280 | Algorithmic Fairness in Brain-Computer Interfaces for Motor Imagery Detection

Theoretical and Computational Neuroscience

Author: Bruno Jose Zorzet | Email: bzorzet@sinc.unl.edu.ar

Bruno J. Zorzet , Diego H. Milone , Victoria Peterson , Rodrigo Echeveste

1° Research Institute for Signals, Systems and Computational Intelligence, sinc(i), FICH-UNL, CONICET
2° Institute of Applied Mathematics of the Litoral, IMAL, FIQ-UNL, CONICET

Brain-Computer Interfaces (BCIs) can transmit information between individuals and computers by monitoring their electrical brain activity using electroencephalogram (EEG) in real-time. The interpretation of these signals using artificial intelligence (AI) algorithms enables the categorization of mental states. However, addressing potential biases that may result from such algorithms, which can favor certain population groups over others, is an important case of study in algorithmic fairness that has not received much attention in the field of BCI.
In this study, we present experiments that help to understand the potential presence of bias with respect to sex on EEG signal decoding for motor imagery detection in BCI. We evaluated multiple databases and AI models. Our findings suggest the possibility that these disparities in performance are linked to the detection of sex-related information in EEG signals by AI models. This discovery exposes the urgency to mitigate biases in AI-based BCIs before deployment, to ensure equity in performance as well as to prevent the amplification of inequalities.