AI is transforming education: learning and showing what you've learned

AI is changing how we learn, work and conduct research. And it is doing so at remarkable speed. The UvA is studying this change and integrating AI into its education. This means doing two things: teaching students to use AI critically and responsibly, and carefully assessing what they themselves know, understand and can do.

A chatbot that writes texts, helps with programming or asks questions as a practice partner: AI offers all kinds of opportunities to learn and work differently. At the same time, the technology raises questions. When does AI help you understand something, and when does it take over the thinking you need to do to learn for yourself? And how can lecturers assess what a student has genuinely mastered?

‘A chatbot can write a new text in seconds. Carefully improving our teaching and assessment and introducing new approaches takes a little longer,’ says Rector Magnificus Anniek de Ruijter. ‘Good education calls for a more thorough approach: deciding what students need to learn, developing appropriate assignments and assessing carefully. We are constantly working on this. And we certainly aren't waiting until AI is fully developed, because in the meantime we need to prepare students for the future.’

Learning and showing what you have learned

Preparing students for a world with AI takes more than teaching them how to give a chatbot an instruction. It also means checking whether AI-generated answers are accurate, recognising missing perspectives and taking responsibility for how they use AI. Subject knowledge remains essential: to assess an answer, you need to understand its content.

This responsibility is also reflected in the development of degree programmes. A new learning pathway within the Bachelor's programme in Information Science and the Master's programme in Information Studies combines technical knowledge with critical thinking about people, technology and society. ‘We need professionals who can not only develop systems, but also critically assess, anticipate and shape their impact,’’ says Sennay Ghebreab, programme director of the Master's in Information Studies.

Learning to use AI does not mean it is permitted in every assignment. Students need to know which rules apply and what they must be able to demonstrate independently.

‘AI is changing how we learn, work and assess. The standard for a university degree remains the same,’ says De Ruijter. ‘Even when AI is permitted in an assignment, we must be able to establish that the student has the required level of academic, analytical and critical thinking skills. We design our teaching and assessment accordingly.’

The UvA's Teaching & Learning Centre (TLC) offers lecturers a range of ways to adapt their assessment. Examples include a written assignment followed by a short oral defence, in which the lecturer asks probing questions about the content and the choices made. Another option is a supervised analysis assignment in which students examine an unfamiliar dataset or apply their knowledge to a new case. Practical assessments can also provide insight into what students can do independently.

These are not all new forms of assessment. The change often lies in combining different methods, with students increasingly being asked to explain the thinking process that led them to an answer.

Writing remains a way to learn

With essays, the writing process can also provide insight into students' understanding. Speaking to AT5 (in Dutch), Judith Huigens, programme director and lecturer on the Bachelor's programme in Political Science, describes progress discussions with students and presentations to the group:

‘Writing essays is essential for practising academic skills. Writing is therefore not only a way to assess knowledge, but also a way to learn how to think and construct arguments.’

Huigens: ‘In a discussion, you can ask probing questions: why are you using this source, how did you arrive at this argument, and how does your conclusion follow from it? This gives you insight into the understanding behind the text. At the same time, students must continue to get enough practice writing for themselves. We want to preserve that learning process.’

A signal is not proof

AI also presents new challenges in preventing and investigating academic fraud. Detection tools such as Pangram and GPTZero analyse texts for patterns that may indicate AI use. Their results may provide a reason to examine the work more closely, but they do not prove that a student has committed academic misconduct. Even if AI has been used, it must be established whether this breached the rules for the assignment.

Detection tools are not infallible either: they can incorrectly flag human-written texts as AI-generated and fail to identify AI-generated texts. Frank Benneker, AI programme coordinator and information manager at the University of Amsterdam, says: ‘We cannot blindly trust AI detectors. Research shows, for example, that they can incorrectly flag texts written by people who are not writing in their native language as AI-generated. That poses a risk for English-language theses, for instance. There is also the prospect of a competition between detection tools and tools designed to circumvent them. Ultimately, mutual trust matters too.’

The UvA has therefore decided not to purchase AI detection tools. After all, a detection tool cannot replace an assessment of what a student knows, understands and can do. The focus is on appropriate assessment and grading, a variety of assessment formats to test different skills, clear rules and careful investigation where there are specific grounds for suspicion.

‘At the UvA, we shape our teaching together with students. When assessing their work, we trust that they too want to know how well they have done. Most of our programmes use a variety of assessment methods, giving us ample opportunities to determine whether students meet the knowledge and skills requirements for their degree,’ says De Ruijter.

More than the thesis alone

A degree programme is more than a collection of individual courses. Through assignments and assessments across the programme, students demonstrate the knowledge and skills they need to earn their degree. The thesis is one part of this.

In short, most theses follow this path: from a research question and research design, through research and drafts, to the final text. Along the way, students discuss their progress, incorporate feedback and explain their choices. This means supervisors gain insight into students' understanding well before the thesis is submitted.

In some Master's programmes, for example within the Faculty of Economics and Business, students may also use AI to write their thesis text. They remain responsible for the content and must stay in control of the process. There are therefore at least four mandatory meetings, each preceded by a progress report. Students reflect on their own learning process, the actions they have taken and where they need help. Supervisors ask critical questions about aspects such as the research method, the interpretation of results and the conclusions. The process concludes with a presentation and questions about the research.

In several programmes within the Faculty of Medicine, the oral defence now carries greater weight in the thesis grade. The written work carries less weight, with less emphasis on writing skills. During the defence, students must demonstrate that they understand their research and can explain their choices. They must receive a passing grade for both the written work and the defence.

A well-written thesis alone is therefore not enough: both during the process and at the presentation, students must demonstrate that they understand their research and can explain their choices.

Want to know more?

For students: AI and your studies. What is possible, what is permitted and what should you look out for?

For lecturers: GenAI and assessment, and adapting existing assessments. Practical guidance and examples.

The UvA's rules: Policy framework and guidelines for GenAI in education.

AI is transforming education: learning and showing what you've learned

Learning and showing what you have learned

A signal is not proof

Want to know more?

Cookie Consent