Should Personalization Be the Future of Learning?

Benjamin Riley and Alex Hernandez square off



By and 07/03/2015

5 Comments | Print | NO PDF |

As technology has continued to permeate the classroom, the concept of personalized learning has gained traction among educators and policymakers. In this forum, our authors take a step back to ask a critical question: should personalization be the future of K–12 schooling? What are the risks? What do students stand to gain? Can personalized learning accomplish what we might hope? Benjamin Riley, founder of Deans for Impact, makes the case for an abundance of caution, while Alex Hernandez, a partner at Charter School Growth Fund, supports continued efforts to get personalization right.


ednext-april15-forum-riley-smallPersonalized learning is not the solution

by Benjamin Riley

Sprinkle the phrase “personalized learning” into virtually any conversation or speech regarding education, and you’ll see heads nodding in happy agreement. Although some might view this as evidence of merit, I suspect that the personalization concept has become an empty vessel into which one may pour any number of competing theories or policies.

What many people mean by “personalized learning” is using technology to give students more control over their education experience. “Blended learning involves leveraging the Internet to afford each student a more personalized learning experience, meaning increased student control over the time, place, path, and/or pace of his or her learning,” declares the Clayton Christensen Institute for Disruptive Innovation. We must “empower learners to learn any time, any place, and at any pace, both in school and beyond,” according to a report from the Aspen Institute, “Learner at the Center of a Networked World.” “Instead of organizing students by age and giving them all the same lesson, [students may] initiate their own learning, follow different paths, and seek varied resources to help them meet their goals,” says my friend Alex Hernandez.

There are a number of assumptions and hypotheses nested in these definitions, but I want to focus on two: 1) students will learn more if they have more power over what they learn (“the path argument”), and 2) students will learn more if they have more power over when and how quickly they learn (“the pace argument”).

The problem with the path argument is that it runs afoul of our current understanding of cognition. Put simply, knowledge is cumulative. What a child is capable of learning depends on what she already knows. When a child encounters new information, if she lacks the preexisting knowledge to put the information in context, she will quickly become frustrated. She won’t learn. So to the extent that personalization seeks to devolve responsibility for acquiring new knowledge to students, it mistakenly assumes that many or most students are properly equipped to make sense of new information.

A large body of research shows that not all learners prefer or profit from controlling these tasks and that forcing control on them can be counterproductive. Many learners lack the capacity to appraise the demands of a task and their own learning needs. In other words, learners often regulate their learning poorly, exerting control in a misguided or counterproductive fashion and thus fail to achieve the desired result.

Of course, as Dan Willingham, a cognitive scientist and author of Why Don’t Students Like School? told me, “It’s possible to imagine versions of the path that are cumulative” and tailored to an individual student’s ability or interests. But students are unlikely to find this path if they themselves “pick what to study next.” I am of the view that we created the profession we call “teaching” largely to solve this problem. Students need to be guided down the path of their learning. Teachers should remain central to the activity of imparting knowledge to students.

The pace argument also contradicts a key insight from cognitive science: our minds are not built to think. In fact, our brains are largely oriented to avoid thinking. Thinking is hard and often not fun, at least at first. As a result, we will naturally gravitate away from learning activities that we find hard and unpleasant. Of course, this doesn’t mean we’ll always avoid thinking about challenging subjects, as Willingham cautioned me. But the challenge for people who support personalization is to demonstrate that children will use technology to engage in difficult activities when they have other options that are less challenging yet more immediately enjoyable.

For the same reason, I think it’s a mistake to place children in charge of the speed of their learning, particularly during the early years of their education. Many kids, and particularly those at risk, will choose a relatively slow velocity of learning (again, because thinking is hard). A slow pace will lead to large knowledge deficits among some students compared to the knowledge assets of their faster-moving peers, which will cause the first group to slow down further, until eventually they “switch off” from school. The only way to prevent this slow downward spiral for these students is to push them harder and faster. But they need to be pushed, which means we should not cede to them control of the pace of their learning.

Am I suggesting that we return to the “factory model” of education? If factory model implies the dry recitation of facts to students, no, I am not. But to the extent that the stereotype represents what’s actually happening in classrooms (which I’m skeptical of), the problem is not the seating arrangement or lack of smartphones; it’s the pedagogy. Effective instruction requires understanding the varying cognitive abilities of students and finding ways to impart knowledge in light of that variation. If you want to call that “personalization,” fine, but we might also just call it “good teaching.” And good teaching can be done in classrooms with students sitting in desks in rows, holding pencil and paper, or it can be done in classrooms with students sitting in beanbags holding iPads and Chromebooks. Whatever the learning environment, the teacher should be responsible for the core delivery of instruction.

