Enhancing Accessibility in Education Through Brain–Computer Interfaces: A Scoping Review on Inclusive Learning Approaches

Abstract

Brain–computer interfaces (BCIs) hold promise in enhancing accessibility in education by enabling students with physical disabilities to interact with digital learning environments without barriers. However, no comprehensive review has explored the landscape and role of BCIs in inclusive learning. Hence, this review sets out to identify relevant literature on BCI-based educational technologies, highlight their key themes, characteristics, and research methodologies, and identify research gaps. The secondary aim is to evaluate how these educational technologies contribute to inclusive learning frameworks by fostering communication, collaboration, engagement, and accessibility among students with disabilities. Overall, the reviewed studies demonstrate that BCIs can facilitate assistive communication among non-verbal students and provide motor control support for physically impaired persons. While these interventions show strong potential, challenges remain, including high implementation costs, user adaptability, and ethical concerns related to neural data privacy. Specifically, there is a need to (1) shift from experimental applications towards real-world classroom integration by developing user-friendly, cost-effective, and ethically sound BCI-based educational technologies, and (2) extend ongoing research efforts to include underserved populations to assess the generalizability of current and future BCI-based interventions. More importantly, future work should focus on enhancing BCI usability, improving adaptability for diverse learners, and establishing ethical guidelines for the development of socially responsible and inclusive neuro-educational technologies for all people with disabilities everywhere. This will go a long way in fostering the fourth and tenth United Nations Sustainable Development Goals of Quality Education and Reduced Inequalities, respectively.

1. Introduction

Ensuring accessibility across all levels of education is fundamental to promoting equal learning opportunities for all students everywhere, particularly those with physical, cognitive, and communication disabilities. Traditional assistive technologies, such as speech-to-text software and adaptive keyboards, have helped bridge the accessibility gap. However, these solutions often require manual interaction, which can be a significant barrier for individuals with severe motor impairments, thereby hindering the laudable objective of having inclusive educational spaces for all students regardless of race, age, gender, or disability. Interestingly, BCIs have emerged as a transformative approach to overcoming this physical limitation by allowing learners with motor-related disabilities to interact with digital learning environments solely through neural mechanisms. BCI functions by detecting and interpreting brain signals via non-invasive techniques such as electroencephalography (EEG) [1,2,3]. It enables users to perform tasks such as text input and cursor control without requiring physical movement and/or contact.

The application of BCIs in education has the potential to revolutionize learning spaces by providing alternative learning and communication methods, facilitating independent digital interaction, and enhancing classroom participation and collaboration. For example, non-verbal students can use BCI-driven text-to-speech systems to express themselves, enabling effective communication that was previously challenging [4]. Moreover, BCIs empower physically disabled students to control computers, access e-learning platforms, and navigate assistive technologies using brain signals alone, reducing reliance on others for routine interactions [5,6]. Similarly, individuals with mobility impairments can engage in learning activities, answer questions, and interact with peers through BCI-controlled interfaces [7], ensuring equal access to educational content without physical or technological barriers.

Despite its potential, the integration of BCI in education remains an emerging field, with challenges such as high costs [8], technical expertise required for setup and maintenance [9], lack of testing with diverse populations [10], and ethical concerns surrounding neurodata privacy and consent [11] limiting accessibility. To explore the landscape of BCI-based educational technologies, this review aims to map existing research and publications on the topic by identifying their key characteristics, challenges, and opportunities for future directions to facilitate the seamless integration of this inclusive technology into learning environments, foster accessibility, and reduce digital inequalities in education. Specifically, this scoping review aims to address the following research questions (RQs):

• RQ1: What types of BCIs have been used in educational settings to promote accessibility for students with disabilities?
• RQ2: What benefits and challenges are associated with implementing BCIs in inclusive learning environments?
• RQ3: What research gaps exist in the current literature on the use of BCIs for inclusive education?

2. Related Work

Literature reviews on the potential of BCIs in several domains exist, with a number of them assessing BCI technology from distinct disciplinary angles, including hardware design, learning outcomes, and ethical challenges. In this scoping review, we cover those reviews focused on education, accessibility, and rehabilitation.

Wegemer [12] critically examined the socio-technical implications of integrating BCI applications into classrooms, highlighting risks of deepening educational inequality due to disparities in infrastructure and access to effective BCI technology. Complementing this, Kögel et al. synthesized ethical and user-centered perspectives on assistive BCI deployment, especially for users with severe motor impairments [13]. Moreover, Xia et al. [14] conducted a systematic review focused on learning regulation via BCI in mainstream environments. However, the review was not focused on students with disabilities. Further, Pan and Cristea [15] advanced the discourse by reviewing personalized and adaptive learning frameworks in BCI-enhanced systems. However, the research did not consider inclusive and/or special education contexts. Jamil et al. [16] and Hosseini et al. [10] presented a machine learning-focused overview of EEG signal decoding, which contributed to the technical understanding of BCI-driven learning and rehabilitation. Meanwhile, Othman [9] presented a rapid review on BCI applications for students with disabilities, directly addressing learning improvements but lacking granular analysis of implementation contexts. Orovas et al. [8] offered a comprehensive scoping review of EEG usage in educational settings, addressing methodological and hardware-specific considerations.

