Generative AI-Driven Practical Teaching Reform of Artificial Intelligence Application Courses

Abstract

With the rapid proliferation of generative artificial intelligence (AI) in higher education, AI-related practical courses generally face problems such as student over-reliance on AI tools, diminished depth of algorithmic understanding, and insufficient code reproduction capability. These issues frequently lead to critical learning gaps, notably the "understanding--generation disconnection" and the "understanding--reproduction decoupling." To address these contemporary challenges and enhance students' comprehensive practical abilities in real project scenarios, this study utilizes the Artificial Intelligence Application course at a southwestern engineering university as a primary vehicle. We construct an integrated teaching mechanism comprising "specification--generation--comparison--verification--evaluation" and carry out a comprehensive teaching reform through two rigorous rounds of classroom action research. Based on 400 valid questionnaires and detailed course test data collected from 120 students, SPSS 17.0 is utilized to analyze scale reliability, validity, and score differences. The empirical results show that in the second round of teaching, students' performance in theoretical understanding, hands-on reproduction, and project presentation significantly improved compared with the first round. Furthermore, prompt strategies and fact-checking behaviors are significantly positively correlated with overall learning outcomes. Ultimately, the study demonstrates that systematically introducing generative AI usage norms, prompt training, and fact-checking mechanisms can effectively mitigate students' substitutive reliance on AI tools, thereby promoting deep learning and sustainable skill acquisition in practical engineering courses.