A comprehensive scoping review examining early childhood education has systematically mapped the algorithmic architectures, implementation forms, and functional roles of emerging intelligent technologies introduced between 2010 and 2024. Published in the journal AI, Brain and Child, the investigation synthesizes findings from twenty-nine peer-reviewed empirical studies to understand how tools such as machine learning, virtual reality, data mining, robotics, and the Internet of Things are reshaping pedagogical practices for young learners. The researchers established that contemporary early learning tools are increasingly moving away from passive screen media toward intelligent, interactive systems that leverage sensor arrays, predictive neural networks, and immersive virtual simulations. However, the synthesis also uncovers critical structural challenges across the published literature, including alarming cybersecurity vulnerabilities in connected classroom toys, a pervasive shortage of longitudinal efficacy tracking, and a frequent failure to ground sophisticated algorithms in established child development theories.
To capture the full scope of technical implementations without sacrificing pedagogical relevance, the research team conducted a rigorous literature search following Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. The investigative timeline commenced in January 2010—a milestone marking the market introduction of the Apple iPad, which redefined touch-screen interfaces and mobile digital interaction—and concluded in October 2024. Guided by an expert in technology-enhanced learning, the authors queried three major interdisciplinary repositories: Web of Science, EBSCO, and ProQuest. These databases were specifically selected for their expansive indexing of computer science, engineering, and biomedical research alongside educational scholarship, whereas education-specific databases like ERIC were deliberately excluded due to their limited indexing of deep algorithmic architectures. The initial query yielded 1,349 total candidate records, comprising 601 from Web of Science, 452 from ProQuest, and 296 from EBSCO. Utilizing Zotero reference management software, the team filtered out 172 duplicates before importing the remaining 1,177 citations into the Rayyan systematic review platform for blinded screening. Title and abstract evaluations eliminated 1,075 non-relevant records, leaving 102 full-text articles for in-depth eligibility assessment.
During the secondary eligibility phase, two independent reviewers assessed each full-text candidate against stringent inclusion and exclusion parameters, resolving differences through structured consensus or adjudication by a third senior reviewer. This phase resulted in the disqualification of 73 articles: seventeen evaluated educational levels beyond early childhood, nineteen lacked empirical data, nine were published in non-English languages, seven presented inaccessible full texts, and twenty-one addressed tangential survey topics such as general educator attitudes rather than direct pedagogical implementations. The resulting core corpus of twenty-nine empirical studies underwent qualitative content analysis using MAXQDA 2022 analytical software. Coding categories captured four principal dimensions: study context, specific digital technologies, underlying algorithmic methods, and applied educational functions.
Across the twenty-nine analyzed investigations, the researchers identified five primary technology categories: nine studies investigated machine learning models, six focused on virtual reality environments, five examined data mining methodologies, five explored Internet of Things architectures, and four evaluated educational robotics. Within the Internet of Things domain, research predominantly concentrated on smart sensor integration to automate physiological monitoring and bolster physical safety. For example, investigative deployments by Osimani and colleagues demonstrated that wearable sensors tracking heart rate and skin temperature can detect early physiological distress in preschool environments. In parallel, Chen and collaborators engineered a crowdsourced localization framework that deployed wearable accelerometers, mobile iBeacons, and networked smartphones to identify hazardous holding-up behaviors and rapidly pinpoint missing children. Smart toys equipped with inertial measurement units have also served diagnostic objectives, as demonstrated by the MoVEAS system developed by Bondioli and co-authors, which captured atypical motor patterns during natural play to flag early markers of autism spectrum disorder, while platforms by Rivera and peers gathered fine motor metrics to detect developmental psychomotor delays.
Machine learning applications formed the largest single cluster within the scoped evidence base, operating through passive multimodal data acquisition to construct automated classification and predictive diagnostic models. Addressing data scarcity in young cohorts, Li and collaborators implemented a transfer learning paradigm that adapted adult movement datasets to calibrate highly accurate child activity models for injury prevention. Beyond physical metrics, machine learning demonstrated diagnostic utility in psychological screening; McGinnis and co-researchers recorded three-minute audio samples during structured pediatric speech tasks, utilizing automated acoustic feature extraction to identify internalizing anxiety and depressive disorders with greater diagnostic precision than conventional parent-reported surveys. In clinical forecasting, Su and team analyzed longitudinal electronic health records to predict pediatric suicide risk, while Harvey and Kumar demonstrated that Random Forest algorithms achieved superior accuracy over alternative classifiers in identifying early asthma development from national health survey variables.
The six analyzed virtual reality studies highlighted the medium’s capacity to construct controlled, highly immersive simulations that foster social cognition, executive function, and emotional resilience among children. Didehbani and colleagues validated a Virtual Reality Social Cognition Training protocol that utilized expressive digital avatars and structured interactive dilemmas to enhance emotion recognition, social attribution, and analogical reasoning in youth diagnosed with high-functioning autism. Similarly, Lorenzo and peers deployed immersive virtual reality systems to train children with autism in recognizing complex social cues within scaffolded environmental simulations. In a sizable intervention involving 94 children aged six to twelve, Ip and collaborators demonstrated that completing a twenty-eight-session virtual reality program across fourteen weeks yielded statistically significant gains in emotion regulation, facial expression decoding, and reciprocal social communication. Furthermore, immersive technologies addressed cognitive screening and educational equity: Nolin and associates developed the Classroom-CPT virtual environment to measure attention and impulse control against simulated classroom distractions, while Wang and team showed that coupling virtual reality with modular Lego construction significantly elevated social competence among students in under-resourced rural schools.
