Teaching and Learning | Feature

The Great Adaptive Learning Experiment

• By John K. Waters
• 04/16/14

Higher education is in the midst of a kind of Renaissance. A flurry of activity and experimentation around adaptive learning is taking place on college and university campuses, thanks to a high-profile, targeted grant program from the Bill and Melinda Gates Foundation; the relatively recent emergence of sophisticated adaptive learning software and platforms; and nascent partnerships among schools and learning content publishers. Institutions around the world are engaged in serious explorations of the potential of an approach to instruction and remediation that uses technology and accumulated data to provide customized program adjustments based on an individual student's level of demonstrated mastery.

That last sentence contains a pretty good definition of "adaptive learning," but the term is so often used synonymously with "personalization" or "personalized learning" that it's reasonable to ask, What's the difference?

"Personalized learning is really an umbrella term," said Adam Newman, founding partner of Education Growth Advisors (EGA). "It's a way in which learning is modified, adjusted or customized to meet an individual learner's needs and objectives that does not depend on adaptivity. So you can think of adaptive learning models as one approach along a spectrum that enables personalization."

That umbrella term covers a range of approaches and models, Newman said, including competency-based learning, differentiated instruction and tutorial models, as well as adaptive learning.

EGA is a strategic advisory and consulting firm and an investment bank focused exclusively on the education sector, founded by Newman and Christopher Curran in 2010. The company provides research-based market analysis to the K-12, postsecondary, corporate and lifelong-learning sectors, and drew a lot of attention recently with the publication of two white papers funded by the Gates Foundation: "Learning to Adapt: A Case for Accelerating Adaptive Learning in Higher Education" and "Learning to Adapt: Understanding the Adaptive Learning Supplier Landscape." In the first paper, the EGA analysts -- Newman among them -- explained how adaptive learning solutions could break the "Iron Triangle" of cost, quality and access in higher ed. In the second paper, the analysts provided a snapshot of the current crop of adaptive learning solutions.

They also offered some usefully specific definitions: Personalized learning, they wrote, is a "pedagogical method or process that draws on observation to inform tailored student educational interventions designed to increase the likelihood of learner success." Technology isn't strictly required for personalization; a professor personalizes a student's experience, for example, when she takes him aside and recommends extra reading. Adding the tech makes it possible to personalize at scale. On the other hand, adaptive learning is enabled by technology. It takes "a sophisticated, data-driven and in some cases, nonlinear approach to instruction and remediation, adjusting to a learner's interactions and demonstrated performance level, and subsequently anticipating what types of content and resources learners need at a specific point in time to make progress."

An adaptive learning system can be "facilitator-driven," which means it provides instructors with actionable student and cohort profiles, typically via a dashboard; or it can be "assessment-driven," which means it provides close-to-real-time ("dynamic") adjustments of the instructional content that lay out a student's learning path. Facilitator-driven systems provide information instructors can (and have to) act upon; assessment-driven systems make their own adjustments and allow students to move through a course individually or in a group. The two approaches are not mutually exclusive, Newman said, and both might be found in a single product or system offering.

For Dale Johnson, senior business analyst at Arizona State University and manager of the school's Adaptive General Education Program, both personalization and adaptive learning describe a tech-enabled continuum of responsiveness to the learning needs of individual students.

"We don't really distinguish between personalized and adaptive," he said. "I think it's a bit of a semantic game in the context of the marketplace. There are a lot of vendors out there that are trying to position their products in different ways, so they go for hair splitting around terminology."

Early Pilots

Johnson has been taking a practical look at adaptive learning approaches and systems since 2010, about the time ASU began working with adaptive learning platform provider Knewton to redesign a developmental math program. Launched in fall 2011, the Knewton Math Readiness program was built on the Knewton platform and aligned to the Common Core Standards for mathematics. The system was both facilitator-driven and assessment-driven. It used student data to "figure out what a student knows and how they learn best," and then "recommend what concept in the course each student should study next," Knewton explained in a statement. But it also provided instructors with real-time reports that allowed them to "detect gaps in knowledge, create adaptive study plans for each student and focus lessons around concepts where students need the most help."

The Gates Foundation cited the preliminary findings from that ASU/Knewton project in 2013 when it launched its Adaptive Learning Market Acceleration Program (ALMAP), pointing to a reported 18 percent increase in pass rates and a 47 percent drop in student withdrawals. At the time of that launch, the ASU program had helped the university to hold onto $12,000,000 in what would have been lost tuition.

The objective of ALMAP, the Foundation has stated, is to get the adaptive ball rolling in higher ed by encouraging "both rigorous research and acceleration of the adaptive learning marketplace, and to showcase important successes and innovations." Last spring the Foundation awarded 14 grants to accredited U.S. colleges and universities with the most promising adaptive learning proposals. The winning schools are required to use the grant money to create vendor partnerships and launch adaptive courses over a 24-month period. The Foundation hopes the project will yield 10 "exemplar implementations" of adaptive courses.

