Securing assessment in an AI-enabled world: a program-level approach for UNSW teachers

The rapid rise of Generative AI over the past few years has fundamentally reshaped the landscape of higher education assessment. Tasks like essays, reports and some forms of problem solving, that once provided reliable evidence of student learning, can now be outsourced to Generative AI. In most instances, this outsourcing can be difficult to detect. This new reality has forced us to ask: how can we assure that students are genuinely achieving program learning outcomes in an AI-enabled world? 

The answer, increasingly, lies not in policing GenAI use or tightening controls around individual tasks, but in rethinking assessment at the program level. 

The challenge: validity in a low-control environment 

As recent research (Fleckenstein et al., 2024; Kofinas et al., 2025) highlights, AI-generated work can be indistinguishable from authentic student submissions in unsupervised or low-control contexts. This undermines the validity of traditional assessment formats and raises concerns about academic integrity, fairness, and the credibility of qualifications. 

However, the issue is not simply about “cheating”. It is about evidence. How do we ensure that the work we assess truly reflects student learning? And more importantly; how do we design assessment systems that remain meaningful and defensible in the face of technological change? 

Moving beyond the course-level mindset 

A key shift is moving from isolated, course-level assessment design to a coordinated, program-level approach, as we are doing at UNSW. Programs should ensure that learning outcomes are assessed securely at multiple, meaningful points across the degree. 

This involves aligning assessment with Program Learning Outcomes (PLOs) and ensuring that students demonstrate their capabilities in varied and robust ways over time. A thoughtfully designed mix of assessment tasks across courses can create a cumulative, reliable picture of student achievement. 

What makes an assessment “secure”? 

In this context, “secure” assessments do not mean those that are surveillance-heavy or restrictive. Rather, “secure” refers to assessment design that is difficult to outsource to AI or third parties and provides strong evidence of student learning. Think of it as a continuum, and remember that no assessment type is completely secure – students have in the past managed to evade controls if they are motivated enough. 

Assessments are likely to be situated on the more secure end of the continuum if they have at least the first of the following features, plus one other:  

1. Being there (embodiment): Tasks that require physical or synchronous presence, such as labs, performances, or in-class activities.
2. Defending thinking dynamically (dialogue): Interactive oral assessments, viva voces, or interviews where students must explain and justify their reasoning in real time.
3. Showing change over time (process): Iterative tasks that capture the development of ideas, drafts, and feedback.
4. Owning judgment (accountability): Tasks where students must make and justify decisions, demonstrating ownership of their work.  

Importantly, the most effective assessments often combine two or more of these elements. 

Expanding the repertoire: beyond traditional exams 

While invigilated exams remain one form of secure assessment, they are far from the only option. In fact, an overreliance on invigilated exams can limit the diversity and authenticity of student learning experiences. 

Alternative approaches include: 

• In-person lab work with verification processes
• Studio-based assessments in disciplines like art, design, and performance
• Clinical or practicum observations in professional programs
• Fieldwork with validation of presence, and production of artefacts
• Structured, supervised in-class workshops or problem-solving activities  

These formats not only enhance security but also align more closely with real-world practice in many disciplines. 

The power of dynamic dialogue 

One of the most promising strategies is the use of interactive oral assessments. These can take the form of viva voces, project defences, or live problem-solving sessions. 

Such assessments allow academics to probe student understanding with follow-up questions like “Why did you choose this approach?” or “What would you do if the conditions changed?” This dynamic interaction makes it difficult to rely on AI-generated responses and provides more insights into student thinking. 

Designing for the future 

For the higher education sector, the implications are clear: assessment design must evolve. This does not mean abandoning all existing practices, rather, it requires the integration of new approaches that actively strengthen the validity and reliability of evidence. 

At a practical level, this might involve: 

• Collaborating with colleagues to map assessment across a program
• Embedding more low-stakes, process-oriented tasks throughout a course
• Incorporating interactive oral components into major assessments
• Redesigning laboratory sessions or tutorial workshops so that students’ actual work can be assessed 

It is important to emphasise that these changes should be guided by pedagogy, not panic. The goal is not to “outsmart” AI, but to design assessments that genuinely capture learning. 

A collective responsibility 

Securing assessment in an AI-enabled world is not the responsibility of individual academics alone. It requires coordinated effort at the program level and across the university, including support for curriculum redesign, professional development, and policy alignment. 

By taking a program-level approach and embracing a broader repertoire of assessment strategies, the higher education sector can not only address the challenges posed by AI but also create richer, more authentic learning experiences for students. 

In doing so, we move closer to a system where assessment is not just secure, but meaningful, fair, and future-ready. 

Interested in finding out more? Contact us at assessmentandfeedback@unsw.edu.au