That's how most things work in education. Students who are motivated and self-disciplined can learn from live teachers, online classes, books, videos -- and now bots. The educationally rich get richer. But most students aren't "independent learners."
Since human tutors are in short supply and many students are way behind grade-level expectations, it would be great if AI tutors could fill the gap, they write. It comes down to motivation.
In a 2018 Khan Academy study, only 11 percent of students used online classes for 30 minutes a week, as recommended, Geraghty and Goldstein note. Those students made learning gains. The vast majority of students didn't use the resource very much and made no gains.
In the pre-AI era, computer-based tutoring -- think Zearn and Dreambox -- has been less effective than human tutors, research has found. In part, that's because "ed tech is just an instructional intervention, not a motivational one," they write.
AI appears to be better at instruction than the old educational software, write Geraghty and Goldstein. "Already we see that Khanmigo helps a confused (yet motivated) student more than OG Khan Academy."
But they doubt that "red-light" students -- those who actively resist tutoring -- will be helped significantly by even the smartest, cuddliest bot.
"Yellow-light" students, who are up and down, depending on their mood, may flip to green because AI's "improved instruction generates some learning success, and that in turn leads to higher motivation," they predict. There's "promise" for autistic students. But the big gains will come for motivated "green-light" kids who are actually going to use AI to improve.
Geraghty and Goldstein imagine replicating Khan's 2018 study with AI-powered Khanmigo in 2025.
They think 20 percent of students will use AI tutoring, up from 11 percent, and will make larger learning gains. It will be "amazing." For some.