1. Introduction and problem

In the 3-to-6 age band—kindergarten, preschool, CENDI, early childhood school—a package of three artifacts has been installed that purport to stand for motor development. The first is a pose-estimation system that scores a jump or a run as “correct/incorrect”: a keypoint pipeline that compares the body’s trajectory with a template and declares that the child “already executed the skill well.” The second is a wearable that flags “motor milestones”: a device that logs accelerations, postures or thresholds and presents the alert as if it were pedagogical motor progress. The third is an app that classifies FMS from video: a model that labels running, jumping, throwing or catching and delivers the score as if it were gross- and fine-motor development. All three are visible and cheap in coordination time. They allow centers to exhibit that they “already do motor development with artificial intelligence.” The leap—from obtaining a keypoint score, receiving a milestone alert or accumulating FMS classifications to asserting that motor development is present—is authorized neither by computer-vision evidence nor by what motor pedagogy measures when there is situated practice of locomotion, object control and stability in contexts of active play and adult mediation.

The thesis of this article is restrictive. A pose-estimation system that scores a jump or a run as “correct/incorrect,” a wearable that flags “motor milestones,” or an app that classifies FMS from video do not constitute motor development in early childhood education. At this stage motor development is situated practice of locomotion, object control and stability—fundamental movement skills cultivated in active play, guided practice and teacher mediation in yard and classroom; pedagogical observation of movement that interprets quality of displacement, not only presence of the gesture—; not a keypoint score, a milestone dashboard or an automatic “correct” label. Zheng, Ye, Korivi, Liu, and Hong (2022), in a systematic review and meta-analysis of FMS in children aged 3–6, locate fundamental motor competence as a real domain of locomotion and object control, not as a classifier output. Zhang, Tang, Geng, Li, Liu, and Cai (2025) synthesize active-play interventions on FMS: the craft is active play, not the dashboard. That does not authorize translating “pose estimation is installed” as “motor development is present.” It authorizes asking what was measured: often keypoint error or correlation with human evaluation on motor tasks, not the reiterated, mediated practice when a four-year-old runs, jumps, throws and balances in the yard with an adult who observes, models and extends.

The problem is aggravated by five category confusions that product sheets omit and that this article separates rigorously. First: motor development is not STEM or educational robotics—kits, sequences or “programming the body” as a computational analogy. Second: it is not executive functions as inhibitory control, working memory or cognitive flexibility dressed as “attentive motricity.” Third: it is not SEL as a program of global socioemotional competencies. Fourth: it is not generic free play without mediation or motor intentionality. Fifth: it is not academic formative assessment—language or mathematics rubrics applied to the yard. This work does not recycle articles on interactions, executive functions, SEL, STEM, artistic creativity, free play, documentation, numeracy or orality. The question is about motricity: what counts as motor development when an early childhood center “does AI and motor skills.” Confusing the product—pose estimation, wearable, FMS app—with the process is the error this work names. Jahn et al. (2025) demonstrate technical ceilings of pose estimation in infants; Lei et al. (2025) correlate machine vision with human evaluation of coordination; Andarage et al. (2023) propose ECAMS as a TGMD-2 monitoring toolkit. Pedagogical inference, marked as such: the market for “AI for motor development in early childhood” inherits that evaluative distribution and turns it into a pedagogical promise. It converts automated measurement of movement into the whole of the motor craft.

The contributions are three: to reconstruct the state of the art that separates motor development (FMS; active play; yard/classroom mediation; pedagogical observation of movement) from keypoint scoring, milestone alerts and video-based FMS classification; to examine three families of empirical cases; and to offer four tests for deciding when a kindergarten may assert that motor development is present, and not only a pose-estimation system, a wearable or a classifying app.

