As ASQA prepares to release new principles for on the responsible use of AI – the sector will not only need to be focussed on ensuring academic integrity is strengthened (ie students are not cheating and do have the skills that their testamurs say they do) but also that providers are keeping ahead of new and emerging uses of technology that can both help and undermine learning.
A new paper “On AI glasses and wearable AI in assessment” argues that AI-enabled glasses and other wearable devices could become a major new challenge for assessment in tertiary education. The authors, Thomas Corbin, Sue Sharpe and Phillip Dawson, from the Centre for Research in Assessment and Digital Learning at Deakin University, argue persuasively that these devices are no longer futuristic: they can display AI-generated prompts in the wearer’s line of sight, listen through built-in microphones, read materials through cameras, and do so without obvious signs to an observer. This is important because many institutions have responded to generative AI by returning to invigilated exams (particularly in higher education), supervised performances and interactive oral assessments as ‘safer’ ways to verify student learning.
In VET, providers have thus far been encouraged to engage with the resources developed by the higher education regulator, TEQSA, to help safeguard academic integrity. And for some there has perhaps been seen to be less of a need for assessment re-design because VET is competency-based and students need to demonstrate their competence in their assessment tasks.
The paper upends some of that thinking.
Its central concept is “dual transparency.” That is, the authors use two slightly different definitions of ‘transparent’ to explain why this technology is a challenge for educators.
The first definition is that wearable AI is “phenomenologically transparent” to the user – that is, it is something they are freely using and is becoming part of their everyday way of engaging with the world. It is not a separate AI tool that they consciously pick up, consult, and put down – but something that “can become incorporated into the wearer’s ongoing activity.”
The second definition is that this type of AI is “observationally transparent” that is, there are no reliable external signals that AI is being used: no phone under the desk, no laptop screen, no gaze shift, no typing, and no visible pause for a query.
The authors argue that this breaks two assumptions behind assessment security: that AI use can be separated from the student’s own work, and that AI use can be observed or detected.
They go on to point out that “common capabilities across these devices include real-time object recognition and scene interpretation, voice-activated or silent conversational AI, live captioning and language translation, contextual information retrieval, and environmental navigation, all delivered within the wearer’s sensory experience rather than through a separate screen. Interaction can be entirely silent.”
That means that in the context of assessment, AI glasses can:
- hear spoken questions
- read text in front of the wearer
- retrieve contextual information
- translate in real time, and
- present prompts or scripts within the user’s line of sight.”
The authors go on to point out that this means that students doing oral assessments can get “live assistance without consulting any visible device.” And they also point out it is not just AI glasses which students (and non-students) are incorporating into their lives.
AI is also available in wearable forms as earbuds, pendants, smart rings, hearing aids, or neural bands. The authors use the example of vision impaired people learning to use a cane to demonstrate how in time people who use wearables are likely to start to think differently – not because they are cheating but because their bodies have adapted to having the technology. They point to research showing that novice users of a cane are “acutely aware of the object itself, its weight, its position, the effort involved in sweeping it. In other words, they perceive the cane itself.” But in time as they get familiar with it “the user no longer primarily experiences the cane pressing against their palm. They feel the rough pavement at its tip, the smooth tile, the edge of a step… the cane has been incorporated into the user’s sense of their own body and its relation to the world.”
And that is where we are now in relation to wearable AI: some students are using it in their everyday lives and it is changing how they experience the world and how they learn. This represents a possible ‘second wave’ of the AI assessment challenge. The first wave, led by large language models like ChatGPT, disrupted a lot of written assessments. This next wave might undermine the whole idea that assessment can be made ‘secure’ simply by separating students from technology.
The authors argue persuasively that we need to stop trying to create “AI-free” assessment spaces and tasks, and instead think more about what counts as a meaningful demonstration of capability, what kinds of AI-supported performance are acceptable, and what students should be held responsible for in an AI-rich world?
And then beyond this paper of course lies the reality that in some occupations wearables are already proving useful to workers in how they undertake their work. The use of AI wearables in the workplace is likely to be yet another challenge for VET, as AI requires us to adapt what we teach and as it changes how students are learning.
