TRV-2026-2921Version 1 · Published

Saved · see the current record

Why this version exists

Certified into the record

The exact text we saved

This is the version exactly as it was saved, character for character. The SHA-256 hash below is a 64-character code worked out from that text. Run the same text through SHA-256 and you get the same code; change a single letter and the code comes out completely different. Each version also stores the previous version’s hash, which chains them together so no version can be swapped out quietly.

TRUVACE RECORD VERSION
record: TRV-2026-2921
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-11-14T18:02:32.294973Z
status: published
lens: g_space
sector: education
headline: Effects of AI-mediated real-time feedback on motivation and motor skill learning in traditional ethnic sports education: testing self-determination theory mechanisms
dek: Introduction Traditional ethnic sports such as Zhuang Daoyin demand precise movement, yet conventional demonstration–imitation–verbal correction limits feedback timeliness and individualization. Artificial intelligence (AI) real-time posture feedback offers a scalable option, but most evaluations emphasize technical outcomes and overlook learning processes. Grounded in self-determination theory (SDT), this study tested whether and how AI real-time feedback improves motivation and motor skill by engaging basic ps…
gain_title: In a 12-week cluster-randomized trial of 307 Chinese university students learning Zhuang Daoyin, the group using AI real-time posture feedback gained more in motor skill (d = 0.30) and competence (d = 0.36) than the group given teacher-led verbal correction.
problem_title: (none)
trace_subject: (none)
gain_reading: In a 12-week cluster-randomized trial of 307 Chinese university students learning Zhuang Daoyin, the group using AI real-time posture feedback gained more in motor skill (d = 0.30) and competence (d = 0.36) than the group given teacher-led verbal correction.
gain_evidence: The AI group showed significantly greater gains in competence ( d = 0.36, p = 0.003), autonomy ( d = 0.40, p = 0.001), motor skill ( d = 0.30, p = 0.010), and learning interest ( d = 0.38, p = 0.003)
problem_reading: (none)
problem_evidence: (none)
quick_read: Twelve intact university classes in Guangxi, China (307 students) learned the traditional Zhuang Daoyin sequence for 12 weeks. Six classes practised with an RTMPose-based system that showed live posture feedback; six received teacher-led verbal correction. The AI group gained more in motor skill (d = 0.30), competence (d = 0.36), autonomy (d = 0.40) and learning interest (d = 0.38).

The effect on motor skill ran through competence and then motivation quality: group predicted competence (β = 0.17), competence predicted relative autonomy (β = 0.34), and that predicted motor skill (β = 0.30). The direct path from AI feedback to skill was not significant (p = 0.097), which the authors read as full mediation.

Relatedness did not change (p = 0.974), which the authors interpret as showing the tool works through feedback rather than general enthusiasm. They recommend that teachers keep providing relational support alongside AI feedback. Effects on the composite motivation index were less consistent across analyses.
limitation: Effects on the composite motivation index were not consistent across analyses, and formal treatment fidelity coding with independent blind raters was not conducted.
tag: Evidence-backed gain
key_points: Twelve intact classes were randomly allocated by computer-generated sequence: 6 to AI feedback and 6 to teacher feedback, with 152 control and 155 AI students analysed. | Both groups had the same 90-minute weekly sessions, and the only difference was feedback during the 45-minute practice phase. Attendance was 94.2% (AI) versus 94.6% (control). | The AI system used a webcam and RTMPose to detect 133 body keypoints with about 120 ms latency, compared against 28 expert-validated templates, and showed a skeleton overlay and colour-coded heatmap. | Motor skill was rated from video by two blinded instructors, with inter-rater reliability of 0.94 at post-test. | Adjusted post-test motor skill means were 3.054 for AI versus 2.914 for control (ANCOVA p = 0.010); the mixed-model group-by-time interaction for skill was b = 0.225 (p = 0.006). | Course satisfaction, concentration and vitality did not differ significantly between groups, and the total effect of AI feedback on satisfaction was not significant (p = 0.296). | The structural model explained 10.0% of the variance in motor skill.
rundown: Zhuang Daoyin is a traditional Zhuang ethnic practice with complex movement patterns, usually taught by demonstration, imitation and verbal correction in which one teacher watches many students. The researchers argue this limits how timely and individual the feedback is, and tested whether live AI feedback could help while also examining students' psychological experience.

Participants were second-year undergraduates at a university of Chinese medicine, with no prior training, who attended a public PE elective. In the AI condition, instructors circulated only for safety and gave no technical corrections, so the system was the sole source of correction. Control students received scripted verbal feedback from the same three instructors, each teaching two AI and two control classes.

Group differences were tested with mixed models and ANCOVAs, and mediation with structural equation modeling. Class-level intraclass correlations were 0.03 or lower. Sensitivity analyses, including per-protocol and cluster-robust checks, left the core mediation findings directionally unchanged.

The composite relative autonomy index favoured the AI group in the ANCOVA (p = 0.042) but not in the mixed-model interaction (p = 0.209), so the effect on overall motivation was smaller and less robust than that on competence and skill.
sources:
- peer_reviewed | Frontiers in Psychology | https://doi.org/10.3389/fpsyg.2026.1958348 | 2026-10-16
prev: 0000000000000000000000000000000000000000000000000000000000000000
SHA-256 hash
28badd4360418a213931824bb8601795fc15b2ef470553f418be4c62373462de
Previous version’s hash
0000000000000000000000000000000000000000000000000000000000000000
Check this record hasn’t been altered
How to check it yourself, without trusting this page

Download any version from /api/record/TRV-2026-2921 and hash its canonical field yourself, for example with shasum -a 256. You should get its content_hash. Each version’s text also ends with prev:and the hash of the version before it (version 1 points to 64 zeros). If anyone had changed a single character of any version after it was saved, those hashes wouldn’t line up.