Abstract
This article presents a comprehensive review of technologies, tools, and applications in the field of brain-computer interfaces (BCI) and artificial intelligence (AI) for developing cognitively adaptive learning environments. Unlike prior reviews that treat BCI hardware, AI decoding models, and pedagogical deployment as separate topics, this review develops an integrated cross-layer taxonomy linking hardware constraints, AI architecture requirements, and classroom deployment conditions. Emphasis is placed on the transition from cumbersome laboratory systems to wearable hybrid solutions combining electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS), as well as textile sensors that offer a favorable trade-off between neurophysiological signal quality and mobility in real-world classroom settings. The review includes a detailed analysis of modern AI architectures, transformer-based models for processing non-stationary neural signals, and Siamese neural networks implementing a few-shot learning paradigm to address the calibration problem associated with new users. It further proposes a unified state-estimation framework integrating cognitive load, attention, and affect into a single formalization for closed-loop adaptive learning systems. The review concludes with a critical analysis of fundamental barriers to scalability, signal instability, mental state privacy issues, and the risk of a neurocognitive divide arising from unequal access to technology. The authors argue that the future of neuroeducation lies in the symbiotic interaction of humans and AI, supported by unified international standards and the convergence of BCI with large language models (LLM) for predictive communication.

This publication has 65 references indexed in Scilit: