A study into eye tracking in VR has found that changes in pupil size and blink rate can identify shifts in mental workload, raising the prospect of XR training that adjusts to an employee’s ability during a session.
Published in Scientific Reports, the research assessed whether eye tracking built into a consumer VR headset could act as a marker of cognitive load during a graded working-memory task.
Its machine-learning models distinguished between low and high workload with up to 88 percent accuracy among healthy volunteers.
Speaking to UC Today, Dr Jordan Tsigarides, Honorary Research Fellow at the University of East Anglia and Senior Clinical Fellow in Rheumatology at Norfolk and Norwich University Hospital, said the opportunity is to make VR experiences more responsive to the individual.
“It’s about being able to take what is an impactful virtual reality experience and supercharging it by being able to provide the information and the content that that person needs at that time,” he said.
The study was conducted in the context of chronic pain, but the underlying approach has potential relevance for enterprise learning.
Rather than giving every employee the same VR module at the same pace, workplace XR platforms could eventually use eye tracking alongside task performance and learner feedback to adjust the experience in real time.
The research involved 84 volunteers, including 42 people living with chronic musculoskeletal pain conditions, such as arthritis and fibromyalgia. Participants undertook increasingly demanding N-back memory tasks in VR while researchers collected data on pupil size, blink rate and gaze direction.
As task difficulty increased, participants’ pupils widened and their blink rates increased. The results indicate that eye signals could offer an accessible way to estimate cognitive workload in headsets that already support eye tracking.
Targeting the Productive Middle Ground
The goal of adaptive XR is not simply to make training more difficult. Instead, it is to keep people in a productive range where the task is engaging enough to hold their attention without becoming overwhelming.
Tsigarides described this through flow theory, where challenge is matched to a person’s ability and their mental capacity at that time.
“We want people to be on task at all times,” he said.
“That looks like having a workload which is not too high, that you are cognitively overloaded, but not too low where you are under-engaged and unmotivated.”
In a workplace setting, this could mean an XR application identifying signs that a learner is struggling with a scenario and responding by slowing the pace, offering prompts or reducing complexity. Conversely, if the learner is under-challenged, the system could remove guidance or increase task difficulty.
That approach could be valuable in areas where training must accommodate varied experience levels, including technical onboarding, safety exercises, complex equipment instruction and simulated customer interactions.
However, the research does not establish a universal threshold for the right level of cognitive workload. A task that is manageable for one employee may be too demanding for another, while the same individual’s capacity may vary from one day to the next.
Combining Signals Could Improve Accuracy
Eye tracking is a useful candidate for adaptive training because it is already available in several commercial VR and mixed-reality headsets. Tsigarides said pupil size change in particular could provide a strong indication of cognitive workload during a task.
But eye tracking should not be treated as a standalone judgement on an employee’s ability or state of mind.
“You would need the ability to measure that person’s subjective experience of that task and have some sort of understanding over time of what that person is able to achieve, because everyone is different,” Tsigarides said.