While prior reviews have either focused on general education [14,15], clinical rehabilitation [10,16], or ethical issues [13], this review is unique in its specific emphasis on using BCI to improve formal educational access for students with motor disabilities. In contrast to prior studies that exclude these populations or merge therapeutic and learning outcomes, we highlight studies where BCIs enable inclusive digital participation, assistive communication, and adaptive learning within educational institutions. Additionally, we use existing frameworks to classify the reviewed BCI applications. For example, we use Zander and Kothe’s binary framework [17] (passive monitoring and active control) to classify BCI applications. Moreover, we use Koehler et al.’s TPACK framework [18] to categorize educational technologies and classify the reviewed applications into three constituent types: technological, pedagogical, and content. The categorical distinctions provided by both frameworks are rarely emphasized in earlier reviews.

3. Methods

3.1. Protocol and Search Strategy

This scoping review was conducted in accordance with a pre-defined protocol registered on the International Platform of Registered Systematic Review and Meta-analysis Protocols (INPLASY) (Registration number: INPLASY202570019; DOI: 10.37766/inplasy2025.7.0019). The protocol specified the databases to be searched, the inclusion and exclusion criteria, and the overall screening process, ensuring that the review was conducted in a transparent and reproducible manner.

To systematically investigate the role of BCIs in enhancing accessibility in an educational context, a comprehensive search strategy was designed and executed. Our objective was to identify and synthesize relevant studies focusing on the intersection of BCIs, accessibility, and education.

3.1.1. Database Selection and Search Strategy

The search was conducted using four major academic databases: ACM Digital Library (Association for Computing Machinery, New York, NY, USA), Scopus (Elsevier, Amsterdam, The Netherlands), IEEE Xplore (Institute of Electrical and Electronics Engineers, Piscataway, NJ, USA), and Education Index (ProQuest, Ann Arbor, MI, USA). Together, these databases extensively cover computer science, engineering, and educational technology literature, ensuring that the most relevant and high-quality studies are captured.

The search was restricted to peer-reviewed journals and conference papers written in English, given the researchers’ language proficiency and the prevalence of English in scientific communication. No date restriction was applied, allowing for a historical perspective on the evolution of BCIs in education.

To capture all relevant studies, we developed a structured search strategy using three sets of terms reflecting the key aspects of the research question. The first set focused on accessibility (e.g., “accessibilit*” and “disabilit*”), the second on BCIs (e.g., “Brain–Computer Interface”, “BCI”, “Brain–Machine Interface”, and “BMI”), and the third on education (e.g., “education” and “learning”). Boolean operators ensured precision and coverage: “AND” was used across sets to retrieve studies at the intersection of accessibility, BCIs, and education, while “OR” was applied within sets to capture variations in terminology. Truncation symbols (wild cards) were included to account for different word endings and spelling variants.

3.1.2. Eligibility Criteria

To ensure the selection of the most relevant and high-quality studies, we applied both inclusion and exclusion criteria. Studies were eligible if they were empirical investigations published in peer-reviewed journals or conference proceedings, written in English, and focused on the application of BCIs to promote accessibility in educational contexts, reporting on educational outcomes, learning experiences, or inclusive spaces using BCI-based technologies.

Studies were excluded if they were non-empirical (e.g., editorials, book chapters, opinion pieces, or non-peer-reviewed sources), if they did not address accessibility or disability-related challenges in education, or if they focused solely on clinical, medical, or entertainment applications of BCIs without educational relevance. No publication date restrictions were applied, as BCI remains a relatively new technology, and none of the retrieved papers were published prior to 1985.

3.2. Data Collection

To ensure a rigorous and comprehensive review, a multi-step data collection process was implemented. This process aimed to identify the most relevant studies, remove duplicate records, and ensure that selected articles aligned with the research focus on BCI applications in education. Our initial search was across the four databases, ACM Digital Library, Scopus, IEEE Xplore, and Education Index, chosen for their extensive coverage of research in BCIs, accessibility, and educational technologies. Duplicate records were identified and removed. The deduplication process involved matching titles, authors, publication years, and digital object identifiers (DOIs). This step ensured that all candidate articles, eligible for title/abstract screening, were uniquely represented in the pool before moving on to the screening phase proper.

3.3. Study Screening

The screening process was conducted in two stages: title/abstract screening (TAS) followed by full-paper screening (FPS). To filter out irrelevant articles efficiently, two reviewers, Mohammed Abdulmawjood (Reviewer 1) and Emmanuel Nzeakor (Reviewer 2), independently applied a predefined protocol outlining inclusion criteria such as a clear focus on BCI applications, educational context, accessibility promotion, and educational outcomes. Studies that focused purely on clinical or entertainment applications of BCIs, without educational relevance, were excluded. Articles at this stage were classified as “Accepted” or “Rejected.” In cases of disagreement, both reviewers engaged in discussions, referring to the inclusion criteria to reach a consensus. Inter-rater reliability for TAS was assessed using Cohen’s Kappa.

Articles that passed the TAS were subjected to FPS. At this stage, the same two reviewers independently assessed each article in terms of methodological rigor, relevance to BCI-based educational technologies, and alignment with the predefined inclusion and exclusion criteria. Each study was evaluated based on design, target population, technology used, and educational outcomes. Discrepancies were resolved through discussions between the two reviewers and, when necessary, consultation with a senior researcher supervising the review. Inter-rater reliability for FPS was also assessed using Cohen’s Kappa.

