Recently, I took my 8th-grade daughter to a horse barn in the name of science—or more precisely, in the name of a project to teach horses to "read." She's fascinated with horses, mucking out stalls and saddling horses at a nearby stable just to be near them and, if she's lucky, take a ride or two. She's learned that horses are smart, complicated creatures with social dynamics as complex as those in any middle school cafeteria.
To test these animals' intelligence, she has designed an experiment (with guidance from her teacher). She shows horses one board painted with a circle and another painted with a rectangle in hopes of teaching them to nuzzle the circle board to receive a treat and ignore the rectangle board (which offers no treat), thus demonstrating their ability to relate abstract symbols to concrete ideas—a form of reading. I've never seen her more invested in a school project. It's a joy to see her eagerly learning about Piaget and Skinner and delving into animal behavior studies.
Yet as we approached the horse barn and her awaiting test subjects, I wondered how effective this sort of learning is. How do we gauge its impact? Moreover, is it equitable? And scalable?
A Complex Mix That Shows Promise
The first thing that becomes apparent upon digging into research on personalized learning is that, generally speaking, it's defined not as a single strategy, but as a constellation of elements, including:
- Using data to create student profiles and set personal learning goals.
- Offering students multiple pathways to demonstrate learning.
- Shifting teaching from providing information to guiding individual learning.
- Providing flexible spaces for large and small groups, mentoring, and independent learning.
Multifaceted interventions like these aren't easy to study, and few scientific studies of personalized learning have been conducted. A recent spate of studies, however, although not quite scientific, suggests some promise for personalized learning:
- A RAND Corporation study compared the achievement of 11,000 low-income and minority students in personalized-learning environments with that of similar peers nationwide and found seemingly positive effect sizes for both mathematics (0.27) and reading (0.19). Students in the personalized environments began below average on national assessments, and within three years they were scoring above average (Pane, Steiner, Baird, & Hamilton, 2015).
- A Stanford University study of personalized-learning approaches in four California high schools serving mostly low-income students of color found that these students outperformed students in nearby schools serving similar populations. In all four schools, students had higher graduation rates, greater gains on state achievement tests, more enrollment in college preparatory courses, and higher college-persistence rates (Friedlaender et al., 2014).
- A Columbia Teachers College study of 4,117 students in 15 schools that adopted Teach to One: Math (a personalized-learning approach for mathematics) found that by the second year of the approach, students' gains were 47 percent higher than national norms (Ready, 2014).
Although these results look promising, it's important to note that none of these studies employed true scientific designs—so the positive results may reflect unidentified variables, such as parental support. Nonetheless, it's worth noting that all three studies focused on students from low-income and minority groups. Moreover, low-achieving students appeared to benefit most from personalized-learning environments. In the RAND study, low-achieving students—those in the bottom 60 percent of their student group—made greater learning gains than did students in the top 40 percent. Similarly, lower-achieving students in the Teachers College study (initially in the bottom third for performance) demonstrated the greatest gains (81 percent higher than similar students in the national sample). Average students (in the middle third) grew slightly better (21 percent higher), and high-achieving (top-third) students stayed on par with similar students nationally.
A Need for Many Shifts
What these promising results mask, however, is wide variability among the schools studied. For example, among the 62 schools in the RAND study, roughly a quarter demonstrated negative effects for personalized learning and another quarter demonstrated no effects. Implementation of core elements of personalized learning was also markedly uneven: Only about half of the schools set personal learning goals for students and less than half used learner profiles to track and guide student progress. Successful schools fully implemented at least three key features of personalization: (1) dynamically grouping students according to learning needs, (2) providing flexible learning spaces to support these groups, and (3) presenting data to students to guide their progress toward learning goals.
In-depth case studies of schools implementing personalized learning suggest that creating true personalized-learning environments requires many shifts—starting with teachers re-envisioning themselves less as information providers and more as learning coaches. At the same time, schools must rewrite the curriculum as learning competencies or pathways that students master at their own pace (Halverson et al., 2015).
Experiment, Reflect, Iterate
Most important, perhaps, personalization appears to require that schools embrace a "fail forward" ethos. As one high school administrator observed, "there's not enough failure in high school. So we need to have structures taken away for the students so they can kind of fall on their faces … I think they learn it that way" (Halverson et al., 2015, p. 7).
Leaders of so-called "next gen" personalized-learning schools highlighted in another recent report (Mead, Schneider, Vander Ark, & Vander Ark, 2014) note that the path to personalized learning is often a winding road and that educators themselves must embrace experimentation. Rather than implementing any single model, such schools knit borrowed ideas into an approach that one leader described as "fail fast, iterate, fix it, keep moving" (Halverson et al., 2015, p. 37).
Ultimately, the complexity and rapid-cycle innovation aspect of personalized learning makes it difficult to pin down—or even study—any one way to do it. Rather, the best approach to personalization appears to reflect practices schools want students to adopt: experiment, reflect, and iterate.
I saw this with my daughter that day at the barn. After many adjustments to her study design, she taught three horses to "read"—to nuzzle the circle board to receive a treat. On the way home, while analyzing her data, the wheels began turning in her mind for larger, more complex studies. She asked where she could study such things in college. Perhaps that's the real power of personalization: it encourages students and educators alike to ask, What did I learn? What else can I learn? How will I do it better next time?
