Research Article
A qualitative study of physical education teachers' perceptions of artificial intelligence and influencing factors based on social cognitive theory
Synthesis: Yang, Chen and Rao interviewed university physical education teachers about how they read AI in their teaching, and the answers divide along a line that is not technical skill. The teachers trust AI with plans, materials and administrative work, and they doubt it can see a hip rotation, correct a landing or pass on tacit knowledge through demonstration. Because that embodied core is what they take physical education to be, resistance sits in the discipline and in resource decisions rather than in a training deficit.
Key Findings
- Confidence tracks the task, not the teacher. The same teachers who feel capable with AI on lesson plans become unsure when it is aimed at movement instruction, so adoption is not one general skill.
- The body is the barrier they name. Physical education rests on bodily perception, on-the-spot correction and demonstration, and teachers treat AI's lack of a body as a difference of kind rather than a stage of maturity.
- The setting does not match the tool. Classes are outdoors, in motion and unstandardized, while the AI they had met was strongest at standard text that still needed heavy reworking.
- A colleague persuades more than a policy. Watching someone use video to analyze a student's form changed what teachers would try; school promotion raised awareness without supplying equipment, funding or subject-specific training.
- Equipment follows the medal teams. Hardware and money attach to squads that win honors, leaving ordinary classes short, and teachers read the allocation as a statement about who the subject serves.
- Use is concentrated and sometimes staged. AI appears mostly in preparation and rarely in live teaching, and some teachers adopt it for demonstration lessons while routine classes stay traditional.
Why confidence collapses at the gym floor
The interviews involved university physical education teachers recruited purposively, ran about half an hour each, and were coded in NVivo against social cognitive theory's three-part structure before open coding added what it missed. The self-efficacy that emerged moves with the task. A teacher who has used a language model to polish a thesis calls himself confident; asked for a term-long plan for his own classes, he describes reworking the output until little remains. The authors call this situated self-efficacy, and it explains why a general readiness survey predicts so little about a gym.
The theme behind it is what they label the repellent effect of embodied disciplinarity. Sport is physical activity, and a system without a body cannot demonstrate, feel or correct in the same second. The setting compounds this: outdoor classes, variable weather and mixed ability do not resemble the standardized text tasks where the tool performs well. Teachers do not treat that as a gap a better model will close. They treat it as a boundary of the discipline, which places the study in the discipline-specific record of adoption.
The resource gap teachers read as a verdict
School encouragement existed, but teachers described it as generic: lectures in group chats, no systematic training, no equipment in the department and funding that did not reach them. Peer demonstration was concrete by comparison: a colleague analyzing a Tai Chi form on video showed what the technology could do in a real class rather than what it promised.
Resource stratification cut across both. AI equipment and annual funding went to competitive teams whose results bring honor to the institution, while general classes lacked large screens, connectivity and budget. Read through subjective norms, the school's expectation is a soft pressure meeting a hard material limit, and the limit wins. One teacher said general courses should focus on getting students to exercise. That is a judgment about whom the institution expects the technology to serve, not merely a complaint about scarcity.
Display, de-skilling and the identity stake
The third theme, performative adoption, explains a pattern that would otherwise look inconsistent. In demonstration lessons teachers bring AI forward to project sophistication; in routine teaching they keep it at the desk. The behavior compromises between two expectations held at once: AI saves preparation time and makes teaching look modern, and it raises the fear of becoming an equipment manager who reviews data while someone else teaches. The authors treat that fear as a finding, not as resistance to be corrected.
What this means for practice
- Instructors: Map your tasks before adopting anything. Keep AI on planning, materials and administration where it returns time, and tell students plainly that movement diagnosis stays with you; confidence here followed usefulness, not encouragement.
- Faculty developers: Run subject-specific sessions, not generic tool training, and design a first project where teachers see AI work with their own students, such as fitness testing or motion analysis. The useful question is not how to prompt but what this can do that my eyes and voice cannot.
- Administrators: Audit where AI equipment and funding land. If the answer is the competitive squads, extend the allocation or expect the department to read adoption as somebody else's project.
- Researchers: Test the proposed mechanism. The claim that peer demonstration outweighs institutional support needs a design comparing the two.
Limitations
- The sample is sixteen university physical education teachers in one region, recruited purposively, so the themes describe that setting, not the subject as a whole.
- Interviews ran about half an hour and relied on self-report; no class was observed, so accounts of practice are unchecked.
- Deductive coding fixed the categories before coding began, which may have kept apart themes teachers experience together.
- Recruitment required prior exposure to digital technology, so teachers who avoid it are absent, and the university focus leaves school physical education unmeasured.
Citation
Yang, W., Chen, H., & Rao, F. (2026). A qualitative study of physical education teachers' perceptions of artificial intelligence and influencing factors based on social cognitive theory. BMC Psychology, 14, 530.