The test came after the advice stopped. In an eight-week randomised trial at Anyang Normal University, 240 undergraduates received structured written AI feedback plus reflection or oral instructor feedback, with the same curriculum, practice and feedback schedule.[1]
Humans supplied shooting and movement observations. AI composed a correction and practice target; students then wrote their next action. Blinded assessors rated technique from videos.[1]
At the no-feedback test
| Outcome | Adjusted advantage | 95% confidence interval |
|---|---|---|
| Successful free throws | 6.1 percentage points | 3.7 to 8.5Source 1 |
| Technique (5–25 scale) | 1.63 points | 1.01 to 2.25Source 1 |
Published 30 September, the study lacked preregistration. It evaluates the combined programme, without establishing an AI-only effect, slower forgetting or long-term transfer.[1]
When help becomes a skill
The distinction between performing with guidance and learning something that survives its removal predates chatbots. In a 1994 experiment, Winstein and colleagues compared physical guidance with information about results while people practised movements towards a target. They also varied how frequently help was supplied. Frequent physical guidance produced the poorest retention; both high-frequency conditions produced the least accurate transfer to the subsequent test.[2]
That experiment provides a reason to test beyond supported practice. It does not prescribe one ideal feedback schedule for every sport. A correction can help someone complete the present attempt while leaving an unanswered question about what they will manage alone. The assessment must make room for that question.[2]
For educators considering AI, the useful ambition is therefore concrete: build a teaching routine that helps a learner notice a problem, choose a correction and eventually perform without another prompt. A fluent explanation is an input to that process. The learner’s next unaided attempt is where its value becomes visible.
Sources & further reading
- From generative-AI feedback to sport skill learning: the roles of perceived competence and autonomous motivation in university physical education Frontiers in Psychology · 30 Sep 2026 · Accessed 28 Oct 2026
- Effects of physical guidance and knowledge of results on motor learning: support for the guidance hypothesis Research Quarterly for Exercise and Sport · Accessed 28 Oct 2026