The integration of artificial intelligence (AI) to assess English writing has become prevalent in higher education, though empirical research on its efficacy has yielded inconsistent results. To clarify this variability and identify moderating factors, this study conducted a three-level meta-analysis synthesizing 234 effect sizes from 43 valid studies. The analysis demonstrated a significant, moderate overall impact of AI feedback on writing performance (g = 0.58, p < 0.001). The finding also revealed that AI feedback was more effective when integrated with human feedback in a hybrid intelligence framework, particularly in cooperating with teacher feedback. Additionally, AI demonstrated the largest effect when providing feedback on narrative writing tasks. These findings support viewing AI as a valuable feedback tool and emphasize the importance of hybrid intelligence in enhancing feedback effectiveness.
LUO et al. (Tue,) studied this question.