Brain-Computer Interfaces NIRS EEG BCI: Science with Evidence, Belonging, and Collective Critical Thinking
Brain-Computer Interfaces NIRS EEG BCI: Science with Evidence, Belonging, and Collective Critical Thinking
Before we “understand” anything, let’s do a two-breath baseline—because a BCI is not an idea floating in a skull. It is a body trying to complete a movement and reclaim a channel of expression. Feel your feet. Unclench your jaw. Notice whether your attention is narrow (Zone 1) or spacious (Zone 2). That shift—tiny, bodily, immediate—is the same kind of shift BCIs must learn to ride.
The researcher’s question (the macro-question)
This paper is a tutorial-style panorama of what the next generation of BCI researchers is actually trying to solve. The authors’ guiding question is practical: what are the core bottlenecks and innovations appearing across the field, as seen through trainee-led “master classes” at the 10th International BCI Meeting?
Instead of one narrow hypothesis, the “experiment” here is the meeting structure itself: the BCI Society’s Postdoc & Student Committee organized 21 master classes where trainees presented specific work, senior researchers (“masters”) gave constructive feedback, and the broader community discussed the path forward. The meeting took place June 6–9, 2023, in the Sonian Forest near Brussels, and the master-class format was intentionally informal: two trainees present 10–15 minutes each, then a master responds, then discussion opens to attendees.
There’s a subtle methodological detail that matters: the paper’s summaries are written so that “we” refers to the trainee and their original abstract coauthors—a reminder that the collective voice in science is often a crafted “a-gente,” not a single narrator.
How the design answers the question
The authors don’t “test” trainees; they map the field by organizing the master classes into themes and summarizing: the question each trainee brought, the methods they used, preliminary results, limitations, and next steps. They group the work into themes such as speech decoding, motor imagery, pediatric BCIs, closed-loop platforms, deep learning, neurorehabilitation, sampling strategies, and novel techniques for performance improvements.
Now, to embody this (BrainLatam style), imagine yourself inside three short experiments highlighted in the paper—each one a different way of turning living physiology into a usable interface. As you read each, notice your own Eu Tensional: do you tighten and chase certainty (Zone 1/3), or do you stay curious and reorganize (Zone 2)?
1) Speech decoding with high-density ECoG: “Which features actually carry words?”
Question: Can we decode individual spoken words accurately, and which neural features are most informative?
Design: In a word-reading task with 12 unique words, each spoken aloud 10 times, the trainee used high-density ECoG and compared decoding performance across frequency bands (alpha, beta, and high-frequency band—HFB) and feature-selection strategies using an SVM classifier.
What it tries to answer: It’s not only “can we decode?” but “what is the minimal, most reliable signal structure?” The results suggest that feature-selection strategy matters a lot, and that HFB features are especially informative, with electrodes contributing strongly along ventral sensorimotor cortex.
Somatic cue: Whisper a word internally—“water.” Feel the pre-movement in tongue, jaw, throat. That is Damasian Mind in action: interoception + proprioception + action tendency. A speech BCI is an attempt to harvest the preparatory geometry of that action and turn it into communication for those who cannot speak.
2) Inner/covert speech with EEG & MEG: “Why does word-level decoding get slippery?”
Question: Can we decode inner speech (covert speech) at the word level using non-invasive signals like EEG and MEG?
Design: The paper notes both the rising interest in covert vs overt speech and trainee work aimed at collecting EEG/MEG during multiple inner-speech paradigms with initial decoding analyses—explicitly addressing the gap in non-invasive covert-speech research.
What it tries to answer: This line of work faces a classic BCI tension: non-invasive signals are safer and more scalable, but often have lower spatial specificity and signal-to-noise for subtle representations. The value of the master-class framing is that it doesn’t romanticize results; it turns difficulty into a research map: Maybe the target representation is wrong (word vs phoneme vs rhythm), maybe timing windows need rethinking, maybe training regimes must be redesigned.
Somatic cue: Try “speaking” a word inside your head without any micro-movement. Notice how quickly your system wants to either (a) subtly activate the vocal apparatus anyway, or (b) collapse into vagueness. That’s your body telling you: covert speech is not a clean, stable object—it’s a dynamic, state-dependent process. In BrainLatam terms, the same “I’m saying it inside” can be Zone 2 (open, exploratory) or Zone 3 (rigid, frustrated), with totally different qualia.
3) fNIRS/EEG movement spellers via transitions: “Extract information from the change, not only the state”
This is where the paper gives a beautiful, very “Corpo-Território” innovation.
Problem: Movement-based spellers need many commands, but non-invasive modalities like EEG and fNIRS struggle to distinguish many simultaneous movements because spatial specificity is limited.
Question: Instead of encoding simultaneous states (“left hand tense” vs “right hand tense”), can we encode transitions (“left→right” vs “right→left”) to multiply command sets and improve accuracy and ITR?
Design: In a pilot study using a NIRx NIRSport system, participants performed left-hand and right-hand tensing in transitional patterns. After filtering and ICA, an SVM achieved 81% for left vs right classification, but 92% for left→right vs right→left classification—suggesting transitions carry richer, more separable information for non-invasive fNIRS-based control.
Then the trainee proposes adapting BrainBraille (a pseudo-binary movement encoding scheme) using transitional encoding, potentially expanding the available commands and increasing the information-transfer rate (ITR) under the same constraints.
Somatic cue: Do this as an internal micro-exercise: imagine tensing your left hand, then releasing, then tensing your right. Feel the passage—the rhythm and direction. The insight is embodied: meaning is not only in “where you are,” but in how you move between states. That’s APUS / Body-Territory: cognition as trajectory, not a point.
The “collective” layer: Jiwasa in science (and why it matters)
This entire paper reads like a complex system in motion: each trainee occupies a different “position” in the flock, specializing in decoding, platforms, ethics, pediatrics, stimulation, and performance—yet the field remains coordinated through shared constraints, shared methods, and shared discussion. That is Jiwasa in an academic bioma: differentiation without fragmentation.
And here the bridge to BrainLatam politics arises naturally from the mechanisms, not from slogans. If BCI is about restoring expression and agency, then a society that keeps people in chronic scarcity pushes bodies toward Zone 1 survival or Zone 3 capture—where belonging often collapses into dogma or enemy-making. A DREX Cidadão logic (as you frame it: “feeding the living citizen-cell”) functions as a neuro-ecological hypothesis: reduce baseline metabolic threat → increase bodily flexibility → expand the chance that “TMJ / we’re together” can live in Zone 2—belonging without surrendering critical sense.
Take-home (in one embodied sentence)
This paper shows that BCI progress is not only about better algorithms—it is about building interfaces that respect the living body, and building communities that keep critical thinking compatible with belonging.
