Generative artificial intelligence (GAI) has brought new ideas for optimizing students' learning. Despite increasing attention on the effects of GAI on learning outcomes (LO), research results are inconsistent. While GAI's educational benefits are qualitatively described, there is substantial debate about its actual impact on students' LO. The study sought to quantify GAI's impact on students' LO, evaluating its overall and average effects, and examining four key moderating factors: functional types of GAI, educational levels, intervention duration, and knowledge domains. Based on the screening criteria, 26 empirical studies were selected from 5,887 peer-reviewed papers. Two researchers collaboratively completed the literature screening and coding process. The research employed a meta-analytic method to calculate the impact of GAI on learners' LO, and examined four moderating factors. GAI exerts a significant but small overall effect on students' LO (g = 0.392), with varying impacts on physical (g = 0.701), social-emotional (g = 0.347), and intellectual (g = 0.372) outcomes. The changes of GAI's functional types have no significant effect on LO, but three other moderating factors do, showing significant statistical differences. GAI more significantly impacts primary school students, especially in supporting their intellectual and social-emotional outcomes. Longer interventions have a greater effect on LO than short ones, particularly intellectual and physical outcomes. GAI's effects vary across knowledge domains, possibly due to its adaptability in different subjects. Long-term GAI in higher education boosts intellectual and physical outcomes, especially in education and humanities and arts, while short-term use in primary education enhances social-emotional outcomes. Integrating diverse learning components and adjusting GAI implementation parameters can optimize its effectiveness in terms of enhancing LO across different levels of education.
Oversett med Google Translate
-
