Adaptive Generative AI Interfaces via EEG-based Cognitive State Recognition

This study proposes an EEG-driven adaptive optimization method for generative AI interaction interfaces, aiming to enhance usability and personalization. Unlike conventional static UIs, the approach integrates real-time brain-computer feedback to dynamically adjust interface parameters, including font size, response latency, and suggestion density. EEG signals are continuously monitored to detect fluctuations in user attention and cognitive load, serving as inputs for adaptive control. A lightweight support vector machine (SVM) classifier identifies cognitive states, which are then mapped to corresponding UI adjustments via rule-based logic. The system architecture ensures low-latency responsiveness and minimal computational overhead. In controlled experiments involving 20 participants across diverse cognitive tasks, the adaptive interface reduced average task completion time by 12.4%, decreased NASA-TLX workload scores by 18%, and significantly improved subjective readability ratings. These findings demonstrate the method’s effectiveness in real-time user experience enhancement and cognitive state alignment. The proposed framework holds promise for deployment in education, healthcare, and other domains requiring sustained, adaptive human–AI interaction.