I do some work with the Stanford University Graduate School of Education. Last week I had a chance see a brilliant board presentation from faculty member Candace Thille on the future of online learning.

The last couple years we’ve seen lots of new developments in the online learning space, from Khan Academy to ambitious higher education consortia such as EdX and Coursera. But so far online education has mostly been focused on the high-leverage advantages of the “broadcast” model – one teacher now can reach a ten thousand students. Which is great but the problem, of course, is that we risk just amplifying bad teaching. What we really want is smarter teaching.

Professor Thille walked us through an innovative new math application aimed at middle-schoolers. The student solves a word problem which the application presents to him, and if he gets stuck anywhere he can click the “hint” button and the application displays a hint which helps him get to the next step in solving the problem. The system tracks every interaction, of course, so we know exactly where the student got stuck and how he eventually solved the problem. Now aggregate that data from 10,000 students, and you have amazingly powerful information on learning habits. Then make a couple changes to the lesson plan, and instantly see how it changes the way students solve the same problem. Optimize and repeat. Aggregate data doesn’t lie.

This is pretty profound, and clearly represents the next generation of online learning. As Candace said “This is the secret sauce of Facebook, Google, and Netflix. They no longer have to guess whether people are understanding content – today’s leading technology companies know how to use empirical data to optimize learning and understanding”.

In many ways, this is the holy grail of education research. Until now, K-12 pedagogy improvement been a pretty slow process: university researchers develop a new way to teach math to middle schoolers, then slowly work to get it published, then classroom teachers slowly begin to use the new methods, and maybe 5 years later the standardized tests might start to show whether the new math teaching method is working.

Now we can iterate and improve, iterate and improve. On the fly. At 1000x the speed.

It’s “big data meets education”. And it’s pretty exciting stuff.