AI Has Changed Assessment; Higher Education Has Not Caught Up

AI has already transformed how learners think, solve problems and produce knowledge. Higher education assessment, meanwhile, remains largely unchanged. A clear gap is opening between how students actually work and how they are evaluated.

Students use AI to accelerate learning, iterate rapidly and reach knowledge beyond the formal curriculum. In many cases they are becoming more adaptive and efficient than the systems designed to assess them. Yet they stay discreet about it, largely because current models still penalise or misread AI-assisted work.

A fundamental misalignment

We claim to value critical thinking and real-world readiness, yet we continue to assess controlled, decontextualised outputs.

Frameworks such as learning analytics and computerised adaptive testing show that more dynamic, process-oriented assessment is possible, while automated scoring systems highlight both the scalability and the risks of current approaches. At policy level, the OECD and UNESCO keep calling for competency-based, ethical and transparent systems. Institutions remain slow to adapt.

If this continues, assessment risks becoming ever more performative, while students become more strategic and less transparent.

Final thought: the real issue is not that students are using AI. It is that they may already be learning and evolving faster than the systems meant to measure them.

Beyond Mock Exams: Audio Briefings for the AI Era

Mock exams are not a tenable method of assessment in the age of AI and digitalisation. As an educator in multilingual education and AI-assisted pedagogy, I am challenging the memorisation paradigm that has dominated our classrooms for too long. Mock exams belong to a different era, rooted in the standardisation and industrial paradigm that treated education like assembly-line production. Here is why I reject them, and what I do instead.

Why mock exams fail

Mock exams promote rote cramming and turn students into prisoners of repetition. They kill cognitive dynamism and ignore how multilingual skills, such as contextual fluency and cultural adaptation, demand real-world application rather than one-shot tests from an outdated industrial model.

My approach: progressive mastery, in class and remotely

I favour continuous activities and mini-projects that build competencies situationally. Students apply their multilingual skills through collaborative tasks, simulations and AI-enhanced workflows, developing a deep, layered understanding along the way.

The innovation: NotebookLM audio guides

Today I used NotebookLM to create an audio overview of the final exam structure: an elaborate narrative linking each question to the full network of lectures. Students engage multisensorially, listening, associating content (the module on multilingual practices and ethics, for example) and reflecting. This immersive preparation boosts retention without a single mock exam.

In the AI era of multilingual education, tools like these augment cognition, reduce anxiety and spark motivation. Yet existing structures and policies, especially those plagued by micro-management and control-mania, do not let such illuminating methods be nurtured and implemented.