CAPPS - Avocacy and Communication Professional Development

California Association of Private Postsecondary Schools

Learning to Adapt

06/13/2014

Inside Higher Ed. June 13, 2014.

The term "adaptive learning" can mean many things, even to colleges that have begun experimenting with it.

A broad definition is the use of software -- often driven by data collection -- to create a more individualized learning experience for students. Experts say adaptive learning has promise in tailoring coursework and supports to students, which could boost retention and graduation rates.

The language around the technology “seems to be up for grabs,” said Peter Stokes, executive director of postsecondary innovation at Northeastern University’s College of Professional Studies.

The distinction between “adaptive” and “personalized” learning is unclear, he said. Even old-fashioned tutoring or extra-credit projects can be called personalized learning, as can data-driven courseware that morphs in response to each student.

The more aggressive forms of adaptive learning technology vary widely, each with multiple potential uses. Ultimately, colleges may use several overlapping versions of adaptive software simultaneously.

Larger institutions may have a leg up when it comes to with adaptive learning. Stokes said the technology holds the most promise when used across larger numbers of students. That way institutions can better see what works, and also get the most value out of their investment.

As a result, for-profit chains in particular appear to be giving adaptive learning a whirl. The Apollo Education Group, the American Public University System, Career Education Corporation and others have begun broad adaptive projects.

Part of personalized learning’s promise is that it “atomizes” or breaks down the professor’s role, said Stokes. The use of automated software to take on some of those tasks, such as tutoring, is controversial. Many in higher education are leery of using computers to do things that people did in the traditional college model.

But for-profits typically have more leeway for experimentation, given their corporate structures and less prominent voice for faculty in governance.

“The genetic character of the for-profits permits a mutation of the role of instruction,” Stokes said.

Not Just 'Big Data'

Apollo is “jumping in with both feet” on adaptive, said Alan Drimmer, the chief academic officer for the company, which owns the University of Phoenix. But Drimmer said Apollo is being cautious and methodical with its experiments. Part of that strategy is trying several different techniques.

“The spectrum of approaches is really important,” he said.

Three years ago the company spent $96 million to buy Carnegie Learning, a publisher of math curriculums, and related technology. Carnegie Learning’s adaptive math software, dubbed Cognitive Tutor, provides immediate feedback to students about skills they are trying to master. Apollo has funded the tutoring software’s continued development, and is using it in Phoenix math courses.   

The work is as laborious as it is expensive, Drimmer said. That’s because subject-matter experts must determine which concepts students should learn -- such as how to graph quadratic equations -- as well as the best ways to learn them. Those strategies are programmed into the software, which collects data on students’ actions and approaches to problems.

“We are definitely seeing positive results,” he said.

Apollo is currently developing a Cognitive Tutor application for English classes, with an eye toward helping students improve their grammar and writing.

In addition, Phoenix has used a different form of adaptive technology in automated personal study guides for 50 of its courses. Those “Knowledge Checks” steer students to materials to help them brush up on learning concepts. 

The American Public University System (APUS), a for-profit chain, also is well along in its work on adaptive learning. APUS has begun incorporating the technology into the instructional design of some of its courses, all of which are online.

One of the system’s more advanced experiments is the “semantic mapping” of coursework across its business school curriculum. University researchers created an open-source tool that searches for gaps between course content and learning goals at the course and program levels.

APUS’s work on adaptive learning isn’t just about “big data.” Phil Ice, the system’s vice president of research and development, said the system’s approach features a heavy dose of qualitative analysis, as opposed to just being focused on the quantitative, or data-driven aspects.

The system uses learning analytics to ask which course content is working and whether it is helping retention, Ice said.

Eventually APUS hopes to use a range or “stack” of adaptive software, he said, perhaps even tapping outside vendors. But getting there will be complicated.

“We’re past the point where the cool stuff is going to come quickly,” he said of adaptive learning, adding that the various tools are “so difficult to piece together.”

Next Steps

Knewton is the most high-profile of adaptive learning companies.

The company blends data science, psychometrics and machine-learning elements to create a platform for “personalization at massive scale.” Perhaps most notably, Knewton collaborates with Arizona State University in a broad effort to “power” university courses with adaptive software.

Knewton has also partnered with publishers such as Pearson and Cengage, which are embedding Knewton in products that traditional colleges use. Likewise, McGraw-Hill Education incorporates its LearnSmart adaptive platform into e-textbooks.

Jose Ferreira, Knewton’s founder and CEO, said it is “incredibly expensive” to create adaptive engines.

“They’ve got to be gigantic” to work well, he said. “We lose money on all our direct relationships.”

The company has competition, however. Education Growth Advisors, a consulting firm, last year released a study that analyzed adaptive offerings from 70 companies. The report, which the Bill & Melinda Gates Foundation funded, included case studies on a range of providers, including Cerego, LoudCloud Systems and Smart Sparrow, among others.

Two of the earliest and most commonly used adaptive learning approaches in higher education are Pearson’s MyMathLab and McGraw-Hill’s ALEKS. The computerized learning systems respond to individual students as they progress through mathematics coursework, seeking to determine what students know and have yet to master.

Ferreira is skeptical that any college could create an adaptive platform like Knewton’s. That would cost well over $100 million, he said. Apollo, however, has spent close to $100 million on its Cognitive Tutor.

A large-scale adaptive project at a for-profit also could feature technology from Knewton or a similar company, such as Ireland’s CCKF, which is working with Career Education. Those sorts of adaptive learning tools can be add-ons to what Apollo or APUS are doing, lending another, deeper layer of data analysis to courseware.

Later this year Knewton plans to release a more accessible platform. Ferreira said colleges and even individual professors will be able to use it, which is a departure from the sort of broad partnership the company has built with Arizona State.

“There will be ways for individual institutions to work with us more directly,” he said. “We haven’t made it very easy yet.”

Ice said adaptive learning would hit a “tipping point” when Knewton-style software can work in conjunction with the sort of content-specific adaptive coursework that APUS is building.

For now, however, both Ice and Drimmer said figuring out how to use adaptive learning is a long slog with an uncertain conclusion -- but also one worth trying.

“If you keep working, it’s going to show results,” said Drimmer.