Advances in Explainable Analysis Methods for Cognitive and Computational Neuroscience

In the fields of cognitive and computational neuroscience, the past decade has witnessed a significant surge in the application of Artificial Intelligence (AI) techniques to decipher neural activity recorded with neuroimaging and neurophysiology. This has resulted in remarkable performance enhancements in neuroscience tasks, encompassing Brain-Computer Interfaces (BCIs) (e.g., motor imagery BCI and affective BCI), understanding complex cognitive processes (e.g., music and speech perception and visual cognition), and clinical applications (e.g., seizure detection and sleep staging). Nonetheless, the sophisticated nature and lack of transparency within these computational models pose challenges, as they often obscure the neural processes they aim to explore. This lack of transparency undermines their potential for clinical translation and hampers the progression of theoretical neuroscience.

This Research Topic aims to illuminate and enhance the interpretability of analysis methods in cognitive and computational neuroscience. It proposes to address critical questions about the underlying neural activities deciphered by AI models. The objective is to examine groundbreaking methodologies that promise to improve model transparency and reliability, thus enhancing confidence in the application of AI-driven diagnostics and neurotechnologies. This exploration aims to delve into the nuances of individual neural variability and to reveal latent processes that traditional analysis might overlook.

To gather further insights within the boundaries of explainable analysis methods in cognitive and computational neuroscience, we welcome articles addressing, but not limited to, the following themes:

- Implementation of advanced eXplainable AI (XAI) frameworks in neuroscience tasks, including ante-hoc and post-hoc XAI methods.

- Development of parameter sensitivity and feature importance analyses in neuroscience.

- Advancement in statistical attribution methods and their applications in neuroscience.

- Exploration of causal inference modeling and its integration with neural systems.

- Utilization of innovative neurobiologically plausible models.

In addition, studies incorporating EEG, MRI, eye-tracking, or multimodal neuroimaging paradigms, with applications to neuroscience tasks, are particularly encouraged. We invite submissions showcasing these techniques and their potential to deepen the understanding of neural mechanisms and achieve more trustworthy technology applications through more interpretable methods.