The first in a series outlining my concerns with the “two-lane” approach to assessment security. Here, I explain how this approach rolls back years of hard-won progress towards inclusive education.

Universities in Australia are beginning to converge on a strategy for assessment security in the so-called “Age of AI”. It’s very simple: two lanes.

Some assessments are “open”, meaning the conditions for students completing them are unrestricted. While scaffolding, recommendations and guidance should be offered, ultimately it’s not possible to ban generative AI in an open assessment. There’s just no way to tell for certain. And that’s ok! It’s the real world, ain’t it? These tasks are assessment for learning, so they’re about development, not proof.

Other assessments are “secure”, meaning the conditions for student work are tightly controlled and supervised. These involve tasks like live practical and written exams where the use of generative AI can be proscribed or limited to strict and specific applications. These tasks are assessment of learning, so they can be used to prove students have learned what they were supposed to.

This solution is proposed to address the tidal wave of cheating that is rendering university degrees worthless as currency for the job market.

However, I consider this approach and its rationale to be disingenuous, misguided and unethical. I’ve dedicated some time to articulating the reasons why.

In this article I’ll focus specifically on the issue of inclusive education, as it is an area I care about deeply.

It will roll back inclusive education practice.

The reality is that, in a two-lane approach, the only type of assessment that will count for defending award of degrees is the “secure” type, which are characterised by the University of Sydney as assessment of learning. “Open” assessments are assessment for learning, which actually means active learning with feedback. According to the NSW Government, assessment for learning is formative. This kind of assessment is for the purpose of enabling teachers to observe their students’ progress, to determine how and what and whether learning is happening.

But in the Age of AI, if we need to be sure it’s the student doing it, assessment must be “secure”.

At the University of Melbourne, for example, at least 50% of every subject will need to be assessed by “secure” methods. That means it’ll be high-stakes, in-person, invigilated and tightly delimited.

I have been told again and again and again that this doesn’t just mean exams. So what does it mean? According to the University of Sydney, it could mean:

• exams — written, practical or oral

• tests — written, practical or oral

• in class — written, practical or oral

• placement, internship or supervision — task, observation or clinical exam

The University of Melbourne’s list of four secure assessment types also includes “performance”, in the sense of a live music or dramatic performance.

Leaving aside the fact that the above lists at least half a dozen formats of exam, what we’re looking at here is a list of live invigilation methods requiring performance under pressure in a time- and resource-restricted environment.

It is well understood that these are conditions which cause many students to experience severe anxiety and which specifically disadvantage those with disabilities like vision impairment, motor impairment, dyslexia and ADHD.

While universities do offer provisions like reasonable adjustment and special consideration, these practices are administratively and logistically complicated, involve extra time, extra resources and delays to student progress in the meantime.

They also require students not only to disclose but to provide (usually medical) proof of their needs. Some students fake these documents. Many, many more opt not to disclose at all due to stigma and administrative burden, thus receiving no support and compromising their own educational outcomes.

Inclusive education advocates have fought for decades to move us away from a model reliant on self-disclosure and special treatment. A student shouldn’t need a diagnosis for their experienced needs to be respected, but they will need something tangible to protect themselves as the university assessment environment grows increasingly hostile towards students seeking flexibility.

Ultimately, however, disclosure only helps if the conditions can be adjusted. The question remains whether secure assessment types will be possible to adjust meaningfully, at the scale required, to accommodate the real needs of the 1.6 million or so university students who may now need to request them.

To be clear, I don’t think any of these “secure” assessment methods are inherently terrible. Obviously, an oral assessment is amazing for assessing someone’s communicative and expressive skills. Performance is literally essential if you’re assessing performing arts. And exams… well, I really do not like them, but they make a heck of a lot more sense in foundational subjects with a lot of declarative knowledge requirements.

The question is whether a timed, invigilated environment actually does make sense for what’s being learned (and why). Otherwise, it is an invalid method that will produce confounding results in student performance, such as:

• bodily symptoms like sweating, heart palpitations and nausea

• anxiety-induced cognitive issues like impaired recall, reasoning and speech fluency

• observer effects, such as changes in student behaviour while being watched, as well as observer bias

• inability to physically write at speed due to issues with writing by hand or perception difficulties like dyslexia.

Further, time-limited assessments produce measurably skewed results based on student characteristics. Men perform better on exams than women. There’s also considerable evidence that exams disadvantage students of lower socio-economic status, non-dominant race and ethnicity (particularly Indigeneity), and disability. And, of course, when all assessment is required to take place in person, this disadvantages rural, remote and working students (including those who do caring and parenting work). Attending all sessions in person is easy when campus is in the city, you live in the city, and you are able to control your own schedule. If not — not.

Until about 18 months ago, we (educators) were practically unanimous on this point. High-stakes invigilated assessment is both invalid and inequitable, unless it is assessing capabilities for which time pressure is essential.

Now it seems as though universities across Australia are using the AI moment to abandon moves away from inequitable conditions. Suddenly, the invigilated assessment is a proud bastion of academic integrity and assessment security, assuring that only students who “deserve” their degrees can access them.

Without robust strategies for ensuring inclusive assessment conditions, far more versatile and relevant options for “secure” assessment, and provision of massively expanded support infrastructure for assessment, the “secure” lane presents a profound risk to higher education equity across Australia and beyond.