3.4. Snowball Method for Additional Sources

In addition to the systematic search, we employed the snowball method to identify additional relevant literature. This involved reviewing the reference lists of the selected studies and the reviews from the related works section to find related research that met our inclusion criteria. References that frequently appeared across multiple studies were prioritized for further examination. These papers went through the same process of passing our inclusion criteria and were then put through a full-paper screening.

3.5. Data Extraction and Synthesis

Data extraction was performed on the final set of articles using a standardized data extraction form. Variables collected included study characteristics such as title, authors, year, country, and age group. BCI modalities, including EEG, virtual environments, neural network models, and others, were noted. Educational applications, such as the type of educational activity and platform, were also recorded. Outcomes measured included learning evaluation, effectiveness, and other performance metrics. A thematic analysis approach was applied to synthesize the extracted data, identifying recurring themes, challenges, and research gaps in the literature.

3.6. Quality Assessment

To ensure a systematic and transparent evaluation of methodological rigor, we applied the Mixed Methods Appraisal Tool (MMAT), version 2018 (Canadian Intellectual Property Office, Ottawa, ON, Canada) [8]. The MMAT is specifically designed to appraise the quality of empirical research across five methodological categories: qualitative research, randomized controlled trials, non-randomized studies, quantitative descriptive studies, and mixed methods research. Its flexibility made it well-suited for our review, which included studies employing diverse research designs.

Methodological rigor was evaluated by scoring each study against five core MMAT criteria tailored to its methodological type. For instance, qualitative studies were assessed based on (1) the appropriateness of the qualitative approach to the research question, (2) the adequacy of data collection methods, (3) the depth and transparency in data analysis, (4) the coherence between data sources, analysis, and interpretation, and (5) the substantiation of findings by the data presented. For quantitative descriptive studies, rigor was assessed by examining the relevance of the sampling strategy, representativeness of the sample, validity and reliability of measurement tools, risk of nonresponse bias, and the appropriateness of statistical analysis. Mixed-methods studies were evaluated using criteria that measured not only the rigor of the individual qualitative and quantitative components, but also (1) the justification for using a mixed-methods design, (2) the integration of findings, and (3) the consistency in the interpretation of results across methods.

Each criterion was scored as “Yes”, “No”, or “Can’t tell”, following MMAT’s standard protocol. Two independent reviewers applied the MMAT to all 19 eligible studies. To calibrate their use of the tool and ensure consistency, the reviewers first conducted a joint pilot rating on a subset of studies. Any discrepancies in scoring were resolved through discussion and consensus; where disagreement remained, a third reviewer was consulted to arbitrate the final judgment.

The assessment revealed that most included studies demonstrated moderate to high methodological quality. A majority met at least four of the five criteria applicable to their study type. Common strengths included alignment between study aims and design, appropriate use of EEG-based BCI technology, and valid outcome measures. Nevertheless, several studies did not clearly report sampling strategies, data collection fidelity, or procedures for integrating mixed-methods data, which limited confidence in certain findings.

By incorporating the MMAT as a structured evaluation tool, we ensured that methodological quality was systematically and transparently considered during data synthesis and interpretation. This process strengthened the overall credibility and validity of our review findings on the role of BCIs in supporting accessible education for individuals with motor disabilities.

3.7. Risk of Bias

To ensure a rigorous and objective review, we implemented predefined inclusion and exclusion criteria. However, potential biases may still exist in our study selection process. One key limitation is selection bias, as our search was restricted to peer-reviewed journals and conference papers, excluding grey literature and non-indexed sources. This may have led to an over-representation of studies demonstrating positive results on BCI applications in education. Additionally, publication bias is a concern, as studies reporting significant improvements in accessibility and learning outcomes are more likely to be published than those with negative or inconclusive findings. To mitigate this, we included studies across various educational and technological contexts. Another important consideration is database bias. Our search strategy relied primarily on databases such as IEEE Xplore, ACM, Education Index, and Scopus, which may favor technical and biomedical literature while under-representing educational or social science research on BCI. This may have influenced the thematic scope of our included studies, skewing findings toward technology-centric outcomes. Although using Education Index was a conscious choice to include studies with a focus on educational pedagogy. Reporting bias was also evident, as some studies provided limited details on methodology and outcome assessment, making cross-study comparisons challenging. To minimize this issue, only studies with clear intervention descriptions and measurable outcomes were included. Furthermore, we addressed researcher bias by conducting independent screenings and resolving disagreements through consensus. Despite these measures, limitations in the representativeness and generalizability of our findings should be acknowledged.

4. Results

4.1. Overview of Selected Studies

The result of the article selection process, which was based on the PRISMA guidelines, is shown in Figure 1. Upon completion of the four steps in the PRISMA protocol, we arrived at 19 included articles for the final analysis.

Table 1. Inter-rater reliability for title/abstract screening (TAS) and full-paper screening (FPS).

Metric	TAS	FPS
Both Reviewers Agree	150	11
Both Reviewers Disagree	1779	134
Reviewer 1 Agree	66	2
Reviewer 2 Agree	49	2
Total	2044	149
% Agreement	94.37%	97.32%
Cohen’s Kappa	0.692	0.831

shows the result of the inter-rater reliability. For the TAS, the Cohen’s kappa ($\kappa$) was $0.692$ (substantial agreement). Moreover, for the FPS, the Cohen’s kappa was $0.831$ (almost perfect agreement).

