development and educational contexts, it is essential to understand students' experiences and difficulties. This research examines the integration of GitHub Copilot, a tool driven by a large language model (LLM), into an introductory programming course at a Caribbean university. Design/methodology/approach Utilizing an exploratory case study through action research, an instructor employed sentiment and content analysis of feedback from thirty-seven first-year undergraduate students to garner the effectiveness of GitHub Copilot tool in their learning experiences. Findings The findings reveal numerous technical and ethical issues related to the accuracy and utility of the tool's output, sentiments regarding technical challenges highlighted students' difficulties in comprehending complex code, receiving inaccurate suggestions, and handling erroneous outputs. Students expressed ethical concerns centered around plagiarism risks, diminished independent thinking, and excessive reliance on AI-generated answers. Practical implications The study emphasizes the importance of teaching prompt engineering and establishing clear policy guidelines to ensure the ethical and effective use of LLMs in higher-education settings. Originality/value Even with these challenges, the results suggest that LLMs, such as GitHub Copilot, have the potential to foster critical thinking, enhance conceptual understanding, and support self-assessment in programming education when utilized effectively.