Adaptive learning technology such as intelligent tutoring systems enable the personalisation of students’ learning using similar approaches: they detect the knowledge (or knowledge gaps) of students; they diagnose the next appropriate steps for students’ learning; they act by providing new exercises, new curriculum units, some form of instruction, or just notifying the teacher. This approach is now being expanded beyond mere knowledge acquisition and factoring in behavioural dimensions such as learning self-regulation or style (Chapter 3).

As keeping students engaged and motivated is key to learning effectiveness, a new domain of technology development focuses on measuring engagement and interventions to keep students engaged, both in digital and physical learning environments. Measuring engagement is difficult but a host of new automated approaches have been developed, from eye trackers to the monitoring and analysis of other facial features. Improving engagement typically takes two routes: proactive approaches try to stimulate engagement with incentives, gamification, etc.; reactive approaches do it in a more sophisticated way by continually monitoring engagement, detecting when engagement is waning, and adapting instruction to address periods of disengagement (Chapter 4).

While smart technologies focusing on personalising learning for individuals are probably the most pervasive, another approach is to consider the classroom or rather what happens in the classroom as the subject of the learning analytics. The objective is to support teachers in orchestrating the learning in their classroom and to propose rich and effective learning scenarios to their students. Some classroom analytics techniques provide teachers with real-time feedback to help manage transitions from one task to the next as their students work individually, in small groups or collectively, for example. They also give feedback to teachers on their classroom behaviour so they can reflect on and learn from their practice (Chapter 5).

Social robots are also being increasingly developed for learning uses. Usually powered by the personalisation systems mentioned above, they support teachers in different ways: as instructors or tutors for individuals or small groups, but also as peer learners allowing students to “teach” them. Telepresence robots also allow teachers or students to teach or study remotely and offer new possibilities for students who are ill and cannot physically attend class. They can also mobilise a remotely located teaching workforce, for example teachers from another country to teach foreign languages (Chapter 7).

Technology also enables students with special needs to participate in education and to make inclusive education a reality. With well-known applications such as speech-to-text, text-to-speech, and auto-captioning, etc., AI allows blind, visually impaired, deaf and hard-of-hearing students to participate in traditional educational settings and practices. Some smart technologies facilitate the diagnosis and remediation of some special needs (e.g. dysgraphia) and support the socio-emotional learning of students with autism so they can more easily participate in mainstream education (Chapter 6).

Those smart technologies usually assume and require a human-in-the-loop: a teacher. The level of automation of actions and decisions should be conceived of as a continuum between actions that are fully automated at one end and, at the other end, actions over which humans have full control. As of today, AI systems remain hybrid and request human intervention at a certain point in the process.

Smart technologies powered by AI and learning analytics also allow for the management of education organisations. They can be used for a variety of purposes; for example, to enhance an institution’s curriculum based on an analysis of students’ learning and study paths. While this is still a nascent trend, a whole-of-organisation adoption of learning analytics can transform educational institutions’ culture (Chapter 8).

Early warning systems that identify students at risk of dropping out from high school are a good use of the administrative micro-data that are increasingly being collected by education systems and organisations. While identifying a good set of early warning indicators remains difficult, a few systems have shown a high level of accuracy and enriched thinking about the reasons students drop out. In order to avoid the risks of student profiling, open and transparent algorithms are important (Chapter 9).

Game-based standardised assessments also build on smart technologies and smart data analysis techniques to expand assessment to skills that cannot be easily measured by traditional (paper-and-pencil or computer-based) tests. These include higher-order skills (e.g. creativity) or emotional and behavioural skills (e.g. collaboration, behavioural strategy). Game-based tests may analyse eye-tracking data and audio recording, and process natural language and information such as time-on-task or use simulations (Chapter 10).

Finally, as a “verification infrastructure”, blockchain technology opens new avenues for credentialing in education and training. Blockchain technology enables the validation of claims about an individual or institution, including their characteristics and qualifications, and to do this instantly and with a very high level of certainty. This helps eliminate diploma (and other records) fraud, facilitates the movement of learners and workers between institutions and geographies, and empowers individuals by giving them increased control over their own data. Many blockchain initiatives are underway across the world, which may transform how education and lifelong learning systems manage degrees and qualifications (Chapter 11).

There are good reasons to believe that smart technologies can contribute to the effectiveness, equity and cost-efficiency of education systems. At the same times, there are a few important aspects of smart technologies to keep in mind to reap those benefits:

• Smart technologies are human-AI hybrid systems. Involving end users in their design, giving control to humans for important decisions, and negotiating their usage with society in a transparent way is key to making them both useful and socially acceptable.

• Smart technologies support humans in many different ways without being perfect. Transparency about how accurate they are at measuring, diagnosing or acting is an important requirement. However, their limits should be compared to the limits of human beings performing similar tasks.

• More evidence about effective pedagogical uses of smart technologies in and outside of the classroom as well as their uses for system management purposes should be funded without focusing on the technology exclusively. Criteria for this evidence to be produced quickly could also be developed.

• The adoption of smart technologies relies on robust data protection and privacy regulation based on risk assessment but also ethical considerations where regulation does not exist. For example, there is mounting concern about the fairness of algorithms, which could be verified through “open algorithms” verified by third parties.

• Smart technologies have a cost, and cost-benefit analysis should guide their adoption, acknowledging that their benefits go beyond pecuniary ones. In many cases, the identification of data patterns allows for better policy design and interventions that are more likely to improve equity or effectiveness. Policy makers should also encourage the development of technologies that are affordable and sustainable thanks to open standards and interoperability.