Education edition
Three trials in education show where AI can help, and where teachers remain in control.
Mark first, then AI feedback
Trial: AI gives feedback and practice exercises after marking, but no grades
TU Delft is testing AI feedback on exam answers. A teacher marks the exam first. Then a language model gives feedback and practice tasks. The trial aims to reduce teachers’ marking workload. It should also help students learn sooner. A student assistant is building software that reads scans of paper tests. The student council funded the trial in June. The first results are expected in December. So the trial is still under way. We do not yet know how much time it saves teachers. For other schools, this is an example of a clearly limited role for AI. The teacher assesses the work first. Only then does AI give feedback and practice tasks. This week, consider what feedback pupils need after a test. Then discuss whether a trial could help.
The order matters. First, a teacher decides how good the answer is. Then the language model creates feedback and exercises based on the marked answer. This way, AI helps with the next step in learning. AI does not decide the grade. The software reads scans of paper tests. What happens to the data? According to the person behind the trial, it will stay on campus if the model is actually used.
The trial has no results yet. They are not expected until December. So we do not yet know whether the feedback is accurate, useful or time-saving. The teacher is still needed. They check that the comments fit the answer and the course material. Also pay attention to the agreement about data. The person behind the trial says it will stay on campus if the model is actually used. That is not a general guarantee for every AI tool or school.
Background for subscribersThe reports show that AI can help with feedback and getting an overview. But AI is not automatically reliable. Let a teacher assess the value of pupils’ work. Check AI output. Make clear in tests what pupils must show themselves. Also ask what data a school tool uses and what actions it can carry out on its own.
Leiden teachers must assess tests for AI risks
The Faculty of Humanities in Leiden has made rules for the use of AI. Teachers decide for themselves whether students may use AI in a test. But they must consider how likely it is that students will use AI when it is not allowed. They must also try to reduce that risk. The bachelor’s programme must include AI literacy. This means knowing how AI works and where its limits are. In the master’s programme, students must learn about the risks of AI. The faculty council thinks the rules leave too much up to teachers. A working group is looking at how writing assignments can continue. For schools, this is an example of a policy that does more than ban something. It also asks schools to adapt tests. This week, check whether an assignment makes it clear what pupils must show themselves. The rules give teachers a lot of responsibility. The teacher decides whether AI is allowed in the test. They also assess how vulnerable the test is. That means making a clear choice for each type of test. The working group is also looking for ways to keep writing assignments. The source does not mention a detailed approach or a fixed measure yet. The faculty council thinks teachers are being given too much freedom. If every teacher makes their own choice, the rules may be unclear to students. So the tension is between teachers’ freedom and clear, consistent rules for students. Explain what pupils must be able to show themselves in each assignment.
Background for subscribersThis is a taster.
Subscribers read the whole Education edition: every story, each with its background in plain language.
Subscribe →Not ready for that? Read the free letter first →
AI and tech, made simple.
This is the Education edition. Subscribers get the background to every story, in plain language.
Subscribe
inklaretaal · Amsterdam · © 2026