READ / RESEARCH · BRAIN–COMPUTER INTERFACES AND NEUROADAPTIVE DESIGN

This chapter explains the decisions involved in designing a brain–computer interface (BCI): a system that uses measured brain activity to control or adapt technology. It covers the purpose, users, sensors, interpretation, feedback and evaluation. For my project, its value is a way of thinking through the whole interaction, rather than evidence that a particular wearable reduces sensory overload.

Why I looked at this

My notes move between the technical system and the person using it: brain signals, passive sensing, neuroergonomics, feedback, individual differences and consent. Together, they bring me back to a central design question: what should happen after the system detects a change?

I am exploring biosensing to support autistic adults during airport travel. Reading about BCIs helps me question how a system interprets information and responds, without assuming that my final design must measure brain activity.

Passive sensing and the role of the user

My notes distinguish passive BCIs from reactive BCIs. The chapter separates three approaches: active BCIs use intentionally generated brain patterns; reactive BCIs use responses to external stimuli; passive BCIs estimate states from activity that is not deliberately produced as a command.

This makes passive sensing relevant to my thinking. The person would not need to perform an extra mental task simply to communicate with the system. However, “passive” should never mean that monitoring happens without their knowledge or choice.

My note about a neuroadaptive system raises a further question: once a state is estimated, what adaptation would actually help? Detecting a change and selecting a useful response are separate design problems.

A BCI specifically involves brain activity. A wearable using heart rate or skin conductance alone would be a physiological sensing system; I should not call it a BCI simply because it responds to the body.

Neuroergonomics: considering the person in context

My notes describe neuroergonomics as bringing together neuroscience, cognitive psychology, engineering and human factors to understand the brain in relation to tasks, systems and environments.

For my project, this shifts my attention towards the whole airport experience. I need to consider what someone is doing when a change occurs: reading a departure board, moving through security, waiting in a queue or processing unexpected instructions.

The note comparing human–machine and human–human interaction also matters. A wearable would sit within an existing network of staff, companions, signs and announcements. I would want to investigate whether it makes those interactions easier or introduces another thing to manage.

Choosing signals without overstating what they mean

My notes identify EEG and fNIRS alongside additional signals such as skin conductance and ECG. EEG records electrical activity; fNIRS estimates changes in blood oxygenation. Their suitability would need to follow the question being asked, rather than the appeal of a particular technology.

I noted EEG’s portability, speed and relative affordability. Those are reasons to investigate it, not enough to establish that it is the right sensor for this project. Fit, comfort, setup and signal quality during movement would also matter.

Another note highlights muscle artefacts and the difficulty of mapping brain activity to a particular state. This is a useful challenge to any simple “signal equals overload” assumption.

For my design, uncertainty needs an appropriate response. If the system cannot interpret a change confidently, I would explore whether it should wait, ask the user or provide no intervention, rather than presenting its estimate as a fact.

A measured signal is information to interpret. It should not override the person’s account of how they feel.

Feedback: enough information without another burden

My note favours progressive feedback over a binary response. In the chapter, this concerns BCI training: graded feedback can communicate more about progress or difficulty than a simple present/absent signal.

For my project, that prompts an interaction question rather than a direct prescription. Would someone prefer a subtle cue, a short explanation, a choice of responses or no feedback at that moment?

I should not translate “progressive feedback” into automatically increasing pressure, cooling or stimulation. The information a system communicates and the physical support it delivers have different purposes.

Individual differences and evaluation

My notes ask whether a BCI could interpret signals differently across age, gender or culture, and identify relevant participant information such as age, gender and education. I am treating this as a question about who a system works for, rather than claiming that every difference produces a predictable signal.

For my project, I would also want to understand sensory preferences, familiarity with wearables and the person’s preferred ways of regulating. These should help shape design decisions without reducing someone to a demographic category.

I noted computer simulations, immersive environments and XR as ways of representing realistic tasks. For development, a simulated airport scenario could let me explore cue timing or competing demands. It would still need to be distinguished from an actual journey, and the simulation itself would need to be acceptable to participants.

The challenge I wrote down about moving passive BCIs outside the laboratory is especially useful. A successful demonstration in a controlled setting would leave questions about movement, changing environments and ordinary use unresolved.

Consent and the boundary between assistance and control

Two of my strongest notes ask whether users can understand continuous brain monitoring and where helpful assistance becomes intrusive control.

These questions should shape the interface from the beginning. I would want the wearer to understand what is measured, what is inferred, what is stored and what the system can change. Pausing sensing and declining a response should be straightforward.

I also need to keep agency visible. If a traveller rejects a suggestion, that should be treated as a valid choice, rather than a failure to cooperate with the system.

The system may estimate that support could help. The person should retain the authority to decide whether they want it.

What I am taking forward

I am keeping this chapter as a design framework for a responsive system. My priorities are to define the intended support, choose measurements that serve that purpose, account for uncertainty and test the interaction with the intended users.

The most useful outcome is a clearer set of questions about the relationship between sensing, interpretation, response and control. Adding brain sensing would only be justified if it contributes something necessary and usable.

Source

Lotte, F., Dreyer, P., Rimbert, S., Trocellier, D. and Welter, M. (2025). ‘How to build a brain–computer interface from beginning to end’. In Neural Interfaces, Chapter 10, pp. 115–141.

This reflection brings together my photographed reading notes, the chapter introduction and checked manuscript sections on BCI types, signals and feedback. The page labels on my sticky notes have not been converted into published page citations because the available manuscript uses different pagination.