This week my independent reading moved further into the relationship between immersive technology, physiological data and responsive systems. Rather than looking for another airport-specific intervention, I followed references from the adaptive VR research to understand some of the technology sitting behind the idea.

Integrating Artificial Intelligence with Virtual Reality for Cognitive Therapy

This paper was useful because it brought together several areas I had previously been researching separately: VR, physiological monitoring, biofeedback and artificial intelligence. It describes systems where information such as heart rate, HRV, EEG and EDA can be used alongside machine learning to adapt a virtual environment.

The most useful idea for me was the feedback loop between the person and the environment. The system does not only present an experience; it can potentially monitor how somebody responds and change what happens next. This gave me a much clearer technical precedent for the idea of personalised regulation.

However, the paper also made me question whether VR itself should remain at the centre of the project. If the important principle is that physiological information can influence a response, perhaps the valuable part is the responsive system rather than the headset.

Ge et al. — Real-Time Biofeedback in Immersive VR

I then read Ge et al. (2022), which looks at how heart-rate and EEG information can become part of a VR performance. This was useful from an interaction-design perspective because it demonstrates that biosignals can be translated into visual and auditory changes rather than remaining as medical-looking numbers or graphs.

What was equally useful was that more feedback did not automatically improve the experience. Continuous heartbeat audio became distracting for some participants. This feels particularly relevant to my project because any technology intended to support sensory regulation could accidentally create more sensory demand if the feedback itself is badly designed.

This reinforced the idea that the question is not simply whether a system can respond, but how it should respond without adding another layer of cognitive or sensory pressure.

Tayarani-N. and Shahid — Detecting Anxiety via Machine Learning Algorithms

I followed another reference into Tayarani-N. and Shahid's literature review on machine-learning approaches to anxiety detection. The review covers a wide range of data, including physiological signals, EEG, facial information, voice, movement and information collected through wearable devices.

For my project, the most relevant sections were those on physiological signals and wearables. The review describes studies using ECG, heart rate, HRV, respiration, electrodermal activity and other signals as inputs into machine-learning models.

This is useful because it demonstrates that changes in the body can contain patterns associated with anxiety. At the same time, the paper repeatedly highlights problems with context, individual variation, reliability and the assumption that one physiological marker can cleanly identify anxiety. That limitation is extremely important for my project.

Bahameish et al. — Heart Rate Variability in Autism and ADHD

I also began looking more specifically at HRV through Bahameish et al. (2026). Their systematic review and meta-analysis included 20 studies and 2,223 participants, examining HRV during cognitive tasks in autistic and ADHD populations.

The paper is useful because it gives me a much more autism-specific basis for understanding HRV, but it also stops me from treating it as a simple measure of overload. The findings vary between studies and the authors identify substantial methodological inconsistency and heterogeneity.

This is beginning to make the role of physiological sensing much clearer. A biosignal may provide useful information about a changing internal state, but it cannot independently explain why that change is happening.

What I Am Taking Forward

Across these readings, the strongest idea is that physiological data could become one input within a responsive accessibility system, but it should not be treated as a diagnosis of what the traveller is feeling.

The design opportunity therefore seems to sit between three areas: sensing, interpretation and response. I now need to understand each part separately before deciding what the final technology might look like.

Key Terminology

Biosignal — physiological information produced by the body that can be measured, such as heart activity, electrodermal activity or brain activity.

Physiological monitoring — collecting bodily signals over time to provide information about changes in physiological state.

Machine learning (ML) — algorithms that identify patterns within data and can be trained to classify or predict particular outcomes.

Affective computing — computing concerned with recognising, interpreting or responding to human emotional states.

Biofeedback — using information collected from the body as feedback for the user or a responsive system.

Heart Rate Variability (HRV) — variation in the timing between heartbeats, often used as an indirect indicator of autonomic nervous system activity.

Electrodermal Activity (EDA) — changes in the electrical properties of the skin associated with physiological arousal.