ASU is one of those schools. Working again with Knewton and that company's partner, education services company Pearson, the university is creating what Johnson described as an ecosystem that combines adaptive learning technology; reporting and tracking features; and a grouping capability that supports a flipped classroom by allowing small-group work among students. This last component, Johnson said, is unique to ASU.

"One of the benefits of adaptive learning is that it frees up faculty members to spend more time with students, to work with them in small groups and individually, essentially flipping the classroom," Johnson said. "We see this [approach] as part of a much bigger pedagogical picture. The technology is a tool that gets us to our goal of rehumanizing our large classes."

ASU's Adaptive General Education Program is still in development. When it's launched this fall, it will offer beginning courses in biology, chemistry, physics, economics and psychology. The school is building on its experience with the Knewton Math Readiness program, Johnson said.

"What we're doing now is what I would call continuous improvement," he explained. "We're evolving to new disciplines and continuing to improve our math courses as we learn what works and doesn't."

ASU's Tempe, AZ, neighbor, Rio Salado College, is also dipping its toe into the adaptive learning pool via the Gates program. Also working with Knewton/Pearson, the college has developed an adaptive learning pilot program based on Pearson's MyWritingLab. The program will provide two classes aimed at developmental learners in a fully online learning platform for writing.

The system was developed for Rio Salado from a set of competencies and learning outcomes the school assembled and provided to the platform and content vendors. The platform uses these goals as it responds to a student and builds his or her learning path.

The three-semester pilot officially launched in January, but the school implemented a pre-pilot program in November of last year. "Adaptive learning technology is very new to many of us," said Julie Stiak, associate dean of instruction and community development at the college. "The Gates Foundation was very innovative in this regard. They gave us a chance to do this little pilot, which allowed our instructors and designers a chance to work out a better course offering."

"It helped everybody get their feet wet," added Cynthia Maxson, who teaches the two English classes.

About 50 students participated in that pre-pilot, and 200 students are participating in the first semester of the official pilot. Approximately 100 of those students are enrolled in the adaptive program; the school is also monitoring a control group of 100 traditional online learners who won't have access to that class.

"This is a true research project," said Stiak. "That's how we see it, and that's how we're conducting it."

Stiak sees great potential in the adaptive model, and believes the combination of technologies -- the educational content layer on top of an adaptive platform -- is likely to yield positive results. She pointed to the Knewton platform's ability to map the relationships among individual concepts that make up the learning content, and then integrate the preferred taxonomies, learning objectives and student interactions. Knewton refers to this capability as "adaptive ontology."

"Their concept map approach allows the system to constantly personalize the learning and adapt as the student continues interacting with the system," Stiak said. "If we can get this right, it'll be invaluable."

An Adaptive Partner Ecosystem When Arizona State University first worked with Knewton to redesign a developmental math program, the vendor took on all the software development, recalled Dale Johnson, senior business analyst and manager of the school's Adaptive General Education Program. But through partnerships and a shift in its business model, Knewton is now more of a provider of application programming interfaces (APIs) for third-party content specialists. The content of ASU's Adaptive General Education Program, for example, is being developed by Pearson on top of the adaptive capabilities of the Knewton platform. Knewton has been partnering with Pearson since 2011. "Our partners are the experts in their target market," noted David Kuntz, vice president of research and adaptive learning at Knewton. "They create the application and pass us the data. We process that data and make a set of actionable inferences about the students, and then pass those back to the application, and the partner decides how and when to render those for the student." The Knewton platform collects and processes data from real-time streams and maps the relationships among individual concepts within the learning content, which are then integrated into taxonomies, learning objectives and student interactions. It then uses this data to evaluate student proficiencies and generates "insights" and predictions that lead to recommendations and, ultimately, an individual learning path. The list of Knewton's content partners has grown over the past couple of years and now includes (along with Pearson) Cambridge University Press, Macmillan, Wiley, Cengage Learning, Gutenberg Technology, Lelivrescolaire and Houghton Mifflin Harcourt, among others. In 2014, the company added Microsoft to that list. The Redmond software giant has stated that it will make the Knewton API available to its "vast partner and publisher ecosystem."

A Diverse Marketplace

Of course, Knewton isn't the only adaptive learning vendor taking a platform approach to this market -- and adaptive efforts in higher ed aren't limited to the U.S. CogBooks, for example, is a small, Scotland-based startup with an adaptive learning platform it describes as a "micro-adaptive, intelligent, sequence-driven learning system." The platform is "micro-adaptive" because it's designed to adapt to individual learners, rather than groups, according to the company. The platform's intelligence, which is derived from sophisticated algorithms and machine learning methods, identifies the ideal learning path (or sequence) for the learner at multiple steps through the material.

