AI in education – the possibilities are promising

In this Educator we look back at June as the 'Month of AI in Education', proclaimed by SURF Special Interest Group AI in Education, the Acceleration Plan and the Dutch AI Coalition education working group. During this month you could find out, for example, what you can do with AI in education and what ethical considerations you have to deal with. In "AI in Education," the educational process is the goal that is attempted to change and improve. AI applications are looked at and explored how they can be used to support students and lecturers.

For this edition of the Educator, we talked to:

• Prof. Dr. Marcus Specht, Professor of Digital Education at Delft University of Technology and Director of the Leiden-Delft-Erasmus Center for Education and Learning.
• Duuk Baten, Innovation Project Manager at SURF.
• Bart de Grood, project associate in artificial intelligence in education at the Leiden-Delft-Erasmus Center for Education and Learning. Among other things, we asked them what the added value of AI in education is, what the risks are and where the opportunities lie for TU Delft.

What does AI in education look like, for example?

Marcus: "Many different Intelligent Tutoring Systems have been built. In the medical education sector these are very popular, with expert systems to train doctors in diagnosis, for example. We also just started a new project with conversational agents, also called chatbots. You have very simple chatbots, which can only answer simple questions. However, these systems are becoming more sophisticated and are going back to the original idea of an assistant that can really show intelligent behavior."

Bart: "For example, you have an AI chatbot that helps high school students do their homework. When they have a question, students are connected directly to a coach using AI. I think this is a revolutionary app for education."

Duuk: "On the one hand, you can use AI to give feedback during the learning process for example, think of feedback on writing style or the use of references in essays. Another example is the personalisation of education; based on data from the student a better estimate can be made of what kind of education the student can best be offered."

What are benefits of integrating AI in education?

Duuk: "For lecturers, the hope is that with this kind of tool you can save some of the cognitive capacity, allowing the lecturers to spend more time on the things that they find most important. For example, it might be interesting for lecturers and students if the AI could provide feedback on writing style, giving the lecturer the ability to focus her/ his energy on the content. It does depend on what you think is most important as an institution or as a lecturer. You get very different answers when you focus on efficiency than when you focus on lecturer-student interaction, or if you look at it from the perspective of the lecturer or the student."

Bart: "I have parents in education myself and have seen how much administrative work lecturers have. I can imagine AI being perfect for taking away this kind of workload, for example, by automating the entry of grades and making an analysis of this."

Marcus: "A classic advantage in higher education is scalability. In practicing programming, for example, you can give students a lot of coaching by an intelligent or artificial system, which frees up time for complex programs to be handled by human coaches. Currently, we are doing a project together with Feedback Fruits on writing and essay analysis, where AI systems analyze students and give feedback on style and syntax and can let students practice with them. The new AI systems, which work with neural network models, can really understand the context and provide details about each sentence. This makes AI even more useful for teaching. I see this triangle of AI components, learner and lecturer more and more. We do need to learn how to divide the task optimally between lecturers and AI components."

What are risks of using AI in education?

Marcus: "At a higher level, Intelligent Tutoring Systems or personalized systems often make the stronger students stronger, and the weaker students weaker. With all forms of assessment and estimation, you have to be very careful about using these systems. I think we need to develop systems that support all students in the optimal way. "

Duuk: "On a societal level, you see that with tools built by suppliers, there is a risk that you give away the key of the back door to the educational institution and that soon, for example, TU will no longer be needed to train engineers. Another risk is in the scalability of AI. Everyone has probably experienced a lecturer who liked you a little less or more and where you actually scored low or high. There is undoubtedly also a systematic bias towards certain subgroups, but with AI it is also scalable, which could mean that you are judged the same with every lecturer. What I think is another interesting risk is that AI can actually sometimes be less effective than we think. Then you're investing a lot of money and time into a system that doesn't really work or may even work poorly."

Bart: "I think lecturers and the social contact between people is the most important thing in education. I do think that it can be very useful as a student that an AI application can give you insight into your learning process and what you can possibly grow towards. But I think you should never replace that lecturer for a combination of AI applications."

What do current "AI in education" activities at TU Delft look like?

Marcus: "There are some initiatives in writing assessment and programming education. There are also projects with Multi Model Analytics. We have one in the Faculty of Aerospace Engineering and one in Industrial Design, together with Computer Science.  This AI analyzes the cooperation between people and tries to provide feedback. The system uses sensors to record the dialogue between a group of students, for example, and then makes models about who is cooperating, who is making contributions, etc. It tries to support the group in better cooperation."

What opportunities are there for society in general or TU Delft specifically regarding 'AI in education'?

Duuk: "We are at the point where we have to think very carefully about what we want, how we want to do it and create conditions. At the same time, our environment is not standing still. A lot is changing and it is important to keep up; to collect knowledge, to try things out, so you know what works and what doesn't, and to make sure that we safeguard the things we think are important in AI. However, I think it's going to be a while before we really start applying AI in education."

Marcus: "I think the possibilities are promising. We have very strong students who can actually build these systems. In addition, I think this is a great trigger for ethical discussions about the consequences. Because you can build the Black Mirror nightmare systems, but you can also build really good systems. I think by building those systems we learn a lot. So we need to focus on how the intelligent systems are built, but also how they interact with people and how they collaborate with people."

You can read more about AI in education on this website from SURF's Special Interest Group AI in Education.