The scoping review identified five prominent data mining studies that leveraged advanced analytical techniques and large administrative or clinical datasets to optimize health surveillance and resource allocation. Bizzego and colleagues applied data mining algorithms to demographic indicators from low- and middle-income nations, uncovering that children with diagnosed intellectual disabilities suffered disproportionate neglect in essential cognitive caregiving activities. In pediatric pharmacology, Ferrajolo and co-authors mined electronic health records to identify adverse drug reactions and highlight high-risk pediatric medication combinations, while Naydenova and associates applied predictive algorithms to clinical variables to diagnose childhood pneumonia and assess infection severity in resource-constrained clinics. In community-level epidemiological research, Ivančević and colleagues utilized association rule mining via the Apriori algorithm to identify major risk factors for early childhood caries among Serbian preschool cohorts, isolating significant associations with high birth order, low birth weight, and parental language barriers. Similarly, Tesfaye and collaborators applied J48 decision trees and PART rule induction to Ethiopian demographic health surveys, isolating maternal literacy, birth spacing, and infant feeding habits to construct web-based predictive tools for pediatric mortality reduction.
Educational robotics interventions analyzed across four dedicated studies illustrated the power of tangible, playful programming interfaces to cultivate spatial cognition, mechanical design intuition, and collaborative problem-solving. Research conducted by Zviel-Girshin and co-authors indicated that early engagement with programmable robotics fostered technological self-efficacy and successfully minimized traditional gender divides in technological engagement. Expanding on cognitive metrics, Sisman and colleagues completed a 31-week longitudinal study demonstrating that structured robotics instruction significantly elevated spatial visualization and mental rotation capabilities among young participants. In a controlled trial, Canbeldek and Isikoglu evaluated the Productive Children Coding and Robotics Education Program, showing that combining unplugged logic games, Bee-Bot manipulatives, and Scratch Jr. block coding generated statistically significant improvements in creative problem-solving and language development compared to control groups. Beyond standard curricula, social robotics demonstrated valuable clinical and communicative utility: Nagae and Lee paired the humanoid NAO robot with E4 wristband sensors measuring electrodermal activity to interpret emotional states in children with developmental disabilities, while Kim and associates fine-tuned large language models on pediatric dialogue datasets to help companion robots comprehend non-standard speech and phonological mispronunciations.
Despite the pedagogical and diagnostic promises documented across these five technological domains, the scoping review revealed acute, persistent vulnerabilities in cybersecurity, privacy preservation, and experimental scale. Critical investigations into commercial smart toys highlighted severe gaps between theoretical security protocols and real-world deployment. While cryptographic architectures such as AES-128 encryption, sliced message authentication codes, and HMAC integrity checks have been engineered for Internet of Things toys by researchers like Rivera and colleagues, these protective frameworks were validated exclusively in small laboratory testbeds rather than scalable commercial deployments. Furthermore, the scoping authors observed that most empirical investigations concentrated narrowly on isolated technical competencies or momentary task completion, frequently overlooking broader psychological dimensions such as empathetic interaction, collaborative negotiation, and long-term socio-emotional maturity.
The review also delineated significant conceptual gaps that continue to impede the seamless translation of digital tools into early childhood pedagogy. Several reviewed interventions exhibited an evident disconnect between engineering complexity and developmental theory; for instance, virtual reality platforms designed to foster social skills frequently omitted validated developmental psychology frameworks from their simulation scenarios, leading to an over-reliance on technical novelty rather than grounded pedagogical progression. Methodologically, the literature suffers from a notable absence of longitudinal tracking, as evidenced by studies like Sisman and colleagues’ robotics course, which successfully established post-intervention spatial gains but failed to assess subsequent skill retention or durable interest over extended developmental intervals. To resolve these challenges, the authors advocate for interdisciplinary research combining mixed-methods evaluation frameworks, digital twin growth modeling, brain-computer interfaces for neurodiverse expression, and federated learning paradigms that protect sensitive student records while training predictive early-warning models, thereby ensuring that emerging educational technologies advance child-centric learning safely, equitably, and sustainably.
Subject of Research: Algorithmic implementations and functional roles of digital technologies in early childhood education practice
Article Title: Forms and functions innovation: a scoping review of digital and intelligence technologies in early childhood education practice
Article References: Liu, J., Chen, M., He, H., Liu, H., Luo, W., & Li, H. (2025). Forms and functions innovation: a scoping review of digital and intelligence technologies in early childhood education practice. AI, Brain and Child, 1(1), Article 8. https://doi.org/10.1007/s44436-025-00010-6
Image Credits: AI Generated
DOI: 10.1007/s44436-025-00010-6
Keywords: early childhood education, scoping review, machine learning, virtual reality, educational robotics, Internet of Things, data mining, cybersecurity in education, Forms, functions, innovation, scoping
News Source: Blake Davidson. (November 11, 2026). Scoping Review Uncovers Algorithmic Trends and Security Risks in Early Childhood Tech. Scienmag.



