Learning Gain and User Experience of AI Avatar–Based and Human-Presented Explainer Videos: Prospective Randomized Crossover Feasibility Study

In plain English

Thirteen undergraduate engineering students watched the same fuel-cell explainer video presented either by an AI avatar or by a human, in randomised order. Both formats produced substantial short-term learning gains and the gap between them was not statistically significant (median 5 vs 4.5 newly correct items, P=.51), but participants rated the human presenter higher on every dimension of the AttrakDiff2 experience scale while the avatar came out as merely neutral. The authors are explicit that with 13 participants a null result is not evidence of equivalence — read this as a workable study design for anyone wanting to test avatar presenters properly, not as clearance to swap presenters out.