Johan van der Gaag teaches geography at Dalton Den Haag and has been in education for almost 24 years. With 300 students across 14 classes, he knows better than anyone how much time goes into grading tests. He came across Examplary through his school's AI working group. We spoke with him about his experiences, and about the surprise that the feedback for students turned out to be.
How did you come across Examplary?
"This year we have an AI working group at school. Really a brainstorm with colleagues about everything that's possible. One of my colleagues threw out the name Examplary. They dropped it and that was that. So I thought: well, let me go have a look myself."
What made you curious?
"I've been in education for almost 24 years and it's still great fun. But all that grading… yes, it's lovely to see what your students achieve, but it's just a huge pile of work. I have 300 students, 14 classes. For a test you're looking at an hour and a half per class on average. And then you've got practical assignments in between as well. If there's anywhere you can win back time, it's there."
How did grading go with Examplary?
"The first test was mainly an investment. That makes sense, it's new. But by the second test I already noticed: hey, this is going faster. You just have one question on the screen and you go through all the students — one after another, bam, bam, bam. I actually did it that way on paper too, blocks of four or five questions at a time. Now the AI does the groundwork and I only have to check it."
"By the second test I already noticed: hey, this is going faster"
How accurate do you find the AI's assessment?
"For multiple-choice questions and short answers, the computer just does it flawlessly. For open questions you still have to look carefully, and rightly so. I really see it as a grading aid. Say a student has to name three key concepts in an open answer. Then the AI flags it up front: one seems to be missing. I then double-check that. The accuracy is quite high, but I always still read open questions myself. As a teacher you are and remain responsible."
Is that comparable to how colleagues grade among themselves?
"Yes, it's pretty similar actually. We compared it with three other colleagues after an exam week, and the difference with the AI doesn't seem bigger than the difference between us. We share the averages, not every question of course."
"The difference with the AI doesn't seem bigger than the difference between colleagues"
So your colleagues started using it too?
"I sort of talked them into getting started. They're less into AI themselves, but I got them enthusiastic: hey, give this a try, see what it gets you. And they notice it too. My new colleague, almost graduated, said straight away: the grading really goes faster."
How did the students react?
"The first time there was some resistance. It was new for them to take a test that way — at our secondary school, paper is still the standard. That's possible with Examplary too, but we deliberately chose to do it digitally. From our AI working group we want to reduce workload, and it helps if you don't have to scan everything in first. By the second time they were already used to it."
You just mentioned grading goes faster. How much is that when creating the test itself?
"If I make a test entirely myself, without a textbook test as a basis, I'm normally busy for an hour and a half. With Examplary I get through it in about twenty minutes. That's a really enormous time saving. I upload the PDF of the chapter and the AI picks out the subject matter well. It's not always the perfect question yet — you get that when you make it yourself too, by the way — but the level is just good. My colleagues are happy with it as well."
"From an hour and a half to twenty minutes — that's a really enormous time saving"
What surprised you most about Examplary?
"The feedback to students, I hadn't really expected that. When I hand back their tests, they see feedback alongside. I make a point of saying: this is AI feedback, but you can really get something out of it. Especially in geography you have to be thorough and precise in your answers. If something is missing, the feedback indicates that — including why. Students then see: hey, why didn't I get full marks? Where did it go wrong?"
Did you used to give that feedback before too?
"Before, I did that as a class, in general terms. Individual feedback on every question? For 300 students? That was previously unthinkable. I'd have needed a whole day a week for that. Now it's just there. And even though it's AI-generated, there's good stuff in it. I find that a really big added value."
"Individual feedback for 300 students — that was previously unthinkable"
Looking ahead: what could still make a big difference for your school?
"Differentiation. That's always a thing in education. Fantastic in theory, but when you've got 30 students in your classroom, well… in practice it doesn't always happen. If with a tool like Examplary you could really see where a student is struggling, and help that student further in a targeted way while another is challenged instead. That would be really great. That would take differentiation to a whole different level."