The Promise of Adaptive Learning

The Promise of Adaptive Learning

The Promise of Adaptive Learning

Guy Levi, Chief Innovation Officer at the Center for Educational Technology.

[maxbutton id=”1″ ]

The world’s leadership is today being called upon to address the growing gap between the needs of the labor market and the skills that students acquire in the education system. This gap is a key issue that has concerned policy makers and many educators in recent years. The growing labor market of the 21st century is characterized by new professions that require a different kind of functioning, new abilities, and different skills from those that were appropriate for the last century. People are connected in new, previously unheard-of ways that influence the development and growth of the global economy. This new situation demands the development of new patterns of teaching and learning and new models to provide a response to these changing needs. The more advanced education systems in the world are setting such goals as education toward learning, exploring ways of learning ideas that are yet unknown, and being ready to solve problems that cannot yet be formulated. Technology has presented us with a new reality that will allow the creation of these new patterns and models. However, technology alone is not sufficient, as Michael Barber argues in his introduction to Fullan & Donnelly’s (2013) paper: “…The future will belong not to those who focus on technology alone, but to those who position it within a broader context, and see it as but one component in a broad, comprehensive process of systemic change.”

A key component in this broad systemic change is “personalization,” which means the adaptation of the teaching and learning processes to the personalized needs of the student. Hence, the educational and cultural emphases are now on 21st-century skills and competencies and they require new research and developments to meet the new objectives and goals. For example, in mathematics: whereas in the 1970s emphasis was placed on dealing with mathematical procedures efficiently and precisely, today the emphasis has moved to the solution of problems (“problem-solving skills” in 21st-century language), to application, reasoning, creativity, and critical evaluation. These goals, which had previously been almost totally absent from the goals of formal (K-12) education, were made possible through the development of technological tools (which had previously not existed) that could carry out the procedures. This also created an ability to focus the teaching and learning processes on other important ideas.

The challenge of adaptive learning

Today, with the aid of the new technologies, it is possible to develop learning approaches that also include the use of representations, work on those representations, research into mathematical phenomena through dynamic technological applications, and feedback from the computer through mirroring of the outcome of the student’s action (“intellectual mirroring”). The feedback allows students to solve problems, to research and test different alternatives and decide whether they have achieved what they set out to do, and, by testing, to generalize ideas and phenomena. In this regard, feedback is changed from a confirmation of prior knowledge – feedback – to the key to new knowledge – feedforward. This distinction is of utmost importance because, while feedback focuses on current performance, feedforward looks ahead to the next assignment, i.e. the predominance of formative assessment over that of summative assessment. In addition, the technology allows, on the one hand, the assembly of rich content to develop the required concepts and ideas in the field, together with the disciplinary goals and learning skills, while on the other hand it allows the student’s learning abilities to be checked and analyzed using analytical tools applied to “big data,” collected and analyzed on an ongoing basis, and, on the basis of matches between them, to construct teaching and learning processes appropriate to each student.

In the K-12 world of digital learning, there are today more and more solutions based on adaptive learning using various models – either products that promote personalization and offer a complete solution for a given syllabus (particularly in mathematics, although not only in that field), or as solutions that can be integrated into existing environments and products, and which offer significant added value to learning and to the transition to personalization:

the growing use of learning analytics (analysis of learning data) which allows the teacher to obtain a report on each student at any time and over time; the conception that learning has to be relevant and of value to the learner and the instructor, and thus should be active, not passive; and finally, the understanding that learning is possible at all times and is not limited to particular times or places – these explain the broad range of offerings and the variety of solutions that adaptive learning is today beginning to make available.

What is an adaptive learning system?

Technological learning systems are considered adaptive when they change dynamically in regard to each student and in response to data collected in the course of the learning itself, to better adjust themselves to that learning. The system makes use of the data accumulating while the student is working, in order to change, for example, the way in which a concept is presented, the level of difficulty, the sequence of problems or tasks, and the nature of the hints or feedback provided to the student. Thus, the students receive an individualized pace and pedagogic approach and a flexible study track in keeping with their needs, interests, and choices.

Types of adaptive learning systems

There are three types of adaptive processes that can be found in the various systems:

Model based on diagnostic tests – built on the basis of a test administered at the beginning of learning a topic. The student is tested on the requisite knowledge for learning the topic to come and on the knowledge that he is about to learn. The test indicates to both the student and the teacher what the student knows and what he does not yet know. Teaching continues with teacher adaptation of content from the existing content repertoire. This model is applicable to book-based learning, computerized activities, and other media.

The learning content may also be computerized or not, as the teacher chooses. This approach may be classed as personalization since the teaching is adapted to the variation among the students. The bulk of the work of adaptation, as well as significant portions of the teaching process, remain in the hands of the teacher.

Rule-based model – constructed through the use of “if-then” functions. The student is asked one or more questions. If he answers correctly, he moves to the next activity or content unit; if he errs, he receives a hint or an explanation that is somewhat different from the earlier one, depending on the answer that he has chosen. This model is computerized; it makes use of the knowledge of experts in the knowledge domain, who create branched structures and rules for progress regarding the content being learned. All the progression rules are defined in advance by the content specialists. In this approach, both the adaptation and many parts of the teaching process are carried out using the computer.

Algorithm-based model – a computerized model constructed using mathematical (statistical) functions that analyze the student’s performance and collect information on content. The more students who work with the system, the more up to date and precise the data is. As learning progresses, the system learns more and more about the student and about the content and is able to combine what it has learned more effectively. The model is capable not only of assessing what the student has already carried out but also of adding information about what the student knows at a higher level of detail, so as to more precisely adapt content to him. Systems such as this make use of educational data mining, are involved in big data analysis, and create complex algorithms for predicting the probability of a student’s success in learning particular content. The learning tracks are built from computerized analysis of the student’s performance, and an infinite number of such learning tracks can be created. This adaptive learning system may be facilitator-driven and provide real-time data that the teachers can use and based on which they can act. At the same time, the system can be assessment-driven, with the ability to carry out the adaptation itself, thus allowing the students to progress on their own as they proceed through the course. The future of AI promises new breakthroughs in adaptive learning models which will be integrated also into skills and competency-based education.

[maxbutton id=”1″ ]


Fullan, M., & Donnelly, K (2013). “Alive in the Swamp, Assessing Digital Innovations in Education.” Nesta, NewSchools Venture Fund.