Faculty of Social Sciences
Zoek in de site...

Hybrid human-AI learning technologies and the role of the teacher

The last year during corona pandemic, increasingly digital learning technologies were utilized in educational institutions. This was beneficial for shaping home education and gave us a great advantage in comparison to other countries. Together, we learned how to use technology in the best way and for what circumstances human interactions are more valuable. This experience clearly underlined the value of education in the social context of school.

The role of technology & AI in education

There are many different scenarios in which AI can be implemented into education. Over the last 35 years, the focus was on developing an intelligent tutor for every student which fueled the belief that AI aims to replace the teacher. However, as we have learned the past year that technology will not replace education, AI will also not replace teachers. On the one hand,  AI is valuable for analyzing data, classifying certain behaviors, diagnosing students’ knowledge and suggesting appropriate subsequent learning actions. On the other hand, humans are good in creative thinking, solving problems together, and connecting and indicating different perspectives.

In this light, we increasingly see an augmentation view on AI, which supports professionals in education. This is also referred to as Hybrid-Intelligence (HI), i.e. a fusion between human and artificial intelligence. Central in this thinking is how these two forms of intelligence can reinforce each other. A defining characteristic in such a hybrid Intelligent system is the fluctuating boundaries between AI and human decision-making.

Digital learnings technologies & adaptive learning technologies

In this light, we make an an important distinction between digital learning technology, technologies in which learning materials are made available through technology, and adaptive learnings technology, technologies which are smart in a certain way and use a form of AI.

The "intelligence" in adaptive learning technologies is taking form in the following ways:

  • it detects data in the learning environment, for example the student's answer to a question or the time it took to answer
  • Based on this data, it diagnoses behaviors and/or knowledge, for example current French words known by a student
  • This diagnosis is translated into an overview with information (dashboards) or into actions, for example feedback on their answer


If we take a look at the actions these adaptive learnings technologies often take, we detect three forms:

  • Feedback on the task students perform, we categorize this as the step adaptivity
  • The selection for the next task, this can be an explanation or an assignment that fits the current level student’ knowledge . We categorize this as task adaptivity
  • The selection for the most appropriate subsequent learning objective. We categorize this as curriculum adaptivity. This can also regard learning objectives that are not yet sufficiently mastered by students.


Our perspective on future education and how intelligence in learning technologies can assist

The model displayed below helps to think about the division of roles between teachers, students, AI, and the degree of teacher control in different levels. The 6 levels of automation highlight different stages in development towards full automation. Originally, this model was developed for the car industry to visualize the steps towards a self-driving car. This model has already been translated to health care and we recently translated it to education as you can see below. Central in this model is the connection between human and AI intelligence.

Automation model EN

  1. Teacher only

In level 1 the teacher has full control over the learning environment and learning technologies have no organizing function.

  1. Teacher assistance

The second level of the model is teacher support. At this point, the teacher receives additional information from the learning technology, such as in the well-known dashboards. The teacher receives additional insights from the learning technology but are in full control. The learning technologies only inform, but does not perform any tasks.

  1. Partial automation

The third level is partial automation. The learning technology adopts small tasks from the teacher. An example is the adaptivity task where the learning technology selects the next best fitting task, making the learning process more efficient. Teachers will have more time for other important tasks , such as providing elaborative feedback.

  1. Conditional automation

The fourth level is conditional automation where the technology performs a broader set of tasks and notifies and advises the teacher to take action. An example is the learning technology like Mathia. Mathia includes both extensive feedback on students’ answers and determines the most appropriate subsequent assignment. Different forms of adaptivity are combined and therefore more tasks are transferred to the AI.

  1. High automation

The fifth level, high automation, describes a learning technology that works largely independent, but asks the teacher for input or additions if necessary. An example is MathSpring which is an intelligent tutoring system that guides the learner in selecting learning goals and offers personalized instruction, practice opportunities and feedback driving towards the learning goals. The functions of the learning technology in high automation are to steer learners and only in exceptional cases notify the teacher to take action.

  1. Full automation

The last level is full automation where the learning technologies work autonomously. An example is ALELO, an American system for learning a foreign language. This system combines all three forms of adaptivity and can independently shape the interaction with the student. The role of the teacher is completely taken over.

The model shows that the roles of teachers and artificial intelligence change increasingly and different forms of Hybrid Intelligence (HI) are emerging. Our focus for future research will be to further understand the role of teachers in interaction with artificial intelligence and develop ways in which they truly reinforce each other.