Test-prep company Knewton takes online courses to next level: university

Knewton, a company that provides personalized help to boost scores on tests such as the GMAT, is taking online education to the next level: It will now power actual university math programs for Arizona State University (ASU) students.

The announcement marks a shift for the New York based startup company, which to date had only provided test-prep programs. ASU’s decision to use online, instructorless remedial education raises the question of just how much new technologies could disrupt the traditional university model.

ASU students who score below a certain threshold on the math portions of a preliminary assessment will be required to use the Knewton-powered adaptive-learning program. The web-based program will generate homework assignments based on each student’s individual proficiency levels and learning styles, and adapts as students score better in a certain type of problem. Based on the data from the online program, ASU will also provide virtual and in-person tutoring. Once students demonstrate college readiness in mathematics, they will advance into ASU instructor-led math courses. The university hopes this will boost retention and graduation rates.

In addition to the remediation program, ASU’s two introductory math courses will also incorporate Knewton. Course professors can assign individualized homework through Knewton’s platform, such that a certain percentage of incorrect answers on a preliminary problem set will trigger additional problems catered to a student’s personal weaknesses.

Knewton’s move into higher education core curriculum is one that CEO Jose Ferreira tells us was in the business plan from the beginning. It’s a step along the path to launching an open, adaptive-learning platform that Ferreira hopes will be used by students and content producers of any subject. Such a platform could provide individuals with an educational profile that tracks their proficiency levels and learning styles across disciplines, not unlike the way Facebook tracks a person’s ever-changing social profile. The platform could also be used as a discovery engine for the most effective curriculum. “We can algorithmically determine who has the top performing content,” Ferreira says. “If 10 people create the best content on how to convert fractions to numerals … we can psychometrically determine what is most effective.”

Ferreira’s ambitions are big, but not everyone thinks improving education is a matter of data and algorithms.

Karin Forssell, Program Director of the Learning, Design, and Technology Masters Program at Stanford University says adaptive learning engines might prove less effective outside of clear-cut subjects like math. “The bigger question is, what happens when you move from the disciplines where there are “right” answers (traditionally more lecture- and test-based) to those in which argument and interpretation are key (generally more discussion- and essay-based),” she wrote in an email to VentureBeat. “Assessing those ‘fuzzy’ subjects by computer is extremely challenging.”

Ferreira acknowledges that not every subject can be taught with adaptive-learning engines, but he says only the most abstract like philosophy and complex law fall out of its purview. And the usefulness isn’t limited to hard sciences, he insists, as is evidenced by the fact that ASU recently asked Knewton for reading and writing programs as well. “If people can agree on what proficiency means, it can work with Knewton,” he says.

Jennifer Carolan, associate partner at New Schools Venture Fund (not a Knewton investor) and former high-school history teacher, says adaptive-learning technology has promise in the humanities. “Great teachers break down their subjects into learning bits which can become the basis for adaptive learning engines,” she wrote in an email to VentureBeat. “While it’s true that math and the hard sciences are more conducive to this format, there is still a lot of learning bits living in the heads of great English and history teachers that could populate adaptive learning engines in the humanities and facilitate faster learning there as well.”

Knewton is backed by a long list of angel investors, including Ron Conway, Reid Hoffman and Chris Dixon, and has raised a total of $21 million since its first round of financing in 2008. The company, which was named a 2011 Technology Pioneer by the World Economic Forum at Davos, aims to launch a first version of its open API for programmers in March or August and hopes to have a version that is usable by any teacher or content creator by some time in 2012.