Each learner tends to respond differently to the same instruction—and sometimes a given learner will respond differently at different times. We believe that personalized instruction, i.e. instruction that is adapted to individual learners, is a promising way to address the challenges of these differences. However, not all approaches to personalization are effective in improving educational outcomes. To be successful, they must consider more than just a learner’s fixed characteristics.

The emergence of personalized learning

Personalized learning approaches fit the instruction to the individual learner, taking into account factors such as existing knowledge or motivation. The first to examine personalized learning from a psychological perspective was the psychologist Lev Vygotsky, in the early 20th century. He developed the concept of the “zone of proximal development,” which encompasses all the tasks or challenges that learners cannot accomplish on their own, but can accomplish with support.

According to Vygotsky, optimal instruction is always situated within this zone. Clearly, these zones are very different for each learner. Teaching that addresses a group of learners all at the same time risks being too challenging for some learners and not challenging enough for others. Vygotsky provided good theoretical reasons for personalization, but does this approach actually improve learning?

Prominent evidence of the promise of personalized education came from Benjamin Bloom in the 1980s. Bloom found that one-on-one tutoring produced larger learning gains than did regular classroom instruction. This is generally assumed to be because tutors working one-on-one can tailor their teaching so that it sits in the learner’s zone of proximal development. The goal of personalized instruction is to bring the benefits of one-on-one tutoring to larger groups of learners, without providing an individual tutor for each.

One way to achieve this goal is through educational technology, such as an intelligent tutoring system. These digital systems assess several learner characteristics – often repeatedly – and adapt instruction to them. Intelligent tutoring systems are more effective in improving educational outcomes than are other forms of computer-assisted instruction, such as programs that simply present increasingly difficult tasks, without adapting to each learner.

“Intelligent tutoring systems assess several learner characteristics – often repeatedly – and adapt instruction to them.”

In regular classroom instruction, too, repeatedly assessing learner knowledge leads to greater learning gains, probably because teachers use the information gained from assessments to adapt their instruction to individual needs. In any event, regular assessments can help both digital programs and teachers to provide tailored instruction that helps children learn within their zone of proximal development.

Personalizing to account for dynamic characteristics

Not all learner characteristics are helpful for the purpose of personalization. The notion that learners have preferred learning styles that enable them to learn more effectively, for example, is not supported by evidence. Students don’t learn any better when the mode of presentation or organization of learning materials is in keeping with their preferences.

Successful personalization takes into account the fact that learners change during the learning process, and through interaction with that process. Personalizing to account for “static” characteristics, such as learning styles, appears to be much less effective than adapting to dynamic characteristics that change over time, such as knowledge or interest.

“Successful personalization takes into account the fact that learners change during the learning process, and through interaction with that process.”

Existing knowledge can be inferred from the learner’s performance on educational technology, or assessed separately and entered into the educational technology program. Motivational characteristics such as interest are usually measured through learner self-reports and then entered into the program. Ideally, dynamic personalization should measure these relevant factors and adjust instruction to reflect the most recent measurements as well as the underlying trajectories for each learner.

Considerable research remains to be done on personalization. Little is known about how learner characteristics interact to determine the effectiveness of instruction. For example, how does learning differ for someone who has considerable subject knowledge but low motivation, relative to someone with little knowledge but high motivation? And how should instruction take account of those variables? Because learners differ in multiple dimensions, truly personalized instruction requires considering a variety of learner characteristics simultaneously.

Keep up to date with the BOLD newsletter