Algorithms to help teachers teach and students learn

Over the last few decades there's been a push to make teaching more personalized for each learning style. But how can one teacher tailor lessons to 20 or 30 kids?

Joel Rose ran into this problem when he was a fifth-grade teacher in Houston more than 25 years ago. Joel's solution was found in algorithms that track student performance and recommend lesson plans. He co-founded New Classrooms, the organization behind the "Teach to One" math program. Teach to One's software tools help build schedules for each student based on daily assessments. It's been implemented in dozens of schools in the US.

It's part of a much broader trend towards blended learning, where part of education happens online. The trend takes advantage of the fact that algorithms can see how a student is progressing and customize the learning so students progress at their own pace. 

Rather than a teacher at the front of a large class of student, Teach to One uses a combination of stations, or modalities. "At some stations kids work with teachers, in some they work with software, and in some stations they might work with one another on different types of projects," Joel says. "When you walk in, you look up and you see a big TV monitor like you might see at the airport." The monitor directs the students to different assigned stations.

"You might spend the first 10 minutes at station number two working with a teacher on how to calculate the area of a triangle. The next 30 minutes is station number six with some software on how to calculate the area of a triangle. then in the last 10 minutes you take an online assesment."

Joel's system then uses that daily assessment to create a new schedule for the student, specifically tailored to their needs. 

The system is powered by a set of algorithms. "We want to make sure kids are getting a broad range of modalities or approaches... but what we're now evolving to is we're able to use the data we get everyday not just to figure out what's working for each student but to actually figure out for the thousands of students that have been the exact same situation, what did they do? What worked? What didn't work? And how do we used all that information to inform each student's schedule the next day."

One criticism of algorithm-based educational tools is that they make students better at standardized tests - not critical thinking. But Joel believes that Teach to One can better prepare struggling students.

"In the traditional model, especially in the US, teachers are responsible for the results on the state test. So they need to teach the curriculum on the state test regardless of where the student is. What we're doing is meeting every student where they are... and that, I would argue, is a much healthier experience to experience in school than is the current model."