Am I against technology in schools? Not at all. There are many ways in which technology can improve education, as schools and educators are demonstrating worldwide. Here’s one example that strikes me as particularly interesting: using a tablet with a screencasting app, teachers can record their students grappling with a problem and reflect on what led to their understanding (or failure to understand). And the ubiquitous use of technology will make our schools rich with almost real-time data on student learning that can be quickly analyzed and acted upon in beneficial ways, provided appropriate privacy safeguards are in place.

But technology can also hinder education. I’ve visited technology-rich classrooms where virtually no learning appeared to be taking place. The fact that students make use of electronic devices does not make them good users of the media they have at their disposal. They can Google but lack the skills to find the information they need or to assess the relevance or truth of what they find. This leads to essays on Baconian science with texts about the 20th-century British artist Francis Bacon and on the problems that Martin Luther King had with Pope Leo X and Holy Roman Emperor Charles V.

Just to reiterate: my argument is not against technology. My concern is that many personalization enthusiasts support a theory of instruction that is difficult to harmonize with our current understanding of how the mind works. If “children are more alike than different in how they think and learn,” as Willingham argues, then one of the major pillars underlying the push for personalization crumbles. We should adopt a more skeptical posture regarding some of the claims about the benefits of personalizing learning and shift the burden to its proponents to explain how it will improve student learning in the face of scientific evidence to the contrary.


ednext-april15-forum-hernandez-smallUnderstanding the promise of personalized learning

by Alex Hernandez

Personalized learning theory is built on the twin pillars of 1) differentiated learning pathways for students and 2) feedback that enables students to make informed judgments about what they’ve learned, how well they’ve learned it, and what to learn next. The importance of these two pillars for effective education is well established, yet traditional schools struggle mightily with both, mostly because there are only 24 hours in a day and educators are human.

Built on these two pillars, personalized learning has the potential to fulfill some of the most basic hopes families have for their children: “I just want my child to get what they need, when they need it.” “I want my child to grow and develop from where she is today.” “I want my child to experience success and grow in confidence as a learner.”

Differentiated learning paths honor student variation in both background knowledge and ability. Effective teachers assign work to students that is “appropriate to their current levels of competence.” But most teachers’ ability to manage multiple learning paths in multiple subjects is limited, at best. Arbitrary, age-based academic standards and fixed pacing guides only exacerbate the problem.

Substantially reorganizing school to honor the variation in student readiness is an option that educators only infrequently employ. At my children’s elementary school, the entire student body learns math at the same time, and students across the school are regrouped based on their needs, not their grade or age. I’m not arguing that this practice is the only answer to our problems, but it is clearly much easier to say “we differentiate for every child” than to actually put differentiation into practice.

Feedback has a powerful impact on student achievement, and providing it is entirely within the school’s control. In traditional classrooms, teachers are the bottleneck in giving student feedback unless there are other feedback loops students can access directly. In personalized learning environments, students theoretically have access to ample, frequent, and actionable feedback from multiple sources, including content, peers, and teachers. Teachers can focus their energies on 1) providing the feedback that only they can provide and 2) making sense of the feedback generated by other sources.

Together, differentiated learning paths and feedback create the basic conditions for personalized learning to occur. In addition, students need to own their learning. Every student should have the opportunity to go to college, where students are primarily responsible for their learning. This lies in stark contrast to K–12, where teachers are primarily responsible for student learning. Ensuring college readiness requires improving executive function and ramping up background knowledge, neither of which are core competencies in today’s schools. It is unrealistic to assume that in a traditional school, students will receive a seven-step lesson for every chunk of background knowledge they’ll need to succeed beyond the 4th grade. We need to organize our schools in such a way that students acquire background knowledge, accrue expertise, and develop independent learning skills.

Among the issues frequently raised by skeptics are that the research base for student-driven learning is abysmal. Individualized instruction has shown particularly weak effects on student achievement as have other methods that put the teacher in the role of facilitator. I suspect insufficient attention to background knowledge is the Achilles heel of those who believe teachers should be the “guide on the side.” I remain optimistic nonetheless, for two reasons. First, now that I have a deeper appreciation for what it takes to develop stable personalized-learning environments, I’m skeptical that prior efforts had the resources, capacity, or technology necessary for those initiatives to succeed. Second, the new wave of personalized learning draws on a set of instructional strategies that have shown particularly large effects on student achievement: feedback, peer tutoring, mastery learning, goal setting, and even direct instruction.

