Enhancing university level English proficiency with generative AI: Empirical insights into automated feedback and learning outcomes

Abstract

This paper investigates the effects of large language model (LLM) based feedback on the essay writing proficiency of university students in Hong Kong. It focuses on exploring the potential improvements that generative artificial intelligence (AI) can bring to student essay revisions, its effect on student engagement with writing tasks, and the emotions students experience while undergoing the process of revising written work. Utilizing a randomized controlled trial, it draws comparisons between the experiences and performance of 918 language students at a Hong Kong university, some of whom received generated feedback (GPT-3.5-turbo LLM) and some of whom did not. The impact of AI-generated feedback is assessed not only through quantifiable metrics, entailing statistical analysis of the impact of AI feedback on essay grading, but also through subjective indices, student surveys that captured motivational levels and emotional states, as well as thematic analysis of interviews with participating students. The incorporation of AI-generated feedback into the revision process demonstrated significant improvements in the caliber of students’ essays. The quantitative data suggests notable effect sizes of statistical significance, while qualitative feedback from students highlights increases in engagement and motivation as well as a mixed emotional experience during revision among those who received AI feedback.