2. State of the art: from situated motricity to the artifact on display

It is useful to separate four strata that the market for “AI for motor development in early childhood” usually mixes. The first is the construct of motor development at ages 3–6 as competence in locomotion, object control and stability—FMS; gross and fine motricity; coordination—cultivated in play and practice contexts (Zheng et al., 2022; Zhang et al., 2025; Zhu et al., 2025; NAEYC, 2022; OECD, 2021). The second is the pedagogical craft that cultivates it—active play, guided FMS practice, teacher mediation in yard and classroom, pedagogical observation of movement—(Zhang et al., 2025; OECD, 2021, 2023; NAEYC, 2022). The third is the evidence on pose estimation, machine vision, action monitoring, fine-motor computer vision and handwriting apps, plus mappings of AI in ECE (Jahn et al., 2025; Lei et al., 2025; Andarage et al., 2023; Kim and Neville, 2023; Butler, Pimenta, Tommerdahl, Fuchs, and Caçola, 2019; Li, Fu, Zheng, Gou, Yu, Kong, and Wang, 2023; Chen, 2024; Su and Yang, 2022; Ljungcrantz, 2026; Nikolopoulou, 2025). The fourth is the rights, systems and developmentally appropriate practice framework that treats the 3–6-year-old as a subject of situated motor practice, not as a vector of keypoints for a scoring pipeline (UNESCO, 2021; Miao and Holmes, 2023; European Commission, 2022; U.S. Department of Education, 2023).

In the construct stratum, Zheng et al. (2022) meta-analyze gender differences in FMS among children aged 3–6 and locate locomotion and object control as empirical domains of fundamental motor competence. Status: synthesis finding on real FMS in the kindergarten age range, not on AI. Zhu et al. (2025) review with COSMIN methodology the measurement properties of the Test of Gross Motor Development-3 (TGMD-3): the test is an instrument for gross-motor assessment, not everyday yard pedagogy. Status: metrological finding. Pedagogical inference, marked as such: a TGMD score—human or automated—is not, by existing, motor development in classroom life. Zhang et al. (2025) include nine studies of active-play interventions on FMS in typically developing children: seven report a significant effect and two do not; interventions last 45 to 60 minutes, once to four times per week, for periods from four weeks to six months. Status: synthesis finding that active play can improve FMS, with heterogeneity. Inference: the craft is mediated, situable AP; not a dashboard.

In the craft stratum, OECD (2021) anchors ECEC quality in meaningful everyday interactions; OECD (2023) requires that digitalization empower without replacing those interactions. NAEYC (2022) requires developmentally appropriate practice: the 3–6-year-old is a subject of play, movement and relationship, not of a pipeline. Inference: a pose-estimation score does not run with the child. An adult who organizes active play, models a throw and observes the quality of foot support does. Pedagogical observation of movement interprets intention, effort, transfer across contexts and safety; it is not exhausted by detecting whether an ankle keypoint crossed a threshold.

In the artifact stratum, Jahn et al. (2025) compare marker-less 2D infant pose-estimation methods with 4500 annotated frames from 75 recordings of spontaneous motor functions between 4 and 16 weeks of age; ViTPose stands out among generic estimators; the top-down view outperforms the diagonal; infant-specialized estimators do not generalize well to other infant datasets. Status: empirical finding of a technical ceiling. Explicit transfer: the sample is infants, not ages 3–6; it is not kindergarten motor pedagogy. Lei et al. (2025) propose evaluating children’s motor coordination ability via machine vision with keypoints and Dynamic Time Warping, correlated with human evaluation: scoring objectivity, not pedagogical motor development. Andarage et al. (2023) present ECAMS to detect sitting up, running, walking and jumping integrated with TGMD-2 in early ages (3–5): a monitoring toolkit, not a substitute for mediated active play. Kim and Neville (2023) demonstrate accuracy and feasibility of computer vision for assessing fine hand motor skill via object tracking: a technical finding; transfer marked when experimental ages do not match kindergarten 3–6. Butler et al. (2019), with N = 125 kindergarten children (experimental 58 / control 67), find that a handwriting app with stylus improves manual dexterity versus worksheets alone; cited here only as a technological limit/transfer outside the 2021–2026 core: a tracing app does not authorize equating app = integral motor development. Li et al. (2023) integrate visual-motor tracking in behavior analysis for developmental coordination disorder: clinical/rehabilitation context; transfer to kindergarten marked. Chen (2024), Su and Yang (2022), Ljungcrantz (2026) and Nikolopoulou (2025) map AI in ECE without equating it to yard motor development.