A second contention is that students are incapable of making good decisions about their learning. I agree that many students, without any structures, supports, or feedback, will make poor decisions about their education. But schools that move to student-driven environments create supports so students can function well. These include increased coaching, goal setting, feedback on progress, tiered supports, peer tutoring, small-group instruction, and other resources that together enable a greater percentage of students to manage their own learning effectively. For example, the first thing most schools do is put in a “minimum pace” to help address the “slow velocity” problem.

Opponents argue that students will avoid hard thinking in personalized environments without teachers. I generally agree that teachers are best positioned to lead cognitively challenging activities like Socratic seminars, deep reading, and math talk. Personalized learning environments are better suited to teach basic skills and background knowledge than to teach critical thinking. That said, students with significant background knowledge are capable of hard thinking when they take more ownership of their learning, and we need to honor those capabilities.

Similarly, opponents assert that personalized learning precludes great pedagogy like argument, discussion, and debate. This is a false choice. Students can engage in personalized learning for a portion of the day and spend the rest of their time in rich learning activities that only teachers can provide. The bet here is that if students can drive their development of basic skills and background knowledge, teachers can “trade up” and focus their energies on challenging tasks and compelling experiences. Teachers should be in the business of creating “aha” moments for children, not figuring out seven different math lessons for 25 different students.

Ben Riley argues that one of the important roles of the teacher is to “pick what to study next,” and Dan Willingham acknowledges that “craft knowledge trumps science” when it comes to differentiation. But clinical diagnoses have extraordinarily high error rates. In his book Thinking, Fast and Slow, Daniel Kahneman cites a study in which experienced radiologists contradicted themselves 20 percent of the time when they saw the same chest x-ray on separate occasions. A recent Johns Hopkins study found that 28 percent of ICU patients had at least one missed diagnosis when they died. It is not unreasonable to believe that error rates are higher in education because of the lack of reliable heuristics and a relatively thin research base. This is not teacher bashing. If education is anything like other professions that rely on clinical judgment, it is likely that one in four education diagnoses (probably more) is incorrect. Students are best served when they have access to both expert judgment and the types of algorithmic supports possible in personalized learning environments.

I may be overly optimistic about the promise of personalized learning, but I think critics overestimate the effectiveness of the average traditional classroom.


ednext-april15-forum-riley-smallBen Riley responds

No one can reasonably dispute that students vary in their cognitive abilities. The question is whether we should be orienting our pedagogical practices primarily around these differences, as I think most personalized-learning supporters would urge, or instead take note of the many ways in which students are cognitively similar, and make these shared characteristics the focus of our education policies and practices. The former approach is not without its romantic appeal, but the latter finds greater support from those who study how our minds work.

Does this mean I’m against any and all efforts to differentiate instruction? Not at all. I believe we should set common expectations around what students should know at particular points in time, and when some students fall behind these expectations, their teachers should identify strategies to help them.

Alex argues for personalized learning on the grounds that it can provide more feedback to students. The question is whether these varied sources can and will provide meaningful feedback in ways that improve student learning.

These are “testable” claims, and at least thus far I believe the evidence is lacking. Indeed, Alex acknowledges that the evidence in support of personalized learning is, in his words, abysmal. To the extent that personalized learning has generated testable hypotheses, the evidence suggests the theory is flawed.

Alex posits that students need to “own their learning.” Respectfully, I think this simply reiterates the aspirational goals of personalized learning, rather than offering any substantive evidence on its behalf. Of course we need to organize our schools to promote “acquiring background knowledge, accruing expertise, and developing independent learning,” as Alex urges, and who could argue against children growing and developing confidence as learners? The relevant question, however, is whether personalized learning strategies are an effective way of achieving these ends.

In my view, the burden of persuasion should rest with those who would seek to dramatically reorient our education system despite the historical and present-day evidence suggesting their favored approach won’t work for many students, particularly students living in poverty.

I can’t help but wonder what would happen if we spent a greater proportion of our education-related efforts on incorporating the insights of cognitive science into our education decisions. Doing so would require suspending some of our ideological predispositions and focusing on building bridges between those capable of providing scientific insight into how children learn and the educators responsible for imparting that learning.

