A motivational-meaning model of AI feedback in university physical education

Synthesis: Wang and Wang (2026) ask why the same AI movement feedback reads as evidence to one student and as a demand to another, and answer with a model, not a study. It is a Hypothesis and Theory article with no sample, no data and no measured effect. Its contribution is a motivational-meaning model for university physical education that joins self-determination theory, Feedback Literacy, Self-Regulated Learning and motor-learning research around one claim: an output has no fixed motivational function apart from the framing a teacher gives it, the audience that can see it, and the evidence a learner can bring against it. Blocked translation is a temporary failure to find a defensible basis for the next attempt, not a third meaning; appropriation is the shift to independent monitoring. Seven proposition groups carry the predictions; whether appropriation is distinct from established self-regulation constructs is, by the authors' account, an empirical question.

Key Findings

The paper reports no data; these are its propositions, none tested.

1. Proposition group 1 separates three contrasts: mastery rather than pressuring framing should raise informational and lower controlling appraisal; named peer disclosure raises controlling appraisal with reference information and grading held constant; withholding verification raises next-step evidence uncertainty, more for discordant output.
2. Proposition 2: an intelligible next action added to a valid, attainable discrepancy raises later competence satisfaction beyond the same discrepancy described alone, with score, valence, specificity, practice dose and frequency matched.
3. Proposition 3: an effective right to challenge raises autonomy satisfaction and improves discrimination in reliance across valid and invalid outputs, with explanation, teacher contact and choice matched.
4. Proposition 4: named peer disclosure in assessed PE raises controlling appraisal and later evaluation anxiety against private display, more strongly at low baseline perceived competence.
5. Proposition 5 makes the interaction the prediction: discrepancy-specific review should shorten evidence-selection uncertainty more for discordant than concordant output, so neither compliance nor rejection confirms it.
6. Proposition 6: repeated informational rather than pressuring arrangements improve independent monitoring before support is withdrawn, and those indicators prospectively predict retention and changed-task transfer beyond baseline skill.
7. Proposition 7 is secondary: repeated informational appraisal should predict voluntary post-course activity through autonomous regulation, over and above baseline activity and opportunity.

Why a movement cue is not a written comment

In written feedback the learner meets a representation of the work; in PE the learner also has the movement, its felt timing and balance, and its visible consequence, so an output enters a relation among four fallible sources rather than standing as the record. Performance is also public, though the paper is careful: voluntary sharing in a cooperative climate differs from a named leaderboard rather than being a milder version of it. The output arrives inside pedagogical authority as well, where the same number can be introduced as fallible evidence or as machine-certified fact. AI changes the scale of these properties rather than creating them, and the five outcome domains stay apart because they diverge: a learner can consult a system daily and practice badly.

Two meanings, and a state that is neither

The decomposition is temporal: appraisal of what the output says or demands comes first, need satisfaction or frustration after it, the next attempt after that. Informational meaning is what the learner understands the message to permit, not a report of feeling competent; controlling meaning is an expected consequence of noncompliance, not a report of anxiety. They coexist and are scored separately, not sorted into types.

Blocked translation names the state where conflicting evidence leaves no defensible basis for the next adjustment. A discrepancy is an antecedent, not proof of it; the state is coded before later compliance or performance, so it is not defined by its own outcome; and it must beat confusion, epistemic uncertainty, distrust and need frustration. It is not a third meaning: a learner can be unsure what is warranted yet comply.

Appropriation, its rival and its tests

Appropriation is where the model stakes its claim and where it is most exposed. The proposal is a within-person transition: criterion generation, discrepancy checking, cue selection and revision, and evidence seeking that begin as externally prompted and become independent. Indicators must be process measures taken before the retention test; scoring transfer as evidence of appropriation would be circular. The authors name the competitor: the guidance hypothesis holds that frequent external feedback can support practice while retarding independent error detection. If scheduling explains the difference after withdrawal, the motivational account is unnecessary, so tests must match frequency, fading and error-contingent thresholds.

The proposition-to-test map is the most useful part: interviews and stimulated recall first, then content review and separate exploratory and confirmatory samples, then experiments isolating one contrast at a time, analyzed at the class level where disclosure reaches classmates. Ethics too: minimum necessary deception, low-risk task, no grade effect, and repair that can correct the system as well as the learner.

What this means for practice

Instructors can use the model as a design checklist. Say what the output is for, what it is compared against and where it fails, before showing it, and attach a discrepancy to an attainable adjustment rather than to a number. Default to private display and make named disclosure something to justify. Offer evidence and a route to contest it, and fade prompts against demonstrated self-monitoring. Administrators should hold consent to be recorded apart from consent to public display or reuse.

Limitations

• There is no empirical test of any kind. Construct definitions, candidate items and falsification criteria are proposals, not a validated scale.
• The evidence is borrowed. Written feedback, consumer quantification and tracker research suggest possibilities that may not reproduce in embodied PE.
• Both novel labels may collapse into existing ones: blocked translation into confusion, appropriation into feedback literacy. Each should be merged if it adds no prospective prediction.
• Continued participation is distal and contingent on prior activity, health, habit, identity, access and opportunity, so a null there says little about the event-level account.

Citation

Wang, X., & Wang, Y. (2026). AI feedback does not motivate by itself: a motivational-meaning model for university physical education. Frontiers in Psychology, 17, 1951652.