In the systems stratum, UNESCO (2021) requires human oversight. Miao and Holmes (2023) set pedagogical validation and age thresholds for generative AI. The European Commission (2022) and the U.S. Department of Education (2023) agree not to replace professional judgment. Inference: a four-year-old is not the input of a keypoint pipeline nor the passive recipient of a milestone wearable. The child is the protagonist of situated motor practice that a guaranteeing adult organizes, models and observes.

3. Review method

A critical narrative review was conducted, not a primary meta-analysis. The purpose was not to estimate a homogeneous effect size of pose estimation, but to articulate a pedagogical category argument with verified sources. Inclusion criteria: (a) 2021–2026, with transfer explicitly marked when the sample is not 3–6, and with Butler et al. (2019) only as a handwriting-app limit; (b) motor development, FMS, gross/fine motricity, coordination, active play, pose estimation, machine vision, action monitoring or AI in early childhood education; (c) relevance to kindergarten, preschool, CENDI or ages 3–6; (d) peer-reviewed journal, DOI or report from NAEYC, UNESCO, OECD, European Commission or education department; (e) verifiable DOI or editorial page. Axes already used in this series were excluded as central objects—CLASS/serve-and-return interactions, pedagogical documentation, SEL as substitute program, executive functions, STEM/robotics, artistic creativity, generic free play, numeracy, orality, generative tutoring, gaps, privacy, UDL, academic formative assessment, family–school, continuous training—although some appear as category boundaries.

The search was executed on 28 August 2026 (slot 09:02 America/Mexico_City) on DOI pages, Crossref, Springer, Nature, BMC, MDPI, IEEE Xplore, Tsinghua Science and Technology, OECD iLibrary, UNESDOC, JAIR, NAEYC and editorial sites. Each source was verified against at least one of those pages; authors of Butler et al. (2019) and Zhu et al. (2025) were corrected against DOI/Crossref before citation. Empirical finding, conceptual or normative framework, and pedagogical inference marked as such were distinguished. Priority was given to the distinction motor development / STEM / EF / SEL / free play / formative assessment, and to the caution against translating scoring–human correlation into motor pedagogy.

4. Case 1. Pose estimation, milestone wearables or FMS-classifying apps do not constitute motor development

Jahn, Flügge, Zhang, Poustka, Bölte, Wörgötter, Marschik, and Kulvicius (2025) publish in Scientific Reports the comparison that best names the ceiling of the first artifact when it is presented as motor development. They compare marker-less 2D estimators—generic and infant-specialized—on 4500 annotated frames from 75 recordings of spontaneous movement between 4 and 16 weeks; they evaluate error against human annotation and percentage of correct key-points; they find that ViTPose, trained on adults, is the best generic model; that retraining on own data improves; that infant estimators generalize poorly to other datasets; and that the top-down view outperforms the diagonal used in general movement assessment. Status of the evidence. Empirical finding of pose-estimation accuracy in infants. Explicit transfer: not a 3–6 sample nor kindergarten pedagogy. Pedagogical inference, marked as such: this is the gesture a center copies when it “does motor skills with AI that scores the jump.” Video is recorded, keypoints are estimated, “correct/incorrect” is declared. What exists is a technical vision ceiling. Motor development, in Zheng et al. (2022) and Zhang et al. (2025), asks for FMS practice in active play, not a high PCK on frames.

Lei, Shu, Yu, Shi, Li, and Chen (2025) saturate the portrait of coordination scoring. In Tsinghua Science and Technology they propose a machine-vision method with keypoints and Dynamic Time Warping to evaluate children’s motor coordination ability, correlated with human evaluation. Status: finding of scoring objectivity and feasibility. It is not a finding that installing a coordination scorer develops locomotion, object control or stability in yard life. Inference, marked as such: correlating with human judges authorizes saying there is more reproducible measurement; it does not authorize declaring “motor development is already present.” Andarage, Fernando, Lokuarachchi, Athuluwage, and Wijewickrama (2023) present ECAMS at IEEE PRDC: computer vision and machine learning integrated with TGMD-2 to detect sitting up, running, walking and jumping in early ages, as a real-time monitoring toolkit for educators, trainers and families. Status: system-design finding for monitoring. It is not a finding that the toolkit replaces mediated active play. Inference: detecting the action class is not teaching the skill nor mediating practice.

