Rethinking the Bachelor’s Thesis in the Age of AI

Through the AI track within the Visible Learning Trajectories Programme (ZLP), Moss and his colleagues critically examined existing assessment practices, exchanged ideas with peers and experts and explored how AI is reshaping academic learning and what this means for educational design. One of the key insights from the ZLP track was that assessment should focus not only on what students produce, but also on how they arrive there. The result is a new assessment model that makes students’ thinking visible throughout the thesis process. Rather than relying heavily on the final written product, students are asked to demonstrate their understanding, reasoning and decision-making at key moments during their structured thesis trajectory.

The redesign allows supervisors to gain deeper insight into student learning while ensuring that critical thinking and scientific reasoning remain visible and assessable in an AI-supported learning environment.

The experience of the Psychology programme illustrates the value of the ZLP track: providing the expertise and space for educators to translate emerging challenges into concrete educational innovation.

“The question is no longer whether students use AI. The real question is how we can assess whether they actually understand what they are doing.” – Moss Shukla, Thesis Coordinator, Bachelor’s Programme in Psychology

The Bachelor’s programme in Psychology consists of six programme groups, each with its own research focus and staff, including a thesis coordinator. While these groups operate relatively independently, all students complete the same Bachelor’s thesis as their final academic milestone. Ensuring consistency in learning outcomes, assessment criteria and supervision is therefore essential.

To achieve this, the programme introduced a structured workshop model several years ago. Across five workshops spread over three months, students work through the key stages of the thesis process, from developing a research question to collecting and analysing data and reporting findings.

The workshop model proved successful in creating alignment across the programme and ensuring all students receive consistent guidance. Yet the emergence of generative AI exposed a new challenge: are we still assessing the learning we value most?

Currently, 66% of the final grade is based on the written thesis, while 34% reflects the process through participation and drafts. While this model worked well in the past, the rise of AI prompted a critical reflection on what those grades actually represent.

For Moss, the challenge became increasingly clear: “If a student can earn most of their grade through a written product that was largely generated by AI, without engaging with the underlying thinking, what does that grade actually tell us? We need to assess whether students truly understand the why behind their choices.”

During the ZLP track, Moss and fellow educators explored a challenge faced by universities worldwide: how can we assess learning when AI can support, or even perform, many of the tasks traditionally associated with a Bachelor’s thesis? The concern was never the use of AI itself, but the possibility that students might outsource essential thinking processes. If students use AI to generate hypotheses, analyse data and write discussions without engaging critically, what are they actually learning? Can students still explain why they formulated a particular hypothesis, selected a specific statistical analysis or interpreted findings in a certain way?

The discussions during the AI track highlighted a broader issue: the programme was primarily assessing the quality of the final product, while the most valuable learning often takes place in the reasoning behind it.

Moss and his colleagues were eager to design a new assessment model to make the students’ learning process more explicit. Existing thesis workshops were redeveloped as assessment opportunities where students demonstrate and receive meaningful feedback on their reasoning and decision-making.

Rather than relying heavily on the final written thesis, assessment is integrated into the existing workshop structure. Before each workshop, students prepare evidence of their understanding by justifying methodological choices, explaining hypotheses or comparing their own literature summaries with or without the help of AI-generated outputs. During the workshops, they discuss and defend these decisions through short oral assessments and targeted (unknown in advance) questions. Examples include:

• Why does this [specific hypothesis] logically follow from the literature?
• Can you think of circumstances under which your [specific result] might contradict instead of supporting your [specific conclusion]? What would happen if [specific key assumption] of your analysis were violated?

The aim is simple: students should demonstrate mastery of core academic skills, regardless of whether AI was involved in the process. Moss explains: “We’re not trying to ban AI. We’re creating opportunities for students to show that they can think critically with AI, not simply rely on it.”

The programme is also exploring a different balance between product and process assessment, reducing the weight of the final thesis and placing greater emphasis on demonstrated understanding throughout the trajectory.

Not everyone in the Psychology programme are convinced. Some supervisors expressed concerns about increased workload or the expectation that they would need extensive knowledge of AI. However, one of the strengths of the redesign is that it builds on existing structures rather than creating entirely new ones. The workshop framework remains in place, but its purpose shifts from primarily teaching to assessing and providing feedback. Thesis supervisors do not have to become AI experts. Instead, they continue to focus on the academic skills that matter most: critical thinking, methodological reasoning and scientific judgement.

Although the proposed changes still require formal approval from the coordinators of the various psychology programs and the exam committee, there is broad agreement among the coordinators that the current assessment model needs to evolve in response to AI.

The new approach will not be implemented yet. For Moss, however, participating in the ZLP track demonstrates how AI can act as a catalyst for educational innovation rather than a threat:

“AI is forcing us to rethink what we value in higher education. For us, that means focusing less on the outcome students produce and more on their understanding, justification and adaptability.”

As universities continue to navigate the impact of AI, the Psychology programme offers a powerful example of how educators can use the opportunities presented by AI to create more meaningful, transparent and future-proof learning experiences.