Do we learn better when instructional content is freaking adorable? As someone who’s worked with military audiences for most of her career, this was never really a question I had to answer. The military has long operated on the assumption that warfighters learn best when they are being cranked at, which is definitely not cute. But should they reconsider their position on this? With AI-generated voices becoming increasingly sophisticated, instructional designers have more options than ever for making educational content sound engaging, friendly, and, yes, cute. But does making a narrator sound like a Pokémon actually help people learn? Fortunately, some researchers decided to investigate this very important question.
In a 2024 study published in The Journal of Experimental Education, Jing and colleagues compared learning outcomes from instructional videos narrated by a real human, a formal AI-generated voice, and a “cute” AI-generated voice. Thirty university students watched instructional videos while researchers tracked their eye movements, measured their learning performance, and assessed their perceptions of the narrator. Interestingly, participants didn’t perceive significant differences in the narrators’ personas across the three conditions. Interestingly, the human-narrated videos produced significantly better learning performance and attentional engagement than the AI-narrated versions. But most importantly, the cute AI voice didn’t significantly improve or impair learning compared with the regular AI voice. The researchers confirmed that participants actually perceived the cute voice as cuter, but apparently all that adorableness didn’t translate into better learning.
Before we declare that learning from Professor Pikachu is off the table, there is an important caveat. The researchers didn’t use exactly the same instructional content across conditions. Each video taught a different set of vocabulary words, meaning that differences in difficulty could have influenced learning independently of the narrator. The researchers controlled several aspects of the presentation, but this wasn’t the apples-to-apples comparison I would have liked to see. And, of course, AI voice technology is improving at a ridiculous pace. Still, the findings speak to a broader question for those of us developing AI-enabled learning technologies: Does making instructional content more appealing actually make it more effective? The answer isn’t necessarily yes, and that’s something we should be testing rather than assuming.