Human-AI collaborative lesson design is associated with enhanced student outcomes and planning quality in secondary physical education: a randomized experimental study

Résumé

Background Artificial intelligence (AI) planning tools are increasingly introduced in physical education (PE); however, empirical evidence regarding their pedagogical value remains limited. Objective This study aimed to compare the effects of three PE lesson design conditions—AI-only, Human-only, and Hybrid (AI + teacher)—on students’ basketball shooting performance, documented lesson-plan quality, student enjoyment, and planning efficiency. Methods Ninety secondary school students ( M age = 14.7 ± 1.2 years; 50% girls) were randomly assigned to three groups ( n = 30 each). All participants completed a six-session basketball shooting unit delivered by the same certified PE teacher. The primary outcome was shooting accuracy (/30). Secondary outcomes included documented planning quality assessed via the Assessing Quality Teaching Rubrics (AQTR) and student enjoyment measured using the Physical Activity Enjoyment Scale (PACES). Data were analyzed using one-way ANCOVA, Kruskal–Wallis tests, and repeated-measures ANOVA. Results All groups demonstrated significant improvements from pre- to post-test. The Hybrid condition yielded the largest improvement in shooting performance ( d = 1.42), followed by Human-only ( d = 1.01) and AI-only ( d = 0.78). A significant group effect was observed, with the Hybrid condition outperforming both alternatives. In terms of lesson design, the Hybrid condition achieved the highest total AQTR score, with the greatest advantage observed in the Instructional Guidance sub-dimension. Students in the Hybrid group also exhibited a significantly more positive trajectory of enjoyment across the six sessions compared to the other groups. Additionally, the Hybrid condition required substantially less planning time than the Human-only condition (−67%). Conclusion The Hybrid approach was associated with superior student outcomes, higher-quality lesson planning, and greater planning efficiency. These findings support the use of generative AI as a pedagogical support tool within a co-design framework, rather than as a replacement for teacher expertise. Clinical trial registration Doi: 10.17605/OSF.IO/Y4RZQ .

Citer ce document

Accès au document

Texte intégral en lecture en ligne, réservé aux abonnés SPHAERO et aux membres de l'institution. Se connecter

Auteur(s)

Statistiques

Consultations : 1

Téléchargements : 0