Kim and Neville (2023) name the fine-motor face of the third artifact. They develop 3D capture via computer vision with two action cameras for tracking an object manipulated by the hand and examine accuracy and feasibility for detecting changes in fine hand motor skill. Status: technical assessment finding. Transfer: when experiments do not match the kindergarten 3–6 band, they are not extrapolated as early childhood pedagogy. Butler et al. (2019), outside the temporal core, show with N = 125 that Letter School with stylus improves manual dexterity in kindergarten versus worksheets alone; both groups improve legibility. Status: handwriting-app finding as complement. Restrictive inference: it does not authorize equating tracing app = integral FMS motor development. Li et al. (2023) close with visual-motor tracking in DCD: clinical/rehab utility; not everyday CENDI pedagogy. Chen (2024), Su and Yang (2022), Ljungcrantz (2026) and Nikolopoulou (2025) confirm growth of AI in ECE without evidence that pose estimation, milestone wearables or FMS classifiers replace the craft. Inference: a kindergarten that delivers “FMS correct according to the model” or “motor milestones met according to the wearable” has done measurement or counting. Motor development is verified in whether the child practiced locomotion, object control and stability with adult mediation in active play. It is not verified in the estimator’s PCK.

5. Case 2. What the kindergarten does when motor development is present: active play, FMS and mediation

Zheng et al. (2022) locate FMS at ages 3–6 as an empirical domain with gender differences especially in object control. Status: meta-analysis finding on real fundamental motor competence. Not an AI finding. Pedagogical inference, marked as such: this is the object an early childhood center may call motor development. Motricity is not a keypoint score: it is a growing domain of running, jumping, galloping, throwing, catching, kicking, balancing and coordinating in meaningful contexts. A wearable that flags “jump milestone achieved” without reiterated practice may celebrate a threshold and, at once, empty the yard of mediation. An FMS classifier may label “run” in video and miss whether the adult modeled push-off, adjusted space or invited transfer of the pattern into a chase game.

Zhang et al. (2025) saturate the craft from active play: of 3672 screened records, nine studies enter the review; interventions of 45–60 minutes, one to four times weekly, from four weeks to six months; instruments TGMD-2, BOTMP, PDMS-2 and MABC; seven studies with significant FMS effects and two without; low risk of bias in three and moderate in six. Status: synthesis finding that AP tends to improve FMS with heterogeneity. Inference: when motor development is present in early childhood, there is active play organized with motor intentionality, not an alert feed. Zhu et al. (2025) remind that TGMD-3 has measurement properties reviewable by COSMIN: the test informs assessment; it does not replace daily pedagogy. OECD (2021) anchors quality in everyday interactions; OECD (2023) requires subordinated digitalization. NAEYC (2022) requires developmentally appropriate practice. Inference: teacher mediation in yard/classroom means noticing quality of support, throwing sequence, space safety and opportunity to reiterate; pedagogical observation of movement is professional reading, not a keypoint log. A dashboard does not model a gallop. An adult who organizes FMS stations, plays with the child and observes transfer across corners does.

Guided FMS practice does not deny play: it orients it. Zhang et al. (2025) underline active play’s potential through free choice and ease of dissemination, not through automation. Inference: the kindergarten that “does motor development” opens time, space and mediation for locomotion, object control and stability; documents with a pedagogical gaze if needed; does not declare development completed because ECAMS detected four action classes or because a coordination scorer correlated with a judge. Andarage et al. (2023) may assist monitoring; they do not sign development. Lei et al. (2025) may objectify a score; they do not replace practice. Jahn et al. (2025) show how costly pose estimation remains even under controlled infant conditions: a transfer that brakes the fantasy of a fully “scored” yard without friction.

6. Case 3. Motor development is not STEM, EF, SEL, generic free play or academic formative assessment

The first category boundary is STEM/robotics. A robot kit, a programming sequence or a metaphor of “the body as machine” may coexist in a center; they do not constitute, by themselves, FMS motor development at ages 3–6. Status: pedagogical inference marked as such, anchored in Zheng et al. (2022) and Zhang et al. (2025) as motor construct. This article does not make STEM its central object: it names it to forbid the equivalence. A pose-estimation system is not “motor STEM” either: it is geometric body estimation. Confusing “we have computer vision” with “there is mediated locomotion practice” is the category error the market exploits.

