Assessment

AI, Course Delivery and the New Question of Authenticity

Generative AI has changed what a written submission can tell us about a learner. Assessment design has to change with it.

A few years ago, the biggest threat to assessment authenticity was a copied essay or a friend helping with the homework. Today, anyone with a phone can produce a well-structured report, a reflective journal or a case study response in seconds.

Generative AI hasn’t just added another way to cheat. It has changed what a written submission can tell us about a learner. For anyone who designs or delivers courses, that’s a bigger shift than most of us expected.

What’s actually changed

Many traditional assessment tasks assumed that producing a written response was evidence of understanding. If a learner could explain a process, analyse a scenario or write a reflection, we inferred they knew the content.

That inference no longer holds on its own. AI tools can generate plausible answers to most knowledge questions, short-answer tasks and written case studies. Detection tools haven’t solved the problem either, and false accusations carry real consequences for learners.

The result is that some assessments still look rigorous but no longer reliably measure what they were designed to measure.

What the regulators are saying

Australian regulators have moved from cautious statements to practical expectations.

In higher education, TEQSA published Assessment reform for the age of artificial intelligence, developed with a group of experts led by Jason Lodge, Sarah Howard, Margaret Bearman and Phillip Dawson. It argues that trustworthy judgements about learning need multiple, contextualised approaches to assessment, and it emphasises the learning process, appropriate engagement with AI and secure assessment at key points in a program.

The University of Sydney’s two-lane approach has become one of the most discussed responses. Secure assessments, supervised and usually in person, verify learning at important points. Open assessments allow AI use in scaffolded ways that support learning.

In vocational education, ASQA’s practice guide on assessment lists evidence generated with AI tools among the authenticity risks an RTO must consider. ASQA has also published principles for the responsible use of AI in VET. They keep decisions that affect students with qualified trainers, assessors and staff, and they make clear that AI must not compromise the conditions under which competency is achieved and demonstrated.

The common thread is clear. Banning AI isn’t realistic, and pretending it doesn’t exist isn’t defensible. Assessment design has to adapt.

Vocational education has an advantage

Competency-based training is in a stronger position than many people assume. Units of competency already require learners to demonstrate skills in conditions that reflect the workplace. Observation, practical tasks and workplace evidence are much harder to outsource to AI than an essay.

The risk sits mostly in the knowledge evidence. Written questions and case studies that can be answered from a prompt now need more care, both in how they’re designed and in how much weight they carry in the overall judgement.

Redesigning for authentic evidence

Assess the process, not just the product

Ask learners to show how they reached an answer: drafts, decisions, annotated work or a short explanation of the choices they made. AI can produce a final product. It’s much harder to fake a credible process, especially when an assessor can ask follow-up questions.

Add a conversation

A short verbal discussion about a written submission is one of the most effective authenticity checks available. Learners who did the work can talk about it with confidence. Those who didn’t usually struggle to explain their own answers.

Make tasks specific to the learner’s context

Generic questions get generic answers. Tasks that draw on the learner’s own workplace, a site visit, a practical activity they completed or data collected in class are much harder to answer with AI alone.

Strengthen observation and demonstration

Where the unit allows it, shift weight towards observed performance, simulations and practical demonstrations. They’re closer to the real job anyway.

Decide where AI belongs

Some tasks should allow AI, because using it well is now a workplace skill. Others should be secure. Be explicit about which is which, explain why, and design each task to suit. Clear rules protect learners as much as they protect integrity.

The impact on delivery

Assessment changes ripple into delivery. If learners will use AI at work, courses need to teach them how to use it responsibly: how to check outputs, recognise errors and take accountability for the final result. Trainers also need support, because many are adapting their own practice while redesigning assessment at the same time. Practical professional development on using AI in teaching and assessment can help.

AI can genuinely help with delivery too, from generating practice questions to supporting learners who need extra explanation. The key is keeping professional judgement with the trainer and assessor, which is exactly where ASQA’s principles place it.

Where to start

You don’t need to redesign everything at once. Start with the assessments most exposed to AI, usually written knowledge tasks, and work outwards. Validation sessions are an ideal place to ask the question directly: if this evidence could have been produced by AI, would we still be confident in our judgement?

Practical takeaways

  • Review which assessment tasks AI could complete convincingly
  • Assess the process as well as the product
  • Use short verbal discussions to confirm understanding
  • Tie tasks to the learner’s own context and practical work
  • Be explicit about where AI is allowed and where assessment must be secure
  • Teach responsible AI use as a skill, not just a rule

About the author

Josh Senior is Principal Consultant at Meyvn Collective, an Australian learning design consultancy. He has worked across higher education and vocational training for more than a decade, including as Head of Learning and Development for vocational education at Western Sydney University. More about Josh

Meyvn Collective

Put these ideas into practice

A free, one-hour consultation. Bring a single course or a whole program and we’ll explore your challenges, opportunities and goals, then offer practical suggestions you can use, whether you work with us or not.