Towards User-Centered Design for Motor Imagery Brain-Computer Interface: From MI-BCI Illiteracy to MI-BCI Unfamiliarity

Brain-Computer Interface Illiteracy (BI) has challenged BCI research for decades. Some users negatively impact performance, regardless of decoding robustness. This issue is especially evident in the Motor Imagery (MI) paradigm, particularly during offline calibration without feedback. The literature lacks a comprehensive definition of BI. Extensive experiments with many subjects are needed to thoroughly analyze BI's effect in offline MI decoding.This paper proposes a robust definition of BI by investigating decoding performance differences between MI and motor execution (ME) of the same task. We introduce a BI experiment using the largest MI-BCI dataset. By comparing BCI decoding accuracy for the same users during both movement execution and imagined movement, our study offers a new perspective on understanding and identifying BI. We propose that users' neural networks are less familiar with imagining than executing movement, causing BI in MI but not in ME. We evaluate whether BI relates to subjects' training and behavior during MI tasks. We suggest ruling out other factors like BCI setup, noise, and brain structure differences among subjects. This paper offers a new perspective on BI for the MI-BCI research community. Empirical evidence shows BI depends on subjects' familiarity with MI, influenced by proper training.