The stakes here are high. There is a tremendous amount of energy behind “personalized learning” and in support of policies and practices tied to its loose vision for the future of education. Perhaps it will lead to tremendous learning gains as it unleashes learners to pursue their true passions and learn at a time, place, and pace of their choosing. But if we’re wrong about this, we may actually compound the challenges faced by the students who are most in need of guidance and support from the professional experts we call teachers.


ednext-april15-forum-hernandez-smallAlex Hernandez responds

Ben argues that the science is settled with respect to differentiated instruction. I have a very different read of the same research. Where Ben sees a dead end, I see great opportunity. Cognitive scientists make the responsible, prudent suggestion to focus on commonalities because 1) effectively grouping students is hard, and 2) teaching to multiple subgroups is hard. The “do no harm” strategy is simply to avoid the whole endeavor. Groupings matter. Michael Petrilli cites several studies where student grouping patterns were correlated with student achievement (see “All Together Now?features, Winter 2011). The issues involved are complex, nuanced, and political, but it’s hardly a barren wasteland of evidence.

In my twin boys’ randomly assigned kindergarten classroom, some students entered school reading Magic Tree House chapter books, while other students were just learning their letter sounds, a four-grade-level spread among five-year-olds (pre-K to 2nd grade). I disagree that this class of students should be held to a common kindergarten-level expectation with some differentiation for children below grade level.

Cognitive science is not anti-differentiation. In Why Don’t Students Like School, Willingham offers multiple recommendations for how teachers should address student variation in background knowledge. Cognitive scientists’ skepticism of grouping students stems, in large part, from the emergence of various theories on “learning styles.” Categorizing students by and/or tailoring instruction to learning styles has not increased student achievement. Yet most educators can cite one or more “learning style” theories, and it has not quite sunk in that these theories are ineffective when broadly implemented.

Neither differentiation nor feedback scale without technology. This is the opportunity. I don’t blindly worship at the altar of technology. Montessori-style work cycles, carefully curated print books, peer tutoring, and small-group instruction are all important components of a personalized learning environment. But in traditional classrooms, differentiation and feedback are both constrained by the teacher. This can potentially be solved by hiring more people (e.g., additional teachers, tutors), unlocking the power of peers, or opening up new avenues for differentiation and feedback through technology.

I am also unapologetically pro-standards. Personalization needs the Common Core. Ben wants to know whether a 3rd grader is meeting the 3rd-grade standards—yes or no. I want to know if the 3rd grader is at a 1st-grade level in math and/or a 5th-grade level in reading. This “complete view” of the student allows us to help students reach their full potential and guards against the “race to the bottom” we observed during NCLB when states figured out their proficiency rates would look better if they had easier standards or lower cut scores.

The quality of feedback from sources other than a teacher is of great concern, especially as it relates to online learning programs. Silicon Valley has frequently ignored learning science in developing its products, resulting in serious shortcomings. But these criticisms are equally true of traditional classrooms. Worksheets and print textbooks are hardly feedback-rich learning tools based on cognitive science. Students invest thousands of hours a year reinforcing misconceptions by doing work incorrectly or attempting to complete assignments for which they are ill-prepared. People use Khan Academy because it’s better than looking at the answers to the odd problems in the back of the math textbook and trying to reverse engineer solutions. We should have more learning resources, not fewer.

Finally, Ben suggests my appeal for more student ownership of learning relies on circular logic. Some learning environments require and potentially develop more student agency (or ownership) than other learning environments. Many educators engaged in personalized learning are pursuing hypotheses designed to increase student ownership. The research in this area is still evolving. The lack of a robust evidence base shouldn’t discourage us from using the scientific method to develop new knowledge. Otherwise, we’d never try anything new.

This forum is adapted from a series of 2014 blog posts by Hernandez (thinkschools.tumblr.com) and Riley (kuranga.tumblr.com). The opinions expressed herein reflect their individual views and are not necessarily those of their respective organizations.




Comment on this article
  • WFGersen says:

    I am not convinced that all students can progress through the entire sequence of “college-ready” objectives nor should they do so. Our economy does not require that everyone go to college. Moreover there are many service economy jobs that pay well and do not require college as much as they require self-direction, ambition, and hard work. These qualities should be by-products of a personalized learning environment.

    One thing is clear: the current lockstep method of instruction is undercutting the self-direction, ambition, and work ethic needed for success in the work place.

  • Bror Saxberg says:

    Reading this, the word “personalized” is used in two ways: 1) describing environments that provide control to the learner, and 2) making the learning environment and tasks match features of each learner (which could be decided on by the environment or the instructor, not necessarily the student).

