Addressing the Human Bottleneck in Brain-Computer Interfaces

You are reading version 1 of this direction, frozen 2026-10-04. The latest version is version 2.

Version 1 of 266 papers81 conceptsfrozen 2026-10-04

For decades, the primary challenge in Brain-Computer Interface (BCI) research was viewed as a signal processing problem: if we could just build a better decoder, we could read any mind. However, a persistent "usability gap" has revealed a more fundamental hurdle. Between 15% and 30% of users—and in some studies up to 50%—are unable to achieve reliable control over a BCI even after extensive training.

This phenomenon, variously termed BCI illiteracy or BCI inefficiency, suggests that the bottleneck is not just the machine, but the human in the loop. Researchers like Camille Jeunet and Fabien Lotte argue that current systems fail because they treat users as static signal generators rather than active learners who must acquire a complex new skill.

Predicting Performance Beyond the Signal

If a significant portion of the population cannot use a BCI, the first step is understanding why. Traditional neurophysiological markers, such as the strength of a user's alpha rhythms, are often inconsistent predictors of success. Instead, researchers are looking at the "human factors" that precede the first trial.

Camille Jeunet, Bernard N’Kaoua, and colleagues have shown in Predicting Mental Imagery-Based BCI Performance from Personality, Cognitive Profile and Neurophysiological Patterns that a user’s spatial abilities and personality traits—specifically tension and self-reliance—can predict BCI performance with high accuracy.

Beyond personality, the physical "wiring" of the brain plays a role. Carmen Vidaurre found that the functional connectivity between motor and sensory regions during rest can forecast how well a novice will perform. Even passive stimuli can be predictive: Sébastien Rimbert and Fabien Lotte demonstrated that a quick electrical pulse to the wrist (Median Nerve Stimulation) reveals neural reactivity that correlates strongly with future BCI control.

The User as an Active Inference Agent

To solve BCI illiteracy, we must move away from "trial-and-error" training. Recent work by Côme Annicchiarico, Jérémie Mattout, and Fabien Lotte proposes modeling the user through the lens of Active Inference.

In this framework, the user is an agent building an internal "map" of how their thoughts translate into BCI commands. If the feedback from the BCI is too noisy or the training protocol is too rigid, the user cannot update this map effectively. This perspective shifts the goal of BCI design from "decoding the brain" to "facilitating the user's internal learning process." For example, Hanwen Wang and Yu Qi developed a joint learning framework where both the machine decoder and the human user adapt to each other simultaneously, guiding the user to generate more discriminative signals.

Replacing Abstract Imagery with Familiar Tasks

Standard BCI protocols often ask users to perform abstract tasks, like "imagine moving your right hand." For many, this is cognitively demanding and produces weak neural signatures. A growing trend is to replace these abstract cues with more engaging, culturally or physically grounded tasks.

  • Writing Imagery: Yu Zhang and Andrzej Cichocki found that imagining writing Chinese characters significantly boosts classification accuracy compared to simple motor imagery.
  • Multimodal Cues: Jigang Tong and colleagues combined silent reading and writing imagery, which leverages more distributed neural networks and improves signal-to-noise ratios.
  • Tactile Assistance: Yucun Zhong and Lin Yao used wrist vibrations to "prime" the motor cortex, helping users generate clearer signals during calibration.

The Social and Emotional Dimension of Training

Learning to control one's brain waves is a lonely and often frustrating task. Léa Pillette and Fabien Lotte have pioneered the use of Artificial Learning Companions like PEANUT, an agent designed to provide social and emotional support.

While these companions may not immediately boost raw accuracy, they improve the perceived usability and learnability of the system. Similarly, Jelena Mladenović has explored biased feedback—intentionally showing users they are doing slightly better than they actually are—to manage anxiety and maintain the "state of flow" necessary for high-performance neural control.

By treating BCI control as a pedagogical challenge rather than a purely technical one, this research direction aims to make neural interfaces accessible to the "illiterate" 30%, transforming BCIs from specialized tools into universal technology.

Papers in this version

Showing 60 of 66 papers.

History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 92 papers
  2. Version 1 · 2026-10-04 · 66 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_bci_illiteracy_42d853d1,
  title = {BCI illiteracy},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/bci-illiteracy/rdr_42d853d18fa14669a3034a6ab8a08f1e},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_addressing_the_human_bottleneck_in_brain_42d853d1_v1,
  title = {Addressing the Human Bottleneck in Brain-Computer Interfaces},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/addressing-the-human-bottleneck-in-brain-computer-interfaces/rdr_42d853d18fa14669a3034a6ab8a08f1e/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}