An EEG correlation framework to study state anxiety and learning under uncertainty

Objective.Recent developments in computational neuroscience have shed light on the neural processes underlying altered decision-making under uncertainty in anxiety. These disruptions are partly attributed to impaired encoding of precision-weighted prediction errors (pwPEs), which guide belief updating during learning and decision-making, as described by hierarchical Bayesian models. In this paper, we introduce a gamified paradigm for collecting decision-making data, together with a framework for extracting EEG features linked to computationally relevant variables, drawing on principles from neurofeedback and brain-computer interface research. This approach aims to develop tools that target functionally meaningful brain networks involved in decision-making, with the potetntial to inform future neurofeedback interactions.Approach.Forty healthy participants performed a volatile decision-making task in a game-based, immersive environment. EEG data were analysed to identify spatial filters whose theta- and alpha-band power correlated with pwPEs and state anxiety scores. Both intra-subject (trial-wise pwPEs) and inter-subject (state anxiety) analyses were conducted to uncover distinct neural signatures.Main results.The intra-subject analysis revealed that pwPEs were significantly and positively correlated with theta power, and significantly and negatively correlated with alpha power-supporting the hypothesis that these oscillatory patterns underlie belief updating. In contrast, the inter-subject analysis showed that higher state anxiety was associated with reduced theta and increased alpha power, consistent with attenuated learning and impaired adaptation in anxious individuals. These findings align with theoretical models of hierarchical Bayesian inference and prior evidence of anxiety-related disruptions in uncertainty processing.Significance.The findings validate the proposed EEG framework for identifying neural markers related to belief updating and anxiety-related learning impairments. This approach lays the foundation for personalized neurofeedback procedures that target maladaptive decision-making in anxiety, with the added benefit of using immersive task paradigms for better engagement and translational potential for real-world applications.