The second boundary is executive functions. This article does not convert motricity into inhibitory control, working memory or cognitive flexibility. Li et al. (2023) are retained for visual-motor tracking in clinical DCD, not for kindergarten executive training. Inference: when a product promises “improves executive attention because it classifies the jump,” another construct has been crossed. Motor development may associate empirically with many outcomes; it is not identified with EF.

The third boundary is SEL. Socioemotional competencies may appear in the yard; they are not FMS. Restrictive inference: a wearable that rewards “milestones” with emotional reinforcement does not turn reinforcement into object control. The fourth boundary is generic free play. Free play matters; it does not suffice, alone, as intentional FMS practice if there is no mediation, motor time or pedagogical observation of movement. Zhang et al. (2025) discuss active play with potential for FMS, not leaving the yard without craft. The fifth boundary is academic formative assessment. Language or mathematics rubrics do not measure gallop or ball reception. Zhu et al. (2025) locate TGMD-3 in gross-motor metrology: another instrument, another construct. Chen (2024), Su and Yang (2022), Ljungcrantz (2026) and Nikolopoulou (2025) map AI in ECE; Miao and Holmes (2023) set limits. Inference: the only AI use coherent with ages 3–6 remains on the side of the adult who organizes active play, guides FMS and observes movement—as support for training or deferred professional feedback—subject to pedagogical validation. It does not enter as an autonomous “correct/incorrect” scorer, a wearable that replaces judgment, or an app that declares FMS completed.

7. Inferential framework: four tests for asserting motor development, not an artifact

The framework that follows is pedagogical inference of this article, anchored in the cases and verified instruments. It is not a new international standard. It distinguishes four tests. If a kindergarten, preschool, CENDI or early childhood school fails them, it cannot declare that a pose-estimation system scoring “correct/incorrect,” a motor-milestone wearable or an app classifying FMS from video constitute motor development.

7.1. Test of situated practice of locomotion, object control and stability, not of the keypoint score. Zheng et al. (2022) define FMS at 3–6 as real competence. Jahn et al. (2025) estimate pose in infants with PCK and generalization ceilings. Lei et al. (2025) correlate machine vision with human judges. Inference: evidence of development is verified in whether the child practiced the skills in context. If the center’s “evidence” is a keypoint dashboard or a PCK, the center has done estimation, not pedagogical motricity.

7.2. Test of active play and guided FMS practice, not of the action detector. Zhang et al. (2025) document AP with heterogeneous FMS effects across nine studies. Andarage et al. (2023) detect sitting up, running, walking and jumping with ECAMS+TGMD-2. Inference: detecting the action class does not demonstrate mediated active play or reiterated practice. A monitoring toolkit may inform; it does not sign development.

7.3. Test of teacher mediation and pedagogical observation of movement, not of the milestone alert. NAEYC (2022) and OECD (2021, 2023) require appropriate practice and meaningful interactions/digitalization. Kim and Neville (2023) and Butler et al. (2019) show fine-motor assessment or dexterity apps—with transfer and temporal limit. Inference: “milestone met” may raise a threshold without raising mediation quality. Pedagogical observation reads movement; the alert counts an event.

7.4. Test of category distinction and professional judgment, not of the product catalog. Motor development ≠ STEM, EF, SEL, generic free play or academic formative assessment. Zhu et al. (2025) prevent confusing TGMD-3 properties with everyday pedagogy. Li et al. (2023) remain in clinical/rehab with marked transfer. UNESCO (2021), Miao and Holmes (2023), the European Commission (2022), the U.S. Department of Education (2023), NAEYC (2022) and OECD (2021, 2023) require human oversight, pedagogical validation and not replacing professional judgment. Inference: a center cannot treat the child as a keypoint vector. Motor development is not fulfilled by scoring the jump video better. It is fulfilled by practicing FMS in active play with adults who mediate and observe.

