Kantaro Fujiwara

Creativity fluctuates markedly from moment to moment, even within the same individual, yet the neural dynamics that determines whether a given attempt produces a highly creative idea remains poorly understood. Prior studies have identified static EEG correlates of creative thinking, but these do not explain how brain activity is dynamically organized before and during successful idea generation. Here, we model creative cognition as trajectories through a neural state space using energy landscape analysis (ELA) of EEG recorded during two complementary problem-solving paradigms: the divergent Alternative Uses Test (AUT) and the convergent, goal-directed Fusion Innovation Test (FIT). Across both tasks, creative success was associated with dissociable dynamical signatures in the resting and ideation stages. Before ideation, greater diversity of resting-state patterns of activity, corresponding to possible attractors, indexed a preparatory substrate of creative potential, showing weak trial-level effects but robust subject-level coupling with performance. During ideation, higher creativity was predicted by how the brain traversed its accessible state space: successful trials were characterized by traversal biased toward sustained exploration within stable attractor basins rather than frequent switching between basins of attraction ( β = 0.104, p = 0.004 , trial-level). These findings identify a task-invariant, biologically grounded dynamical mechanism of creative cognition and show that creative performance depends not only on which neural states are available, but also on how neural activity traverses that state space.

Wearable brain-computer interfaces (BCIs) have made it feasible to monitor brain activity for emotion recognition in real-world settings. While deep learning models achieve high classification accuracy on EEG data, they often lack interpretability, limiting their neuroscientific relevance. In this study, we present an interpretable framework for EEG-based emotion analysis rooted in energy landscape analysis. EEG signals from the DEAP dataset were standardized and binarized prior to quantification of neural state transitions. We found significant subject-specific correlations between the number of state transitions and emotional ratings of valence and arousal. Further analysis revealed that certain binary brain states, particularly complementary pairs, were among the most frequently observed and showed emotion-dependent frequency differences. Transitions between these state pairs varied across subjects, suggesting their role as local minima in the brain’s dynamic landscape. Our findings demonstrate that energy landscape analysis provides an interpretable alternative to black-box models, offering insights into how brain dynamics relate to emotional experiences. This approach contributes toward building explainable affective computing systems and supports the use of neural state modeling in emotion-aware BCIs.

Advancements in Brain-Computer Interface (BCI) technology, such as portable and inexpensive electroencephalogram (EEG) headsets, have profound implications for using this technology in monitoring mental health. This study delves into the development of a multi-reservoir computing network for EEG-based emotion recognition, addressing variations across trials, subjects, and contexts. We utilise several widely recognised benchmark datasets in the field, including DEAP[18], DREAMER[17], and GAMEEMO[2], selected for their diverse emotion stimuli (active/passive), EEG acquisition methods (wired/wireless), and trial sizes per subject (few/many). Our approach involves utilising reservoir computing, a variant of recurrent neural networks designed to mimic human brain functioning, in a multi-reservoir network architecture where each reservoir represents the activity of a distinct brain area. We propose a framework where signals from different EEG channels are stacked and fed into different reservoirs, yielding high-dimensional representations. These representations are used to train a support vector machine (SVM) classifier for predicting emotions in the Valence and Arousal domains. To optimise the stacking of EEG channels, we employ a Genetic Algorithm search strategy and conduct a comparative analysis to identify key EEG channels essential for efficient and computationally inexpensive predictions. We observe average classification performance across trials: 0.87 (Valence) and 0.86 (Arousal), across subjects: 0.84 (Valence) and 0.83 (Arousal) and across contexts: 0.74 (Valence) and 0.7 (Arousal). Subtle differences in classification performance indicate the importance of tuning generalisation vs personalisation when developing an application. Building upon these findings, future extensions of this work aim to facilitate the development of real-time prediction models that require monitoring of fewer EEG channels, thereby advancing the field of emotion recognition.

Our research focuses on the smart pole interaction unit (SPIU) as an infrastructure external human-machine interface (HMI) to enhance pedestrian interaction with autonomous vehicles (AVs) in shared spaces. We extensively study SPIU with external human-machine interfaces (eHMI) on AVs as an integrated solution. To discuss interaction barriers and enhance pedestrian safety, we engaged 25 participants aged 18-40 to brainstorm design solutions for Pedestrian-AV interactions, emphasising effectiveness, simplicity, visibility, and clarity. Findings indicate a preference for real-time SPIU interaction over eHMI on AVs in multiple AV scenarios. However, the combined use of SPIU and eHMI on AVs is crucial for building trust in decision-making. Consequently, we propose innovative design solutions for both SPIU and eHMI on AVs, discussing their pros and cons. This study lays the groundwork for future autonomous mobility solutions by developing human-centric eHMI and SPIU prototypes as ieHMI.