Assessment in the age of AI: why practice, judgement and logic matter more than ever

Until recently, a correct answer, an error-free bibliography or a polished presentation were considered signs of strong academic competence. Today, however, they no longer show on their own how much a person truly knows. Generative tools can write reports, summarise documents or suggest solutions in seconds, and this progress forces higher education to reconsider a fundamental question: which new methods can genuinely verify that someone understands a subject and is able to apply it competently in practice.

The final result of a test or theoretical assessment still has value, but it is no longer enough as the only indicator. A flawless text may conceal limited understanding, while a less polished but better-justified presentation may reveal greater command of the subject. It is therefore worth broadening the focus to include the process, practical application and the ability to respond to unfamiliar situations.

Assessing processes, not just results

This is why higher education quality bodies are shifting their assessment methods towards multiple, contextualised and practice-based tasks. In 2025, the Australian agency TEQSA proposed monitoring progress throughout the degree, combining supervised activities with others that allow the responsible use of digital resources and verifying capabilities at meaningful stages.

This approach changes the purpose of an assessment, because asking for a document and grading only the final version is no longer enough. It is more useful to understand the previous decisions, the sources consulted and the changes made along the way. Drafts, project journals, oral defences and demonstrations provide that traceability, which is far more informative than a single finished version.

Observing the journey also prevents every task from becoming an investigation into authorship. Automated detection has significant limitations and should not sit at the centre of a reliable academic strategy. Designing more robust demonstrations of learning offers a more useful route: it shows what each person can do without depending on guessing who wrote every sentence.

Knowing how to act, not only memorise

Remembering concepts and mastering theory remain essential foundations, because without a solid base, interpreting a case or choosing between several alternatives becomes difficult. However, memorisation does not always reveal whether someone can transfer that knowledge to a specific context.

A clinical simulation, a legal scenario or a psychological intervention plan requires students to connect theory, priorities and consequences. This is where it becomes clear whether someone knows how to act when information is missing, a limitation arises or the first option no longer works.

Applied experience brings the classroom closer to professional reality, where problems rarely arrive as closed questions. They require people to analyse variables, collaborate and communicate, as well as take responsibility for the decision made.

Learning to question what is received

Artificial intelligence can produce plausible content while still including errors, overlooking nuances or presenting an uncertain claim as reliable. Knowing how to formulate a good prompt helps, but real competence appears when the information received is reviewed.

Checking a source, identifying a weak hypothesis or recognising that key elements are missing all require disciplinary knowledge. Someone who understands a subject has more resources to identify content that is convincing and incorrect at the same time.

An academic activity can make use of precisely that limitation. Giving students a generated text and asking them to identify mistakes, verify references and explain risks allows them to reconstruct a stronger proposal and develop an increasingly important skill: supervising systems that produce apparently authoritative answers.

Getting the right answer is not enough

Reaching the correct conclusion by chance does not mean that a technique or system has been understood. It is therefore preferable to make the reasoning process visible: how the evidence is connected, what has been inferred and why one option was chosen over another.

A few questions at the end of a project are often enough: what evidence guided the decision?, what would change if new information appeared?, what is the weakest part of the proposal?, which option was rejected and why?

Answering them requires students to take ownership of their own work and also reveals where the reasoning breaks down: an incomplete comparison, an artificial causal connection or an incorrectly established priority. That is where feedback becomes most useful.

How assessment works at Mundae University

At Mundae University, programmes such as Nursing are not assessed through exams, but through continuous project-based assessment. Throughout the academic year, students also work with practical cases, presentations and applied assignments, alongside the resolution of situations connected with their future profession.

This approach requires studying, but also understanding the foundations and knowing how to apply them rigorously. The difference is that learning is measured throughout the entire course, rather than through a single test. At the same time, lecturers assess whether students can interpret information, propose a solution, defend their choices and turn theory into concrete action.

The combination of formats provides a more complete view of progress. It also prepares students for professional environments where they will need to collaborate, use digital resources and adapt to unforeseen situations, as well as take responsibility for the decisions they make.

AI can and should be integrated as support in certain tasks, always under clear conditions. It can help organise ideas, explore approaches or highlight aspects that need further review. However, responsibility and judgement regarding the submitted content still belong to the person whose name appears on the work.

Clear rules for using AI

An effective academic policy clearly defines which uses are permitted, which ones must be declared and which parts students must complete independently. Generic rules are insufficient, because what is reasonable in one subject may not be appropriate in another.

In practice, only a few elements may be needed: a record of the AI prompts used, a note explaining what was changed or how the information was verified. The aim is not to collect screenshots, but to ensure that human oversight remains visible behind the work.

In this regard, UNESCO understands AI literacy as something that goes far beyond knowing how to use a tool. It involves understanding what it can contribute and where it fails, as well as using it with ethical judgement.

Educating students to develop judgement

Higher education still needs to measure knowledge, but through more varied and, above all, more practical assessments. An exam reveals what someone can memorise or recall, but an applied case, a simulation or a supervised intervention shows whether they can use that knowledge when it is needed. It is in this space, where doing matters as much as explaining, that the competence transferred to professional practice becomes visible.

For that reason, applied work should not be an additional element of assessment, but its core. It is in the face of a real problem, with its limits and pressures, that we can see what students are able to do independently.

Assessment in the age of AI requires attention to what students can actually do, not only to what they submit. Practice demonstrates whether knowledge can be turned into action; judgement shows whether reliable content can be distinguished from weak material. Logic reveals whether someone can respond appropriately and according to the context. When these dimensions sit at the centre of assessment, the university verifies something no tool can fully assume: the individual capacity to understand, decide and take responsibility for one’s own work.