Figure 2 shows the distribution of publication years for the included articles. The publication years range from 2006 to 2024, spanning more than one and a half decades, with a marked and steady increase in publication frequency occurring in 2019. Figure 3 shows the distribution of the articles in terms of the countries in which the first authors’ universities or organizations are located. The United States of America turned out to be the most frequent country, followed by India, China, and Japan.

4.2. Study Characteristics

The 19 included studies addressed a range of participant groups and application domains. Nine studies involved individuals with severe motor impairments, such as tetraplegia or mobility-limiting conditions, while ten studies recruited healthy participants for BCI training or experimental control tasks. Clinical-oriented research focused on BCIs for neurorehabilitation, assistive mobility, and functional restoration, including applications in tetraplegic sports competition, smart wheelchair navigation, and integration with functional electrical stimulation (FES) [2,5,6,19].

Non-clinical works explored prototype development, advanced signal processing, and machine learning-based classification to improve BCI performance. This included approaches such as adaptive decomposition [20], ensemble classification [21], transfer learning [22], spectrogram-based motor imagery classification [23], and deep learning frameworks for ERP-based BCIs [24,25,26].

Several studies were set in educational or training contexts, ranging from virtual drone operation [4,27] to assistive wheelchair training [6,19], illustrating the adaptability of BCIs to inclusive learning and accessibility objectives. Cognitive and learning-focused studies implemented passive BCIs to adapt instructional pace based on cognitive load [7], neurofeedback for cognitive control [28], and psychosocial interventions to improve user engagement [29].

Applications across the included studies span from cost–benefit analyses of assistive mobility devices [19] to high-accuracy P300-based speller systems for communication in disabled participants [30]. This diversity underscores the interdisciplinary nature of BCI research, bridging neuroscience, rehabilitation science, engineering, and education. A detailed overview of the included studies is presented in

Table 2. Summary of included studies. MI: motor imagery; PCA: principal component analysis.

Authors	Participants	Aim of Study	Summary of Findings
Hehenberger et al. [5]	Tetraplegic BCI race pilot, 14-month longitudinal study	Improve race performance through transfer learning and adaptation	Runtimes improved (255 → 225 s) and accuracy rose from 46% to 53%.
Kamble et al. [20]	Healthy adults	Develop MI and silent communication EEG-based BCI	Achieved 89.6% binary and 61.1% 7-class accuracy.
Dumitrescu et al. [27]	Healthy participants	Control a virtual drone via MI-BCI	Achieved 95.5% accuracy; successful virtual drone control.
Echtioui et al. [21]	BCI Competition III dataset IVb	Improve MI classification with ensemble learning	RBF-SVM + Linear SVM + Decision Tree yielded k = 0.783, outperforming individual classifiers.
Sharma et al. [6]	Healthy adults	Apply transfer learning (AlexNet) to MI classification	Transfer-learned AlexNet achieved highest MI classification accuracy.
Badajena et al. [19]	100 subjects	Enhance smart wheelchair decision-making with EEG	Feature weighting + AMCBA improved performance.
Ahmadi et al. [31]	EEG dataset (eye states)	Classify eye states for real-time BCI	PCA+CFS with CART achieved 97.9% accuracy.
Choi et al. [2]	Stroke patients	Use MI-BCI to control FES for rehabilitation	Significant improvement in upper limb motor function; high accuracy and user satisfaction.
Brumberg, Pitt [4]	Healthy adults	Investigate N100 suppression in speech-BCI use	N100 suppression suggests speech motor planning aids BCI use.
White et al. [29]	College students with ASD	Test feasibility of VR–BCI for social/behavioral training	Feasible; mixed behavioral outcomes warrant further study.
Hoffmann et al. [30]	5 disabled users	Evaluate performance of ERP-based BCI	Four reached 100% accuracy; bitrates 10–25 bits/min.
Orovas et al. [8]	Chronic BCI users	Maintain performance over long term with deep NN decoder	Sustained >90% accuracy over a year without retraining.
Li et al. [32]	Healthy adults	Develop MR-BCI for robotic grasping tasks	93.0% accuracy; all subjects completed tasks without collisions.
Zhang et al. [23]	EEG dataset, 7-class MI	Apply pre-trained deep models to MI classification	Achieved 84.9% average accuracy.
Fortes et al. [33]	Conceptual	Propose dynamic architecture for adaptive BMIs	Framework supports predictive closed-loop BMI without direct movement data.
Khan et al. [22]	Healthy adults	Transfer learning for imagined speech BCI without calibration	Achieved ~65.7% accuracy for new imagined words.
Santamaria-Vazquez et al. [24]	ERP-based BCI dataset	Develop EEG-Inception deep learning model	Outperformed competing methods by up to 16%; fewer trials needed.
Keizer et al. [28]	Healthy adults	Improve cognitive control via neurofeedback	Gamma training improved binding flexibility; beta training enhanced familiarity.
Belwafi, Ghaffari [25]	Severe motor disabilities	Multi-application control via hybrid MI+P300 BCI	Achieved 82% (MI) and 95% (P300) accuracy.

.

