Introduction
Brain-Computer Interfaces (BCIs) based on motor imagery (MI) allow users to control external devices by imagining physical movements. Despite significant advances in signal processing, a substantial portion of users (10% to 30%) remain "BCI illiterate," unable to achieve reliable control. While research typically focuses on refining EEG decoding algorithms, this study investigates a critical but often overlooked "human-in-the-loop" variable: the experimenter. Drawing from the well-documented "experimenter effect" in psychology, this paper explores whether the gender of the person conducting the training influences a user's ability to master a BCI.
Methodology
The researchers conducted a study with 59 BCI-naïve participants (29 women, 30 men) and 6 experimenters (3 women, 3 men). Each participant underwent a two-hour session following the standard Graz protocol, consisting of a calibration phase (no feedback) followed by training runs with real-time visual feedback.
To capture the nuances of performance beyond binary classification, the study utilized Quality-Weighted Accuracy (QWA). This metric measures the "strength" of the user's mental command by considering the distance from the classifier's decision boundary. Additionally, participants' spatial abilities (Mental Rotation Test) and personality traits (16PF5) were assessed to control for known correlates of BCI performance.
The Impact of Social Context on Learning
The study's most significant finding is that BCI training is not a purely mechanical interaction but a social one. The experimenter acts as a coach, creating a psychological environment that directly affects neurological responses. A 3-way ANOVA on QWA scores revealed a significant interaction between the training run, participant gender, and experimenter gender ($F(3,156)=3.46, p=0.02, \eta^2=0.06$).
As illustrated in Figure 1, several distinct trends emerged:
- The "Women Experimenter" Advantage: Participants training with women experimenters generally exhibited more positive and stable progress.
- The "Men-Men" Hurdle: Male participants paired with male experimenters started with significantly lower performance than any other group, though they showed a sharp "catch-up" effect by the second run.
- The "Women-Men" Decline: Conversely, female participants training with male experimenters initially performed well but experienced a decline in accuracy as the session progressed.
Discussion and Conclusion
While men generally scored higher on mental rotation and participants with male experimenters scored higher on autonomy, ANCOVA analysis confirmed these factors did not account for the gender-based performance trends. The qualitative data suggests that the social support or "demand effect" provided by women experimenters may be more conducive to the sustained focus required for motor imagery.
These findings suggest that the social and emotional context of BCI training is a vital component of user success. The authors recommend that BCI protocols begin reporting experimenter gender as a standard variable. Future work may involve the use of "learning companions"—AI-driven conversational agents—to provide standardized, high-quality social feedback that could mitigate human experimenter bias and improve outcomes for all users.
Summary updated