Adoption, Usage, and Perceptions of LLM-Generated Feedback in Computer Science Education: A Cross-Cultural Study of Students and Academics

This full paper explores the use of Large Language Models (LLMs) for generating assessment feedback in Computer Science (CS) education, comparing adoption patterns and perceptions of students and academics in the UK and Saudi Arabia (KSA). The study investigates (1) differences in LLM use and adoption between students and academics in both countries, and (2) participants' perceptions of LLMgenerated feedback, focusing on benefits, challenges, and areas for improvement. A mixed-methods approach was employed, involving 251 participants. The collected data were analyzed through descriptive statistics and thematic analysis. Findings indicate that students in KSA demonstrate higher usage rates of LLMs, with ChatGPT identified as the most frequently used tool. Although KSA academics show limited use of LLMs, UK academics did not use them for feedback generation. Thematic analysis revealed three key themes: (a) “It saves effort”, (b) “It is not as smart and genuine as a human”, and (c) “Making it more like humans and less like a machine”. The findings emphasize the need for institutional policies to guide the responsible integration of LLM-generated feedback, ensuring alignment with teaching goals and improving feedback quality.