Artificial intelligence and twice-exceptional (2e) learners in educational settings: a scoping review of an emerging evidence base (2020–2026)

The intersection of artificial intelligence and special education has long been heralded as a frontier for personalized learning, yet empirical data evaluating its efficacy for students with complex learning profiles remains remarkably thin. A comprehensive scoping review published in Frontiers in Psychology sheds light on this gap, examining the state of research regarding twice-exceptional (2e) students—individuals who are intellectually or creatively gifted while simultaneously managing one or more disabilities, such as autism spectrum disorder (ASD), attention-deficit/hyperactivity disorder (ADHD), or specific learning disabilities like dyslexia. Conducted by researcher Yasir Alsamiri and published in September 2026, the review evaluates literature spanning from 2020 to 2026, revealing that while the promise of AI is frequently cited, the actual evidence base is in its infancy and heavily concentrated in a narrow slice of educational technology.

Understanding the Twice-Exceptional Dilemma and the Masking Effect

To comprehend the significance of applying artificial intelligence to 2e learners, education specialists emphasize the unique psychological and cognitive profile of these students. Twice-exceptionality is characterized not by isolated traits, but by the complex interaction between high ability and co-occurring difficulties. A defining phenomenon in this population is the masking effect. High intellectual capacity often compensates for and obscures underlying disabilities, while the disability simultaneously depresses standard performance metrics and conceals the student’s true giftedness.

This dual-directional masking frequently leads to delayed identification, chronic misdiagnosis, and inappropriate classroom placement. Traditional standardized assessments, which rely on single-dimensional psychometric measurements, struggle to capture these asynchronous profiles. Furthermore, regular and special education teachers frequently report feeling underprepared to recognize 2e characteristics and unsupported by institutional infrastructure when trying to accommodate them. Proponents of educational technology have long argued that adaptive AI systems could serve as an ideal solution, offering dynamic challenge in areas of strength alongside necessary scaffolding for areas of difficulty. However, the new review questions whether these technological assumptions hold up under empirical scrutiny.

Chronology and Methodology of the Scoping Review

The scoping review was designed and reported in strict accordance with the PRISMA extension for scoping reviews (PRISMA-ScR). The search strategy was deliberately anchored around the student population rather than the technological intersection to ensure maximum retrieval from academic databases.

On August 20, 2026, a primary search was executed in the ERIC database using bounded terminology for twice-exceptional learners, applying a peer-reviewed filter. The initial search yielded 62 records. Recognizing that educational innovations frequently appear in themed journal editions, the researcher also hand-searched a dedicated 2025 special issue of the Journal of Advanced Academics focused on artificial intelligence in gifted education, recovering three additional records.

After screening 65 total titles and abstracts—and filtering out items published prior to 2020, those lacking an AI component, and non-peer-reviewed materials—only five full-text papers met the strict criteria for evaluation. Of those five, exactly two empirical studies met primary eligibility, both of which were published in 2025 within the same themed journal issue. Three other papers were evaluated at the full-text stage but ultimately categorized as boundary cases because they either addressed gifted education broadly without a specific 2e sample or explored conceptual frameworks rather than reporting direct learner outcomes.

Evaluating the Small Evidence Base

The two primary studies that cleared the eligibility hurdle represent starkly different approaches to educational technology, highlighting the fragmented nature of the field. The first study utilized a human-AI hybrid tutoring system—combining an intelligent tutoring framework with direct human instruction—over 32 writing sessions for 12 twice-exceptional bilingual adolescents diagnosed with motor dysgraphia and ADHD. This study reported measurable gains in handwriting fluency, fine motor control, and composition, with qualitative focus groups indicating high user acceptability due to immediate feedback and socio-emotional support.

The second included study took a synthetic approach, utilizing large language models (LLMs) to construct evidence-informed synthetic learner profiles representing twice-exceptional and multi-functional neurodivergent students. While this study demonstrated the potential of generative AI in helping educators rehearse differentiated pedagogy, it simultaneously exposed significant challenges regarding the consistency, context-appropriateness, and bias of generated outputs, prompting the authors to stress the absolute necessity of rigorous human oversight.

Crucially, the review found a complete absence of empirical studies examining learning analytics or AI-supported identification systems among explicitly identified 2e populations. While generative AI and tutoring applications are beginning to appear in research, the systemic tools most likely to impact identification and tracking at a macro level remain entirely unstudied in this specific demographic.

The Paradox of Personalization and Performance-Keyed Algorithms

One of the most provocative analytical contributions of the review centers on a potential contradiction between how educational AI functions and how twice-exceptional minds operate. Most modern educational AI platforms infer a student’s cognitive state, ability level, and learning trajectory based on observed performance data.

Alsamiri hypothesizes that because 2e learners are defined by masking, performance-keyed AI systems may inadvertently reinforce the very barriers they are meant to dismantle. If an adaptive algorithm reads disability-driven errors as definitive proof of low ability rather than as noise masking a concurrent high capacity, the system may scale down difficulty or misunderstand the student’s needs. Consequently, rather than penetrating the masking effect, the technology might faithfully encode it, treating the student through a distorted lens backed by algorithmic authority.

To structure future inquiries, the review proposes an affordance-risk framework categorized across distinct AI modalities—ranging from generative text tools and adaptive learning platforms to conversational agents and learning analytics. Every proposition within this framework is explicitly labeled as hypothesized rather than demonstrated, serving as an explicit call to action for the academic community.

Implications for Policy, Practice, and Future Research

The publication of this scoping review arrives at a critical juncture for school districts and policymakers rushing to integrate automated learning tools into classrooms. The primary takeaway for educators is a cautionary one: the confident marketing language surrounding personalized AI for neurodivergent and twice-exceptional students currently outpaces empirical validation.

Experts suggest several immediate directions for the field. First, researchers must move beyond general neurodivergent or gifted classifications to isolate twice-exceptional cohorts in empirical trials. Second, developers of learning analytics and adaptive platforms must design validation studies that specifically test whether algorithms can differentiate between actual low ability and masked high ability. Finally, teacher preparation programs must address the digital and diagnostic literacy gap, ensuring that educators interpreting AI recommendations possess the specialized training required to spot complex 2e profiles.

Ultimately, the review reframes the conversation surrounding artificial intelligence in special education. The task ahead is not merely to synthesize an existing body of literature, but to deliberately construct one—ensuring that the technological revolution in schools accounts for the complex realities of students who live at the intersection of exceptional brilliance and profound neurological challenge.