The CogBooks platform also supports an adaptive approach that takes the student on a pre-programmed path in response to his or her input or pre-test results. The system also incorporates consideration of the best learning sequences, based on a student's responses, knowledge profile, learning-style preferences and even context, such as the learner's industry or profession.

CogBooks recently announced a collaboration with the Northeastern University College of Professional Studies (MA) in Boston and Shoreline Community College (WA) near Seattle on an adaptive learning pilot program that will also be funded and guided by ALMAP. The pilot will focus on developing adaptive classes online.

Another adaptive platform provider, LoudCloud Systems (not to be confused with the software-as-a-service company that became Opsware in 2002), launched in 2010 with what was essentially an adaptivity-enhanced learning management system. The company CEO, Manoj Kutty, still refers to the company as a learning platform provider, but that initial LMS offering has evolved into more of an adaptive learning ecosystem for higher education and K-12. The company describes it as "a behavioral-analytics-based teaching and learning platform designed to deliver personalized pathways in education."

Adaptivity is provided primarily through the company's LoudBooks e-reader platform. The e-reader creates a personalized environment -- a "space" in which students and professors can interact with content and with each other. The platform analyzes students' individual learning styles, and then automatically provides a curated set of personalized content. Adaptive tools help students understand what they're learning and offer supplemental content if needed.

Analytics provided by the company's LoudSight dashboard allow faculty and administrators to further personalize the learning experience. The dashboard presents analyses of key behavioral and performance metrics gathered by the system, which identifies "vital predictors of success" to which educators can respond. The company calls it "intelligent feedback at the point of instruction."

LoudCloud is also taking steps into K-12 with a system that taps into data in existing systems within school districts, analyzes that data and then provides recommendations to teachers via a dashboard, explained Harish Joshy, who manages the company's K-12 business. "We find patterns and provide teachers with predictive analytics that allow them to take the next step," he said.

Learning by Doing

Australia-based Smart Sparrow is another up-and-coming adaptive platform provider, and the company's founder and CEO, Dror Ben-Naim, doesn't believe that those who insist on precise definitions in this space are splitting hairs.

"Many of the so-called ‘adaptive learning' platforms are really more like content recommendation systems -- like Amazon or Netflix," he said. "I don't see where the learning is adaptive. The content is not changing in response to the students."

Ben-Naim described Smart Sparrow's Adaptive eLearning Platform as "rich interactive and adaptive" (his emphasis). The platform is designed to be used by instructors to create adaptive courseware that responds to students by providing feedback unique to the individual, which the company calls "adaptive feedback." It then presents differentiated learning sequences for each student based on his or her interactions with the courseware, creating what the company calls an "adaptive pathway."

At the heart of this approach is a belief in the power of learning by doing.

"If you actually look at the educational content that's out there, what you'll see is modalities that are a bit 20th century," Ben-Naim said. "You've got something to read, something to watch and a few multiple choice questions to answer. I think we can innovate there."

Learning by doing is intrinsically adaptive, Ben-Naim argued, because the system that supports it must respond in different ways, based on what the student is doing. Such a system also lends itself to simulative learning, he said.

"Let's say you want to teach engineers how to build a bridge," he said. "Do you want them to read something, watch something and then answer a multiple choice quiz, or do you want them to build a bridge with a simulator that gives them specific feedback and specific activity based on that?"

Smart Sparrow also bucks a trend that has been gaining ascendance in higher ed for some time now, by pulling the teacher back into the center of the educational process.

"I'm a big believer that we should not take the teacher away from the teaching process," he says. "We should build the technology around the teachers to empower them and put them at the center of the story. We want them to be able to leverage the technology, rather than being replaced by it."

Smart Sparrow grew out of years of research at the University of New South Wales in Sydney, in the school's Adaptive eLearning Research Group, which was led by Ben-Naim. The academic project spun off into a commercial entity in 2010. Most of the implementations of the Smart Sparrow platform have been in Australia, but Ben-Naim, who is now based in San Francisco, expects to change that soon. UNSW is currently collaborating on a project with ASU and other universities to build rich interactive and adaptive online courseware for high school students. The project will be led by professors and will "bridge the gap between the science lab and the science school," Ben-Naim said.

Iain Martin, vice president and deputy vice chancellor (academic) at UNSW, believes that the seemingly sudden flurry of activity around adaptive learning is the result of "a confluence of factors."

"The technology is now cheap enough and powerful enough for this kind of approach to be applied effectively and widely," Martin said. "There's much more of an acceptance among students and other consumers of education of online education -- it's no longer seen as something that is intrinsically inferior [to campus-based education]. And every college system everywhere is grappling with how to deliver high quality education, which inherently requires a degree of personalization, but can't be delivered in a scalable way without technology."