The framework admits the digital when subordinated to teacher education and validated professional observation (Andarage et al., 2023, as auxiliary monitoring; Lei et al., 2025, as subordinated scoring, not declared development). It rejects declaring motor development via autonomous pose estimation, milestone wearables or FMS-classifying apps (Jahn et al., 2025; Kim and Neville, 2023; Butler et al., 2019, as limit).

8. Discussion

Three tensions organize the discussion. The first is between exhibiting a movement-measurement artifact and enacting motor development. It is a finding that ViTPose and top-down views improve infant pose estimation, with generalization limits (Jahn et al., 2025; transfer); that machine vision can correlate coordination with human judges (Lei et al., 2025); and that ECAMS detects TGMD-2-linked actions (Andarage et al., 2023). It is a framework that motor development at 3–6 is played out in FMS and active play (Zheng et al., 2022; Zhang et al., 2025; NAEYC, 2022). It is not a finding that score, wearable or app produce the craft the yard requires. The three artifacts measure or label what engineering knows how to deliver and declare what only mediated practice would authorize.

The second is between automated assessment and everyday pedagogy. Zhu et al. (2025) show that TGMD-3 is an object of measurement properties; Andarage et al. (2023) automate detection aligned to TGMD-2. Inference: insisting that the kindergarten “already has motor development” because the model classifies enough jumps is inverted pedagogy. An evaluative proxy is made to stand for practice of locomotion, object control and stability. The child becomes a keypoint emitter; the adult, a classifier supervisor.

The third is between a fine-dexterity app and integral motor development. Butler et al. (2019) improve manual dexterity with a handwriting app in kindergarten; Kim and Neville (2023) make fine assessment by vision feasible. Inference: fragmenting motricity into a tracing gesture or object tracking and selling it as “motor development with AI” confuses part with whole. The only AI use coherent with ages 3–6 remains on the side of the adult who organizes active play and observes movement, subject to UNESCO, European Commission and U.S. Department of Education guidance. An autonomous scorer because the system needs a product is not that use.

Additional pedagogical inference, marked as such: the artifact series—pose estimation scoring correct/incorrect, milestone wearables and FMS-classifying apps—shares the same economy of visibility. Each produces an output readable for coordination: a number, a notification, a skill label. OECD (2021) and NAEYC (2022) require, instead, that development be verified in situated practice. Chen (2024), Su and Yang (2022), Ljungcrantz (2026) and Nikolopoulou (2025) show growth of the AI-in-ECE field; that growth does not authorize the equivalence. Miao and Holmes (2023), UNESCO (2021), the European Commission (2022) and the U.S. Department of Education (2023) agree on subordinating AI to professional judgment. In a CENDI or kindergarten, that means the adult remains guarantor of motricity: organizes active play, guides FMS, mediates in yard and classroom and observes movement. The artifact may assist training or deferred feedback; it cannot sign motor development.

9. Limits

This review is narrative. It does not apply PRISMA nor estimate primary pooled effects. Jahn et al. (2025) are infants aged 4–16 weeks: marked transfer, not kindergarten. Lei et al. (2025) validate coordination scoring, not everyday FMS pedagogy. Andarage et al. (2023) are ECAMS toolkit design, not a trial of mediated motor development versus autonomous scorer. Kim and Neville (2023) are computer-vision fine-motor feasibility, with age transfer when applicable. Butler et al. (2019) fall outside 2021–2026 and are used only as a handwriting-app limit (N = 125). Zhang et al. (2025) synthesize nine heterogeneous AP studies, not AI. Zheng et al. (2022) meta-analyze gender in FMS, not artifacts. Zhu et al. (2025) review TGMD-3 properties, not AI implementation in CENDI. Li et al. (2023) are clinical/rehab DCD. Chen (2024), Su and Yang (2022), Ljungcrantz (2026) and Nikolopoulou (2025) map AI in ECE, not yard motricity as primary variable. NAEYC, UNESCO and OECD are framework sources. No Latin American AI trials were located comparing adult-mediated active play versus autonomous pose estimation or milestone wearables in CENDI. Section 7 inferences are pedagogical category hypotheses, not implementation evidence.