4.3. Evaluation of BCI Applications

The included studies were further evaluated based on their primary application domain, learning or evaluation approach, and reported effectiveness. To better structure the review, applications were grouped into thematic categories that reflect how BCIs are used in practice, from competitive training and assistive device control to cognitive enhancement and high-performance deep learning systems. This thematic classification highlights commonalities between studies, facilitates comparison across domains, and identifies research areas where BCIs have shown the greatest promise or where further development is needed.

Table 3. Evaluation of BCI applications in included studies. ADC = assistive device control; TL = transfer learning; CA = classification accuracy; NF = neurofeedback; VR = virtual reality; DNN = deep neural network; BMI = brain–machine interface; CBA = cost–benefit analysis).

Application Type	Evaluation Approach	Outcome	Articles	Count (%)
Competitive and Performance Training	Multi-month training with tetraplegic pilot; user adaptation over time	Sustained improvement in control performance	[5]	1 (4%)
Signal Processing and Classification Advances	Adaptive decomposition, ensemble classifiers, TL, spectrogram-based classification	Enhanced CA across multiple EEG tasks	[20,21,22,23,24,31,34]	7 (28%)
Assistive Device Control (Physical System)	EEG-based control of wheelchairs, manipulators, and assistive devices; usability and CBA studies	Reliable operation; demonstrated feasibility for real-world use	[6,19,32,35,36,37]	6 (24%)
Virtual and Simulated Control	Motor imagery for VR drone and speech synthesizer control; ERP modulation analysis	Improved precision and reduced ERP amplitude in target conditions	[4,27]	2 (8%)
Neurorehabilitation Applications	FES integration; bibliometric mapping of EEG in rehab	Enhanced motor recovery via FES; literature trends mapping (bibliometric)	[2,38]	2 (8%)
Cognitive and Learning Enhancement	Passive BCI adapting learning speed to cognitive load; psychosocial interventions; NF for retrieval control	Positive impacts on engagement, feasibility, and cognitive control	[7,28,29]	3 (12%)
High-Performance and Deep Learning BCIs	DNN decoding frameworks, ERP-based CNNs, dynamic BMI architectures	Met or exceeded user performance expectations; improved responsiveness	[24,25,26,33]	4 (16%)
P300-Based Communication	P300 speller interface evaluation with disabled participants	Efficient communication with high accuracy	[30]	1 (4%)

summarizes these categories, outlining the nature of the application, the learning or evaluation methodology employed, key outcome measures, and the studies that contributed to each category.

4.4. BCI Active Control vs. BCI Passive Monitoring Applications

Table 4. Evidence synthesis across 19 included BCI studies by implementation mode, technology type, application domain, TPACK dimension, and reported effectiveness.

Criterion	Category	Articles	Count (%)
Implementation Mode	Active control	[2,4,5,6,8,19,20,21,22,23,24,25,27,30,32,33]	16 (84.2%)
	Passive monitoring	[28,29,31]	3 (15.8%)
Technology Type	Invasive	[8,33]	2 (10.5%)
	Non-invasive	[2,4,5,6,19,20,21,22,23,24,25,27,28,29,30,31,32]	17 (89.5%)
Application Domain	Assistive control	[2,4,5,6,19,21,23,24,25,27,30,32]	12 (63.2%)
	Adaptive learning	[20,22,28]	3 (15.8%)
	Rehabilitation	[2,5]	2 (10.5%)
	Miscellaneous	[29,31]	2 (10.5%)
TPACK Knowledge	Technological	[2,4,5,6,8,19,20,21,22,23,24,25,27,28,29,30,31,32,33]	19 (100%)
	Pedagogical	[29]	1 (5.3%)
	Content	[4,22]	2 (10.5%)
Effectiveness	Yes	[2,4,5,6,8,19,20,21,22,23,24,25,27,30,32,33]	16 (84.2%)
	No	[31]	1 (5.3%)
	Partial	[28,29]	2 (10.5%)

synthesizes the 19 included studies across four key dimensions: implementation mode, technology type, application domain, and technological pedagogical content knowledge (TPACK) dimensions [18], alongside reported effectiveness. These categories were selected to capture not only what BCIs do, but also how they are developed, deployed, and evaluated in educational and assistive contexts.

The first dimension, implementation mode, differentiates between active control systems and passive monitoring systems. This categorization provides a clearer understanding of how BCIs mediate learner agency, system interactivity, and accessibility in both classroom and remote learning environments. In passive monitoring systems, BCIs operate in the background to measure cognitive or emotional states, such as attention, workload, or engagement, without requiring the learner to intentionally issue commands. These insights can be used to adapt content delivery, pacing, or feedback in real time. In contrast, active control systems rely on deliberate user input, where learners intentionally generate specific brain signals (e.g., motor imagery and P300 responses) to manipulate applications, navigate interfaces, or control external devices. Moreover, active systems directly translate neural signals into commands for controlling external devices or navigating digital environments, and they make up the majority of the studies (16/19). This predominance reflects the strong emphasis on BCI research focused on direct assistive applications such as wheelchair navigation [6,19], drone control [27], and robotic manipulation [32]. Passive systems (3/19), in contrast, monitor brain activity without requiring intentional control, supporting contexts such as cognitive state tracking or adaptive learning [28]. This distinction underscores how BCIs can either act as explicit control interfaces or as implicit monitoring tools.

The second dimension is technology type, where the included studies are overwhelmingly non-invasive (17/19), using EEG or related modalities to avoid surgical risks. Only two invasive studies were identified [8,33], both emphasizing high-precision control in clinical contexts. This distribution highlights the current research priority for portable, low-risk systems suitable for broader deployment.

