Subject-Specific Analysis of Self-Initiated Attention Shifts from EEG with Controlled Internal and External Attention Conditions
Published in arXiv preprint, 2026
We investigate self-initiated attention shifts through subject-level EEG decoding using a machine learning–based approach with two complementary analyses: frequency-specific topographic patterns and SHapley Additive exPlanations (SHAP) feature attribution. Results show reliable within-subject classification performance, with higher-frequency bands and frontal regions contributing most to model decisions. We note caution regarding potential non-neural artifacts in high-frequency signals.
Recommended citation: Zeng, Y., Hou, D., Zhang, Z., Sun, S., Huang, Y., Tseng, C., & Shioiri, S. (2026). Subject-Specific Analysis of Self-Initiated Attention Shifts from EEG with Controlled Internal and External Attention Conditions. arXiv preprint arXiv:2605.18251.
Download Paper