Hoping to bring its adaptive technology directly to students, teachers, and creators of “open” educational content, ed-tech company Knewton announced Wednesday the launch of a new online platform that relies on big data to match individual learners to the specific instructional materials that are ostensibly best-suited to their own learning needs and styles.
The announcement marks the most aggressive attempt to date to merge such algorithm-based personalization—increasingly popular among both assessment makers and educational publishers—to the burgeoning world of open educational resources, or OER.
In a statement, Knewton founder and CEO Jose Ferreira touted the new effort.
“Knewton plucks the perfect bits of content for you from the cloud and assembles them according to the ideal learning strategy for you, as determined by the combined data power of millions of other students,” he said. “Think of it as a friendly robot tutor in the sky.”
Leaders in the OER community expressed enthusiasm about the general possibility of making “open” material (which is licensed in such a way as to be available for free use, revision, and sharing) adaptive. They believe the type of advanced learning analytics pioneered by companies such as Knewton might help solve one of their field’s central challenges: how to quickly surface the best content from among the millions of open educational resources now housed in numerous content repositories and countless individual websites.
But it remains to be seen if educators, the public, or the OER content creators will embrace the specifics of Knewton’s approach.
Many remain skeptical of big-data and algorithm-driven education, especially in an era of heightened concern around student data privacy.
Educators may also be skeptical of Knewton’s apparent “presumption that content alone can help students master skills and concepts,” said Cable T. Green, the director of global learning for Creative Commons, a nonprofit that encourages the use and sharing of open tools and information.
And, Green added, the “open” community values transparency across the board. That could mean that Knewton’s “closed” algorithms and statistical techniques will be a significant turnoff for many OER providers.
“There’s a red flag raised by any company making claims about the value or effectiveness of learning resources without allowing the public to see what those claims are based on,” he said.
How the algorithms work
Knewton was founded in 2008 by Ferreira, a former executive at test-prep giant Kaplan, derivatives trader at Wall Street powerhouse Goldman Sachs, and strategist for John Kerry’s unsuccessful 2004 presidential campaign.
The company says it has since delivered billions of educational-content recommendations to almost 10 million students in 20 countries, across both K-12 and higher education.
But to date, Knewton’s analytics engine has “powered” only the proprietary content of the educational publishers with which it partners, including heavy-hitters such as Pearson and Houghton Mifflin Harcourt.
In an interview earlier this month at Knewton’s New York City offices, Ferreira said the company’s goal has always been to provide its technology directly to individual users.
Here’s how the company’s new platform works:
Students and other users can sign up directly at knewton.com. After a brief registration process, they can select the subject area, topic, and specific skills they are interested in, then click a button that says “start learning.” Immediately, the user will be introduced to content that is algorithmically selected to diagnose what he or she already knows. As soon as the student begins interacting with that content, Knewton begins harvesting reams of data and comparing that information to the data already collected on the company’s millions of other users. Through a mind-boggling series of probability calculations, Knewton’s algorithms then begin determining which content is most likely to help the user learn what he or she has not yet mastered in the most efficient, engaging way possible.
At the same time, teachers and other creators of instructional materials can upload their lessons, YouTube videos, and the like to the site for free. Teachers might assign their own content to their own students, Ferreira said, or the Knewton platform might assign a new resource to a small initial group of carefully selected users. By comparing how well those users learn with that content to how well other users learn the same skills and concepts with different content, Knewton’s algorithms quickly make determinations on the content’s effectiveness, ability to engage students, and capacity to create valuable data for future use.
Knewton officials boast that their algorithms and statistical techniques are light years beyond those used by other companies in the adaptive-learning field. They also claim that their technology is far better than humans, even experts, at gauging how a wide variety of students are likely to perform with any given piece of content.
That process of statistically “norming and calibrating” content, Ferreria said, is akin to the huge investment that standardized test makers put into developing questions that might appear on an exam such as the SAT.
The end result, he said, is that users get a “free personal tutor” that “can practically read your mind.”
Privacy, over-reliance on data science are concerns
A long list of potential concerns about Knewton’s new platform highlights the challenge of making OER truly adaptive, however.
The company has already received intense scrutiny from privacy advocates over its collection of millions of data points on individual students for creation of extensive “learner profiles.”
And that was in the context of Knewton’s enterprise business with publishers, through which company officials say they don’t collect or maintain any “personally identifiable information” on individual students. The new consumer-facing platform, on the other hand, will require students or their parents to create an account specific to their child and agree to allow their child’s information to be used to improve Knewton’s educational services.
The power of Knewton’s technology could also be limited by federal and state legislation that proposes to place strict limits on how long vendors can retain data on individual students. Company officials say the more you use their new platform, the “smarter” it will become, because it will have ever-more information with which to work.
“It’s madness to want to throw all that data away,” Ferreira said. “Legislation should take into account the probability of that a lot of students won’t want their [information] destroyed.”
But whether consumers are in reality quite so ready to embrace Knewton’s technology remains an open question.
Similarly, leaders in the OER field also expressed hesitance about adaptive technology that relies almost entirely on statistical, rather than human, determinations about the quality of educational content.
Jennifer A. Wolfe, for example, is a partner with the Learning Accelerator, a nonprofit that has helped lead a 12-state effort to procure high-quality OER for classroom use. As part of that initiative’s work, more than two dozen state- and local academic content experts reviewed every unit submitted by approved bidders.
“I’m not a data scientist,” Wolfe said, “but that kind of vetting process by experts just gives me more comfort.”
And in addition to urging Knewton to make its algorithms and statistical models open to inspection by people outside the company, Green, of Creative Commons, also expressed concern over what he described as the company’s narrow focus on educational content alone.
“Can Knewton and its recommendations be useful? Absolutely,” Green said. “But interaction solely with content has a limited effect compared to interaction with content, teachers, peers, and the web.”
Crafting a business model
Knewton, of course, believes that such challenges are either overblown or surmountable.
The company won’t sell or share student data, officials say, pointing out that Knewton is among the original signatories to the industry-led Student Privacy Pledge, a volunteer commitment to secure students’ information and protect their privacy.
When it comes to vetting content, Ferreira said, statistical analysis is actually far more reliable than “expert judgment,” which “has an uninterrupted track record of failure.”
And as for the role of teachers and peer interactions in student learning?
“We don’t claim to replace those things; we wouldn’t want to,” chief research officer David Kuntz wrote in an email. “Right now, Knewton is focused on improving the content side: by using the data generated by students as they interact with content to determine what’s the right material to get to each student, at the right time and in the right manner.”
The company’s new platform launched with nearly 100,000 pieces of content, the majority of which were curated by Knewton staff. Whether teachers, OER creators, and others volunteer to build on that foundation with their own lessons will go a long way to determining the product’s success.
Similarly, it will be a challenge to get students and teachers to choose Knewton from among the many competing online content sites that already exist, some of which have established track records and faithful followings.
And ultimately, the company will have to figure out how to make money off a service that officials swear will remain free for students and teachers. The current plan involves selling licenses to tutoring, test-prep, and corporate-education programs.
Despite their concerns, both Wolfe and Green praised Knewton for taking a leading role in bringing adaptive technology to OER, a move they described as “exciting” and “positive.”
And for his part, Ferreira, ever-brash, voiced no doubt that both the OER and consumer markets will learn to embrace what Knewton is offering, even if hesitation currently abounds.
“We think that long-term, people who build awesome things get rewarded for it,” he said.
Photo of Jose Ferreira courtesy of Knewton.
A version of this news article first appeared in the Digital Education blog.