The third dimension, application domain, categorizes studies by their intended purpose. The largest group is assistive control (12/19), reflecting BCIs’ central role in supporting individuals with severe motor impairments. Smaller groups address adaptive learning (3/19), where neural monitoring informs educational scaffolding [20,22,28]; rehabilitation (2/19), which integrates BCIs with therapies such as functional electrical stimulation [2,5]; miscellaneous applications (2/19), including exploratory work outside these domains [29,31]. Together, this range shows how BCI research spans immediate assistive needs, long-term therapeutic goals, and emerging educational innovations.

The fourth dimension is framed using the TPACK framework [18], which emphasizes the interplay of technological, pedagogical, and content knowledge in the effective integration of new tools. All included studies necessarily engaged with technological knowledge, since they rely on signal acquisition, processing, and interface design. A small number explicitly engaged pedagogical knowledge (1/19) [29], for example, in adaptive instruction. Similarly, only two studies foregrounded content knowledge (2/19) [4,22], embedding BCI within domain-specific learning or task environments. This categorization matters because it reveals that while technological development dominates, the pedagogical and content dimensions remain underexplored.

Finally, studies were categorized based on reported effectiveness. A strong majority (16/19) demonstrated successful implementation, at least within controlled contexts. Two studies reported only partial effectiveness [28,29], typically due to variability across participants, while one reported no measurable effect [31]. The fact that 16 out of 19 papers demonstrate successful implementation underscores the promise of the field, while simultaneously revealing its limitations in robustness and generalizability.

4.5. Summary of Findings and Thematic Synthesis

A total of 19 studies met the inclusion criteria, as summarized in Table 2, with outcomes summarized in Table 3, spanning 2006–2024. Research output increased markedly after 2018, coinciding with the rise of affordable EEG hardware and advancements in machine learning and signal processing [20], addressing RQ1 by showing a diversification in application domains. These domains included motor rehabilitation, assistive mobility devices (e.g., wheelchairs and FES), educational and training simulators, virtual environment navigation, and communication support for non-verbal users. Nine studies involved participants with motor disabilities, while ten recruited able-bodied individuals participants.

Studies originated from a broad range of countries, with the United States contributing the largest share (Figure 3). Of the included studies, 13 reported participant gender, revealing a predominance of male participants, followed by female and other/unspecified categories. Age data were available for 15 studies, with participants ranging from young adults to older adults, illustrating the applicability of BCIs across varied populations.

In terms of RQ2, modalities and technologies, EEG was the primary recording method (18/19 studies), with motor imagery paradigms dominating active control applications. Event-related potentials and hybrid MI–ERP/SSVEP paradigms were also employed, particularly for communication-focused or performance-optimized BCIs. Computational approaches such as convolutional neural networks, ensemble classifiers, transfer learning, and adaptive signal decomposition were widely applied to enhance accuracy and real-time usability.

Finally, addressing RQ3, the studies reported several recurring challenges: high equipment costs, limited portability, need for technical expertise during setup and operation, variability in user performance, small sample sizes, and limited demographic diversity. Ethical concerns, particularly neurodata privacy and informed consent, were acknowledged but not extensively explored. These limitations highlight ongoing barriers to scaling BCI solutions for inclusive educational environments.

5. Discussion

5.1. Interpretation of Findings

This review synthesized evidence from 19 studies investigating the role of BCIs in inclusive education. The studies span a broad geographic spread across North America, Europe, and Asia (Figure 3), underscoring the international interest in using BCIs to promote accessibility. Research activity has accelerated since 2018 (Figure 2), reflecting the combined effects of lower-cost EEG hardware, advances in machine learning, and increased attention to inclusive learning technologies [20,21,24,26,34].

Overall, the included studies illustrate that BCIs are no longer confined to clinical or experimental contexts but are being adapted for diverse educational and assistive applications. Both active control and passive monitoring approaches were represented, which can be seen in Table 4, highlighting the dual potential of BCIs to enable direct interaction as well as to support adaptive, learner-centered environments.

At the same time, challenges in cost, scalability, and methodological consistency remain evident. Reporting of technical specifications and participant demographics was uneven, and economic feasibility is still insufficiently studied. These issues point to the importance of aligning technical development with usability, affordability, and equity considerations.

From a practical perspective, these findings point to the importance of cost–benefit analyses and user-centered design in future BCI development. For educational applications, the following recommendations are made:

• Prioritize low-cost, modular EEG systems that can be scaled in channel count and upgraded over time [19,35].
• Include diverse participant groups, particularly individuals with different disabilities, age ranges, and cultural backgrounds, to improve generalizability [6,8,9].
• Integrate adaptive algorithms capable of adjusting to user-specific neural signatures for sustained performance gains [20,21,24].
• Report hardware and software costs alongside usability and performance metrics to enable cross-study comparisons and support policy decision-making.

By aligning technical innovation with accessibility and inclusivity, BCIs can move from research prototypes to sustainable educational tools, especially in resource-constrained environments.

5.1.1. Discussion of RQ1: Types of BCIs in Education

RQ1: What types of BCIs have been applied in educational contexts to promote accessibility?