    Both approaches have evidence for ways in which they work, and ways in which they don’t. Check out E-Learning and the Science of Instruction by Richard Mayer and Ruth Clark: there’s quite a few studies showing that novices do not pick optimal learning strategies when left to their own devices. Studies by Roediger and Koepicke show that after using a more effective strategy (for long-term retention), students often predict they will retain less – they can’t even tell when they are learning more!

    At the same time, there’s evidence (see E-Learning and the Science of Instruction again) of something referred to as “the expertise reversal effect” at play – things that work well or badly for novices in a domain, may work the opposite way for those who are more expert: e.g., giving someone who is more expert control over their learning choices works fine. And it appears that all learners benefit by being given control over the pace of their learning (although as one of the authors here points out, you do have to address motivation issues, too – learners need to be ready to start, persist, and put in mental effort, and only then do they benefit by being given control over their pace).

    On individualizing to each student’s characteristics, there’s a fair bit of evidence that this can be helpful: if you can connect to what students already have in long-term memory, you can accelerate learning. 1:1 human tutors do this regularly, and Kurt VanLehn’s 2011 meta-analysis of intelligent tutoring systems showed a match between what the better intelligent tutoring system get in lifting student performance and human tutoring performance. This is done by adjusting feedback and, in some cases, problem selection, to match students progress.

    At the same time, as one of the authors point out, there’s a fair amount of myth-making going on about “learning styles” which the evidence just hasn’t backed up yet. We have to be a little careful not to say “it won’t work,” but rather, “it hasn’t yet.” Or, even more carefully, “whenever a good study has been run that compares a “learning styles” approach with a well-designed single instructional approach, no significant difference in learning seems to be visible.” That doesn’t mean there isn’t some magic classification for learning waiting to be discovered: think how complicated the genetic code is, and oncologists are beginning to think it may be a better way to individualize treatment of tumors – we definitely don’t have a “genetic code” for learning to help us (yet).

    The whole discussion reinforces that broad generalizations about learning are probably not the best way to do “learning engineering” (a term I used in a recent Op-Ed in the Chronicle of Higher Education, if you’re interested). It is very inconvenient that our intuitions about learning are often wrong. It’s also inconvenient that our intuitions about health and disease are often wrong, too. We’ve needed decades of science and cultural change to go after health in a different way – we need to get going down that path for learning, too.

  • Raihan kabir says:

    Are we doing what is the best for our students,or are we doing what is most convenient?

  • Jane Jackson says:

    Both of these authors make good points. But personalized learning will fail if it is equated with individual learning as Mark Zuckerberg seems to imply. Personalized learning can succeed when it’s a component of group learning.
    I work with many committed high school physics and chemistry teachers; they take responsibility not only for the academic achievement of their students, but also for their students’ social development. They use a hands-on, minds-on pedagogy called Modeling Instruction, which promotes values like empathy, compassion, and helpfulness; i.e., builds character — with effective science instruction.
    How does it work? These teachers organize instruction into two-week modeling cycles that engage students in building scientific models, evaluating them, and applying them in concrete situations. Rather than lecture, they guide their class to ask questions of nature. To answer the questions, teams of 3 or 4 students design experiments and use the computer to gather data. From their data the teams construct mathematical models and defend them to the class. They apply models to different situations. The course becomes coherent because it is centered on a few basic models. It brings the classroom closer to the workplace because modeling is a central activity of scientists, engineers, and many in business. The teacher is a guide on the side, and the class is a community of learners. Modeling Instruction is described at http://modeling.asu.edu .
    David Brooks addresses this issue in his New York Times op-ed (November 27, 2015) called “Communities of Character”. He concludes, “All over the country there are schools and organizations trying to come up with new ways to cultivate character. The ones I’ve seen that do it best, so far, are those that cultivate INTENSE, THICK COMMUNITY [my caps]. Most of the time character is not an individual accomplishment. It emerges through joined hearts and souls, and in GROUPS [my caps].

  • Chassi Axton says:

    Both arguments make good points in the aspect of personalized learning. Each with it’s own pros and cons. There are also some important concerns and regards for the aspect. So I can agree with both sides on the actions of it.

  • Comment on this Article

    Name ()


    *

         5 Comments
    Sponsored Results
    Sponsors

    The Hoover Institution at Stanford University - Ideas Defining a Free Society

    Harvard Kennedy School Program on Educational Policy and Governance

    Thomas Fordham Institute - Advancing Educational Excellence and Education Reform

    Sponsors