Mind, Brain, and Brain Computer Interface

BCIs have shown great promise in helping society: from movement disorders such as paralysis, to human augmentation, gaming and communication applications. But despite significant progress, the reliability and overall effectiveness of BCI technology remains limited. EEG-based BCIs are barely used outside laboratories due to low reliability in real-life conditions (Benaroch, Sadatnejad, Roc et al., 2021). How might an improved understanding of mind — and brain — help improve the effectiveness, reliability and ease of use of BCI technology?

The brain is obviously a central component. BCIs are controlled via the user’s brain signal. The signal is analyzed, processed, decoded, and classified in real time to control a user’s intended movement, or other external action.

But, what controls the user’s brain signal? Some might say task determines the signal: its location, frequency, amplitude and other characteristics.

I argue task is a secondary variable. The primary, and most direct, variable affecting the brain is the mind. The user’s brain signal is a direct reflection of his or her mental processes — perception, emotion, goals, imagination, attitude, motivation, level of frustration, fatigue, attention, intention etc. Yet, the fact that the user has to volitionally modulate his/her own brain activity (via their state of mind) to operate a BCI has been mostly ignored (Chavarriaga, Fried-Oken, Kleih et al., 2016).

To define and classify the brain signal requires bringing the mind to center stage. Granted, the user’s state of mind is influenced — often strongly — by the task in which it is engaged. But user mental states are affected by many other variables beyond task. Externally, the mind is affected by environment, situation (work/home, solitary/social…), activity, recent life events (good and bad) and other context. In addition, internal factors — i.e. the mind — affect the mind itself. These include level of excitement, reward, frustration, motivation, fatigue, motivation, and degree of concentration.

Though a plethora of external and internal variables affect the user’s mind, the good news is it can be controlled to a large degree. If a user’s state of mind is stabilized, matches the classifiers being used, and the brain signal processed and classified accordingly, the desired action will occur. Mind Target “Hit” → Functional Neural Network Activity → Brain Signal → Decoding & Classification → (Desired) External Action.

For example, imagine a user who could focus their mind consistently and exclusively on an imagining and intention to “move the cursor to the left.” In this case the corresponding functional neural network activity, and brain signal, would then be classified as “move cursor left.” This in turn triggers the desired action.

Better user mind control is only one piece of the puzzle. A more central issue is how to define the mind in the first place. What if the mind could be better-defined? What if it could be defined with greater accuracy, precision, and comprehensiveness than the current cognitive neuroscience paradigm allows? This would inevitably lead to brain signal classification that is more accurate, precise, and comprehensive as well.

Underlying this mind-based view of brain signal classification is a more fundamental issue: the initial defining of mind/brain signal “targets,” or classifiers. To label or classify a brain signal requires a pre-defined classifier. My argument is that a group could use better-defined mind, and corresponding brain signal, signatures as “targets.” The entire team — BCI designer, tester, trainer, and user — could use these to improve their performance. For example, for the user, target states of mind could be defined that are more clear, precise, and personalized; desirable to strive toward: easier to hit (strongly and consistently); and adjusted for specific movements, situations, and other context.

The contents of the conscious mind, contrary to what many in the brain science community might believe, are not mysterious or even difficult to understand. They are the same as our experience of life. All of our awareness, moment by moment, is encompassed by our mind, as it changes (or stays the same) inside the brain. The good news is these contents can be defined — or at least summarized – quite accurately.

The MA Method defines the mind as a set of active general memories, based on past experience. For example, when a user tries to moves a cursor to the left, a “me moving a cursor left” memory will activate. This memory is comprised predominately of (visual & somatosensory) perception: of “my arm & hand moving to the left,” a computer screen, a cursor (a range of cursor size & shapes), and cursor moving left (a range of cursor speeds and directions). These mental contents will activate as part of the larger memory “me moving a cursor left” (itself a general memory).

A memory range = a general memory. For example the general memory “an apple” includes a range of apple characteristics: shapes, sizes, colors, tastes, etc.

The neural correlate of a general memory, I argue, is obvious: a functional neural network (FNN) range. Both are highly associative, function — simultaneously — as signal transmitters and receivers, and dominate their respective (mind/brain) systems.

The user’s state of mind = (mostly) an active general memory set = (mostly) an active set of FNN ranges = (most of) the corresponding brain signal.

Why classify the brain signal based on task when it can be more accurately classified based on the mind? In fact, I argue it’s possible to define, connect, and weight the active memory set — its components and associations — for any common movement & its context. Movement context would include the person, occupation, lifestyle, movement situation (environmental, social, work/home/lab…), and stage of learning (beginner, intermediate, or advanced).

Overall, I argue brain signal targets — based on a clear and accurate definition of the mind — have the potential to greatly enhance BCI design, testing, learning, and use.

References

Benarock, C., Sadatnejad, K., Roc, A., Monseigne, T., Pramij, S., Mladenovic, J., Pillette, L., Jeunet, C., Lotte, F. (2021). Long-Term BCI training of a tetraplegic user: Adaptive Riemannian Classifiers and User Training. Frontiers in Human Neuroscience, Brain-Computer Interfaces v. 15.

Chavarriaga, R., Fried-Oken, M., Kleih, S., Lotte, F., Scherer, R. (2016). Heading for new shores! Overcoming pitfalls in BCI design. Brain Comput Interfaces (Abingdon), 4(1-2), 60.

Comments are closed.