The review shows that EEG overwhelmingly dominates as the recording modality (18/19 studies; Table 4), with motor imagery (MI), the most common paradigm for active control and event-related potentials (ERP), widely used in communication-focused applications. Hybrid MI–ERP systems were also tested, reflecting efforts to improve performance and usability.

Applications clustered mainly around assistive device control (12/19 studies, as shown in Table 3), including wheelchairs, robotic manipulators, and functional electrical stimulation. A smaller but important group addressed adaptive learning and cognitive support through passive BCIs that monitor cognitive load or deliver neurofeedback. This split between active and passive BCIs highlights two complementary directions: enabling direct control for accessibility and providing implicit support for inclusive learning.

5.1.2. Discussion of RQ2: Benefits and Challenges

RQ2: What benefits and challenges do BCIs present for enhancing accessibility in education?

Our review reveals that BCIs offer several benefits in education. The benefits include enabling participation for individuals with severe motor impairments through applications such as smart wheelchairs, P300 spellers, and training simulators (Table 2). Passive BCIs demonstrated value for adaptive instruction and engagement by monitoring workload or providing neurofeedback. Together, these outcomes illustrate BCIs’ potential to enhance inclusivity in educational contexts. Challenges, however, remain. Cost is a major barrier: consumer-grade EEG systems (USD 300–1200) show promise, but research-grade systems often exceed USD 10,000 [19,35,39], limiting adoption in schools. Studies also relied on small, homogeneous samples, reducing generalizability. Technical issues such as noisy signals, variable user performance, and the need for expert calibration persist across studies. While 16/19 studies reported successful implementations (Table 4), these barriers highlight the gap between laboratory feasibility and classroom scalability.

5.1.3. Discussion of RQ3: Research Gaps and Future Directions

RQ3: What are the main gaps and future directions in BCI research for inclusive education?

Several gaps emerged from the review. Demographic reporting was inconsistent, with limited data on gender, age, and disability type. Most studies recruited small numbers of young adults, while children, older adults, and learners from under-represented regions were rarely included; this can be seen in Table 2 and Figure 3. Reporting of technical specifications and costs was uneven, making comparisons across studies difficult. Ethical issues such as neurodata privacy and consent were acknowledged but not systematically addressed.

Future work should broaden participant diversity, standardize reporting of hardware, costs, and performance metrics, and incorporate robust ethical and governance frameworks. Low-cost, modular EEG systems and adaptive algorithms that personalize performance hold particular promise for practical classroom use. Interdisciplinary collaboration and co-design with educators and learners will be essential to translate prototypes into sustainable educational tools.

5.1.4. Overall Implications

Taken together, the findings across RQ1–RQ3 highlight both the promise and limitations of BCI applications in inclusive education. On the one hand, the dominance of EEG-based systems and the diversity of applications, from assistive mobility and communication to adaptive learning, demonstrate that BCIs can meaningfully enhance accessibility. On the other hand, barriers related to high equipment costs, small and homogeneous participant samples, and inconsistent reporting continue to constrain real-world deployment. For BCIs to transition from research prototypes to sustainable educational tools, technical advances must be paired with attention to affordability, usability, and inclusivity. Standardized reporting of costs and performance metrics will improve comparability across studies, while expanding participant diversity will strengthen generalizability. Ethical safeguards around neurodata privacy and informed consent are equally critical to responsible adoption. Ultimately, interdisciplinary collaboration between neuroscientists, engineers, educators, and policymakers, combined with co-design involving learners and teachers, will be essential to move BCIs beyond feasibility demonstrations toward scalable, equitable solutions that genuinely support inclusive education.

5.2. Methodological Strengths and Limitations

A key strength of this review lies in its broad scope, encompassing 19 studies conducted across multiple geographic regions, participant demographics, and application domains. By including both clinical and non-clinical populations, ranging from individuals with severe motor impairments to healthy participants engaged in experimental BCI training, the synthesis reflects the diversity of approaches used in BCI-based accessibility research. The use of the Mixed Methods Appraisal Tool (MMAT) [8] enhanced methodological rigor by providing a consistent framework for assessing studies with qualitative, quantitative, and mixed-method designs. This enabled a balanced evaluation of technical feasibility and user-centered outcomes across heterogeneous methodologies.

Several limitations must be acknowledged. The heterogeneity of study designs, participant populations, and reporting practices makes direct comparison of performance metrics such as classification accuracy, information transfer rate (ITR), and user satisfaction challenging. For example, while some studies reported detailed EEG hardware specifications, signal processing pipelines, and cost estimates [19,35,39], others omitted such details, limiting the ability to assess replicability and economic feasibility. Small sample sizes, common in BCI research, with some studies recruiting fewer than ten participants, further reduce the statistical power and generalizability of findings, particularly when focusing on specialized populations such as individuals with tetraplegia [2,5].

Inconsistent reporting of demographic variables also posed a challenge. Only 13 of the 19 studies reported gender distribution, and only 15 provided age ranges or mean ages. The degree of participant familiarity with BCI technology was rarely documented, despite evidence that prior exposure can significantly affect performance [21,24,26]. Furthermore, publication bias may have influenced the dataset, as studies with positive findings, particularly those demonstrating notable performance improvements through novel algorithms [20,24,34], are more likely to be published than studies reporting neutral or negative results.

