ExplainabilityTransparencyBiascs.LG
Toward the application of XAI methods in EEG-based systems
Andrea Apicella, Francesco Isgrò, Andrea Pollastro, Roberto Prevete
October 12, 2022
XAI.it@AI*IA
23 citations
Key Takeaways
-
XAI methods can help identify which brain signal features remain consistent across different recording sessions, potentially improving how brain-computer interfaces work reliably over time, even though EEG signals naturally change between sessions.
-
Explainable AI revealed that while some important signal characteristics are shared across sessions, others are highly dependent on the specific input, suggesting that a one-size-fits-all approach to brain signal interpretation may not be optimal for responsible AI deployment in medical or assistive technology contexts.
-
The research demonstrates a practical application of XAI beyond just transparency: using explanations from AI methods to actually improve system performance and generalization, rather than only using them for post-hoc understanding of decisions already made.
-
The Dataset Shift Problem—where AI systems trained on one set of conditions perform poorly on new conditions—is a significant challenge in brain-computer interfaces that responsible developers must address to ensure these systems work reliably for individual users across multiple sessions.
Abstract
An interesting case of the well-known Dataset Shift Problem is the classification of Electroencephalogram (EEG) signals in the context of Brain-Computer Interface (BCI). The non-stationarity of EEG signals can lead to poor generalisation performance in BCI classification systems used in different sessions, also from the same subject. In this paper, we start from the hypothesis that the Dataset Shift problem can be alleviated by exploiting suitable eXplainable Artificial Intelligence (XAI) methods to locate and transform the relevant characteristics of the input for the goal of classification. In particular, we focus on an experimental analysis of explanations produced by several XAI methods on an ML system trained on a typical EEG dataset for emotion recognition. Results show that many relevant components found by XAI methods are shared across the sessions and can be used to build a system able to generalise better. However, relevant components of the input signal also appear to be highly dependent on the input itself.