Sal Khan Has Big Dreams for AI in Education. Are They Too Big?

livemint.com:

A small study published 40 years ago set off shock waves in education circles for decades with claims that the right kind of tutoring vaults middling students to the top of the class. Ever since, a number of educators and tech leaders, inspired by the findings, have been chasing these dramatic gains. The latest to take up this cause is Sal Khan, perhaps the country’s leading evangelist for the potential of artificial intelligence in education and whose nonprofit has developed an AI chatbot.

Last year Khan opened a widely viewed TED Talk by describing the remarkable results from the 1984 study. AI could serve as a personal tutor that might eventually be able to “take your average student and turn them into an exceptional student," he said. Other AI enthusiasts have cited the same research—a sign of the heady hopes that the technology could bring an educational breakthrough.

But it isn’t clear whether these ambitions are realistic. The 1984 results are outliers, according to a paper published Thursday by Education Next, a policy journal affiliated with Harvard University. The small-scale experiments were based on two subjects—probability and cartography—that students had little familiarity with, which made their rapid learning improvements more feasible, the paper noted.

Modern studies of human tutoring haven’t typically shown gains nearly as large. While many researchers and educators are optimistic about the potential for AI in education, some worry that current hype will give way to disappointment if the goals are too audacious. “Overpromising can lead to disappointment, and reaching for impossible goals can breed questionable educational practices," wrote University of Texas at Austin education professor Paul von