By Josh Dzieza
Brad McIlquham was tutoring at-risk youth in Durham, N.C., when a former co-worker gave him the educator’s equivalent of the Social Network pitch. What if, instead of teaching at most 50 kids a year, you could help bring personalized tutoring to 100,000, or a million kids?
McIlquham’s co-worker, Jose Ferreira—who had taught SAT and GMAT prep with McIlquham at Kaplan—was proposing an upending of the traditional “teach to the middle” classroom model. When teachers instruct students of varying ability in the same class, some students get bored, while others struggle. And often, teachers don’t discover which students have failed to understand key concepts until their tests get graded. But by then, they’ve already fallen behind. In the meantime, all the potentially useful data from students’ individual homework assignments, quizzes and textbook exercises—everything but the final grade—disappear into the ether.
McIlquham and Ferreira came up with an idea to capture that data and use it to create digital education tools that help tailor the curriculum to each student as he or she learns—by detecting gaps in knowledge early on, recommending the appropriate exercises to help students acquire skill and alerting teachers when students are struggling. “Our goal is to personalize education,” says McIlquham who along with Ferreira, and eight other co-founders, called their new education startup Knewton, “to take educational content and understand not only the ins and outs of that content, but how students interact with it—when students run into difficulties, when they start to forget things—and use it to customize the educational experience.”
Knewton, which launched in 2008, bills itself as an adaptive learning “platform,” a behind-the-scenes service that schools can use to personalize their existing digital coursework. Assisted by Knewton, schools can monitor students’ progress as they work through lessons and make sure students are grasping the material before moving on. In its early years, Knewton was designing its own digital coursework. But since 2011, it has partnered with the textbook publishing company Pearson, combining its analytics tools with that company’s educational material; and last summer Knewton announced a similar partnership with Houghton Mifflin Harcourt publishing company. So far, Knewton’s also received $54 million in funding from Pearson and a collection of venture capital firms.
Knewton didn’t invent adaptive learning: There are a lot of digital education tools that tailor coursework to individual students, giving them more difficult problems as they get better at solving them, for example. The technology has become increasingly popular with the growth of the “flipped classroom,” a way of organizing courses so that students watch video lectures and do reading at home, then do coursework and exercises in class, where teachers are there to help them.
Knewton brings advanced data analysis to this model, looking at factors like how much time students spend on specific questions and whether they consistently fall for certain false answers. “This shows a misconception, that they’re thinking about a concept in the wrong way,” McIlquham says. It’s something that might be easy to fix, but would be difficult to detect from looking at the results of a single test.
McIlquham emphasizes that this kind of adaptive process is a boon to teachers as well as to students, giving them new insight into what lessons are working, what concepts need to be revisited and which students are falling behind. “Teachers will be so much better equipped when they walk into the classroom,” he says.
As Knewton gathers more and more data, McIlquham says, it will also be able to figure out patterns in learning, drawing connections between certain types of students and what learning methods work best for them—a sort of Netflix-style “people who did well with this exercise also did well with this” recommendation engine. “If you’re a similar student and you struggled with something I struggled with, we can see that if I learned it the same way, data suggests you’ll learn well that way too,” explains McIlquham.
Knewton will have a lot of data to work with. Through its partnerships with Pearson and Houghton Mifflin Harcourt, it is now “powering” interactive education programs for 3 million students, and will reach up to 10 million by the end of 2014, from kindergarten all the way through college.
Mcllquham envisions an educational system where grade levels and semesters fall away, and students progress at their own pace, learning key concepts in small groups with help from a teacher. One of Knewton’s earliest test programs involved remedial math students at Arizona State University, which contracted with Knewton in 2011 to design an adaptive math program. ASU had a dropout problem. Some of its remedial students had been away from school for 10 years and needed a quick refresher, while others had never received basic math education in the first place. When these students were dumped into the same classroom, few received the right kind of instruction and many dropped out.
Knewton’s program let more advanced students skip concepts they already understood and focus on ones they didn’t, while an instructor went from student to student giving individual help. Initial figures from Knewton’s adaptive program at ASU showed that withdrawal rates dropped by half after two semesters and the proportion of students getting passing grades rose from 64 percent to 75 percent. Almost half the students finished their classes four weeks early. McIlquham says this sort of variable pacing and small-group instruction could become the norm.
Some observers have pointed out that while go-at-your-own-pace learning works well for some students, it can allow less motivated students to fall far behind. There are also questions about whether adaptive learning can extend beyond basic introductory classes, and whether it would work with less quantifiable, more intuitive subject matter, like literature and philosophy.
McIlquham thinks that adaptive education will free up teachers in any setting for more one-on-one instruction with students, and help them figure out which students need special attention. “Teachers are going to have so much more relevant information about their classes available to them,” says McIlquham. “As a teacher myself I’m excited about that. I’m much happier working with students on problem solving, critical thinking and issues they’re having, than standing up in front of a class and lecturing as if they’re all the same student.”
This article was originally published by NationSwell, a website dedicated to sharing the stories of innovative Americans who are working to effect social change and move the country forward. More from NationSwell:
These Kids Are Designing the Future—And 3D Printing It in Their Classroom
The Next Frontier in Online Education Isn’t What You’d Expect
The Minerva Project: On Online College to Rival the Ivy League
This Controversial Teaching Method Is Transforming Classrooms
Will Bringing Big Data into the Classroom Help Students Learn Better?
Brad McIlquham was tutoring at-risk youth in Durham, N.C., when a former co-worker gave him the educator’s equivalent of the Social Network pitch. What if, instead of teaching at most 50 kids a year, you could help bring personalized tutoring to 100,000, or a million kids? McIlquham’s co-worker, Jose Ferreira—who had taught SAT and GMAT prep with McIlquham at Kaplan—was proposing an upending of the traditional “teach to the middle” classroom model. When teachers instruct students of varying ability in the same class, some students get bored, while others struggle. And often, teachers don’t discover which students have failed to understand key concepts until their tests get graded. But by then, they’ve already fallen behind.