A Study on the Teaching Application Effectiveness of New-Format Physical Education Textbooks in Higher Vocational Education under External Artificial Intelligence Support: Based on Hierarchical Regression and Random Forest Analysis
Year: 2026
Type: conference-paper
Study design: Clinical Trial
Abstract: At present, new-format physical education textbooks in higher vocational education are usually accompanied by QR codes, videos, courseware, and online learning platforms, whereas mature textbooks with natively embedded artificial intelligence functions remain limited. In response to this practical problem, the present study introduced generative artificial intelligence as an external support variable in the instructional application of existing new-format physical education textbooks. Using these textbooks and their supporting digital resources as the carrier, the study combined an 18-week teaching experiment, questionnaire survey, and interviews to compare instructional effectiveness under conventional application and external AI-supported application. A total of 208 valid questionnaires were obtained, including 104 from the experimental group and 104 from the control group; the overall Cronbach's alpha coefficient of the scale was 0.912, and the KMO value was 0.901. Independent-samples t tests showed that the experimental group significantly outperformed the control group in learning engagement, classroom participation, perceived personalized support, extended learning after class, and overall satisfaction. Correlation analysis indicated that external AI support was positively associated with overall textbook application effectiveness (r = 0.517, p < 0.01). Hierarchical regression further showed that external AI support had a significant positive predictive effect on overall textbook application effectiveness (β = 0.350, p < 0.001), that perceived personalized support played a partial mediating role (β = 0.389, p < 0.001), and that teachers' digital literacy and students' technology acceptance both had significant positive moderating effects. To enhance the computational dimension and reliability of the study, a Random Forest model was additionally used for verification. The results showed that perceived personalized support, teachers' digital literacy, external AI support, and students' technology acceptance were the four most important predictors of overall textbook application effectiveness, with a five-fold cross-validated mean R2 of 0.623, an MAE of 0.146, and an RMSE of 0.186. The findings suggest that embedding external AI tools into existing textbooks represents a low-cost, feasible, and computationally interpretable pathway for promoting the intelligent application of new-format physical education textbooks in higher vocational education.
Authors: Qun-li Xing, Chao Wang
Language: English
Subfield: Education
Field: Social Sciences
Domain: Social Sciences
SDG: Quality education
Open Access status: hybrid
OpenAlex ID: W7223788780