Decoding Learning Dynamics: Unveiling AI’s Impact in Video-Based Education Through Multiple Regression Analysis

This study aims to examine whether prior academic performance (Pretest) moderates the relationship—via interaction effects—between student-generated prompts (classified into high-level [HighQuan] and low-level [LowQuan] prompts based on Bloom’s taxonomy) and final achievement (Posttest) in a video-based learning environment, addressing a critical need for personalized AI interventions in higher education. Conducted at the Business and Commerce College in China, the research involved 56 third-year marketing undergraduates (21 males, 35 females) interacting with AI chatbots while watching instructional videos. We collected data on Pretest, Posttest, and prompt frequencies, and analyzed interaction effects using multiple regression analysis. The findings reveal significant negative interaction effects, with prior performance reducing the positive impact of both high-level and low-level prompts on achievement. A novel insight is that higher-performing students benefit less from these prompts, challenging the assumption that they would gain more from high-level prompts. These results highlight the importance of adaptive AI systems tailored to individual learner profiles, offering significant implications for designing effective educational technologies in marketing education.