Addressing these methodological limitations will require standardized reporting guidelines for BCI research in educational and assistive contexts, including the following:

• Detailed technical specifications of hardware and software;
• Transparent cost reporting for both acquisition and maintenance;
• Comprehensive demographic and prior-experience data;
• Consistent performance metrics to allow cross-study, culture, population, and domain comparability.

Such standardization would facilitate meta-analyses, support evidence-based policy, and accelerate the transition of BCIs from laboratory prototypes to practical, inclusive educational tools.

5.3. Implications for Future Research and Practice

The predominance of EEG-based BCIs, particularly those employing motor imagery and event-related potential paradigms [2,4,6,21,24,26,27,34], demonstrates that non-invasive systems remain the most viable option for real-world educational deployment. This can be seen in Table 4, where 17/19 studies have non-invasive systems. While invasive approaches are unlikely to be adopted in education, recent clinical trials have demonstrated the safety and long-term feasibility of fully implanted endovascular BCIs for individuals with severe paralysis [40]. Such advances in the medical domain highlight the rapid pace of innovation in BCI technology and may indirectly accelerate progress in non-invasive systems for broader applications.

The integration of advanced computational methods, including deep learning, ensemble classifiers, and transfer learning [20,22,24,26,34], has consistently improved classification accuracy and reduced calibration times. To bridge the gap between technical innovation and classroom use, these algorithms should be embedded into adaptive educational software capable of adjusting learning pace and content delivery in real time [7,14,15].

BCI research in educational contexts would also benefit from broader participant recruitment, encompassing diverse age groups, cultural backgrounds, and levels of technological literacy. Many current studies rely on small, homogeneous samples, often university students or individuals already familiar with technology [21,24,26]. Expanding recruitment to underrepresented populations, including those with varied disabilities and from low-resource environments, will enhance generalizability and equity of access. Given the ethical considerations surrounding neurodata collection in minors and vulnerable populations, future deployments should incorporate robust privacy and consent frameworks [11,13]. This includes transparent communication with participants and guardians, secure data storage, and clear guidelines for data reuse.

Finally, interdisciplinary collaboration between neuroscientists, educators, engineers, and policymakers will be essential for translating BCI prototypes into sustainable educational tools. Pilot programs co-designed with teachers and accessibility specialists could serve as proving grounds for evaluating usability, learning outcomes, and long-term engagement. By addressing cost, inclusivity, ethical safeguards, and real-world integration, future BCI systems can transition from experimental prototypes to reliable, scalable solutions that enhance accessibility and learning experiences.

5.4. Ethical and Accessibility Considerations

As BCIs transition from laboratory settings to real-world educational environments, ethical and accessibility considerations become critical to their sustainable adoption. The sensitive nature of neural data collected through EEG and other modalities raises important privacy, security, and consent challenges, particularly when involving minors or individuals with disabilities [11,13]. Future research should establish clear governance frameworks for data collection and use, emphasizing informed consent, anonymization, and adherence to applicable data protection regulations. Equitable access remains a significant concern. Many of the reviewed systems rely on high-cost, research-grade EEG devices [19,39], limiting adoption in underfunded schools and low-resource communities. Although low-cost, portable EEG systems have shown promising usability and performance [35,36,37], further validation is needed to confirm that these devices meet the reliability and accuracy standards required for educational deployment. Inclusivity must extend beyond hardware availability. Interface design should accommodate a wide range of user capabilities, including motor and cognitive impairments, to ensure BCIs do not inadvertently introduce new barriers to learning [6,14,15]. Co-design approaches involving students, educators, and assistive technology specialists can help ensure that applications are adaptable to diverse learning needs. Ethical deployment also requires balancing innovation with social responsibility. While BCIs hold considerable potential for enhancing participation and independence in education, implementations must safeguard user autonomy, prevent coercion, and ensure equitable distribution of benefits across populations [11,13]. Addressing these issues early in the development process will be essential to fostering public trust and ensuring long-term sustainability.

6. Conclusions

This scoping review synthesizes the growing body of research on BCI applications in inclusive education, encompassing diverse participant groups, recording modalities, and implementation contexts. The evidence demonstrates that BCIs can support assistive control, adaptive learning, and rehabilitation, thereby improving accessibility for learners with motor and communication impairments or disabilities [2,5,6,14,15,19]. The notable increase in publications after 2018 reflects advancements in affordable EEG hardware [35,36,37], machine learning-driven signal processing [20,21,24,26,34], and hybrid paradigms integrating multiple neural signals [22,24]. Despite this progress, several challenges persist. High equipment costs, the need for technical expertise, inconsistent reporting of participant demographics, and limited evaluation in authentic classroom environments constrain the scalability and generalizability of current solutions. Ethical concerns, including neurodata privacy, informed consent, and equitable access, must also be addressed to ensure that innovation benefits all learners, including those with disabilities [11,13]. Furthermore, the lack of standardized evaluation protocols and longitudinal studies limits the ability to assess the sustained educational impact of BCIs. While BCIs are not yet a mainstream educational technology, their potential to transform inclusive learning environments is clear. By advancing cost-effective hardware, adaptive algorithms, and user-centered design, future research can move closer to making BCIs a practical, ethical, and accessible tool for diverse educational settings and learners. Collectively, the findings address RQ1 by mapping the breadth of BCI applications in education, RQ2 by detailing the dominant modalities and computational approaches, and RQ3 by identifying key challenges and opportunities for scalable, inclusive implementation.