10. Conclusions

A pose-estimation system that scores a jump or a run as “correct/incorrect,” a wearable that flags “motor milestones,” or an app that classifies FMS from video do not constitute motor development in an early childhood education center. Verified evidence does not authorize that declaration. Jahn et al. (2025) show pose-estimation ceilings on 4500 frames from 75 infant recordings (4–16 weeks), with ViTPose and top-down advantages and generalization limits—transfer, not 3–6 pedagogy. Lei et al. (2025) objectify coordination via machine vision correlated with human evaluation. Andarage et al. (2023) detect actions with ECAMS+TGMD-2 as a monitoring toolkit. Kim and Neville (2023) make fine assessment by vision feasible. Butler et al. (2019) improve manual dexterity with a handwriting app in kindergarten, without authorizing equivalence to integral motor development. By contrast, when motor development is present in early childhood, there is situated FMS practice: real domains at ages 3–6 (Zheng et al., 2022); active play with predominantly positive, heterogeneous effects across nine studies (Zhang et al., 2025); TGMD-3 as a measurement instrument, not a yard substitute (Zhu et al., 2025); mediation and developmentally appropriate practice (NAEYC, 2022; OECD, 2021, 2023). Motor development is distinguished from STEM/robotics, executive functions, SEL, generic free play and academic formative assessment. Current guidance requires human oversight, pedagogical validation and not replacing professional judgment (UNESCO, 2021; Miao and Holmes, 2023; European Commission, 2022; U.S. Department of Education, 2023; Chen, 2024; Su and Yang, 2022; Ljungcrantz, 2026; Nikolopoulou, 2025).

Where sources do not measure a kindergarten, this article does not assert it. Where they measure pose estimation, coordination scoring, action detection or dexterity apps, it does not translate them into pedagogical motor development. Accompanying three-to-six-year-olds in motricity is enacting practice of locomotion, object control and stability in active play, with teacher mediation in yard and classroom and with pedagogical observation of movement. The rest is keypoint score, milestone alert and algorithmic FMS classification. It is not motor development in early childhood education, and it must not be presented as what it is not.

Laboratorio Editorial de NEXTECH.IA / Ingeniero Mitre.

References

  1. Andarage, I. S. N., Fernando, D., Lokuarachchi, B. A., Athuluwage, M. G., y Wijewickrama, P. (2023). Early Childhood Action Monitoring and Analytics System (ECAMS). En 2023 IEEE 28th Pacific Rim International Symposium on Dependable Computing (PRDC) (pp. 1–6). IEEE. https://doi.org/10.1109/prdc59308.2023.00056
  2. Butler, C., Pimenta, R., Tommerdahl, J., Fuchs, C. T., y Caçola, P. (2019). Using a handwriting app leads to improvement in manual dexterity in kindergarten children. Research in Learning Technology, 27, 2135. https://doi.org/10.25304/rlt.v27.2135
  3. Chen, J. J. (2024). A scoping study on AI affordances in early childhood education: Mapping the global landscape, identifying research gaps, and charting future research directions. Journal of Artificial Intelligence Research, 81, 701–740. https://doi.org/10.1613/jair.1.16882
  4. European Commission. (2022). Ethical guidelines on the use of artificial intelligence (AI) and data in teaching and learning for educators. Publications Office of the European Union. https://doi.org/10.2766/153756
  5. Jahn, L., Flügge, S., Zhang, D., Poustka, L., Bölte, S., Wörgötter, F., Marschik, P. B., y Kulvicius, T. (2025). Comparison of marker-less 2D image-based methods for infant pose estimation. Scientific Reports, 15, 12148. https://doi.org/10.1038/s41598-025-96206-0
  6. Kim, B., y Neville, C. (2023). Accuracy and feasibility of a novel fine hand motor skill assessment using computer vision object tracking. Scientific Reports, 13, 1813. https://doi.org/10.1038/s41598-023-29091-0
  7. Lei, Y., Shu, D., Yu, M., Shi, D., Li, J., y Chen, Y. (2025). Evaluation method of motor coordination ability in children based on machine vision. Tsinghua Science and Technology, 30(2), 633–649. https://doi.org/10.26599/tst.2024.9010069
  8. Li, R., Fu, H., Zheng, Y., Gou, S., Yu, J. J., Kong, X., y Wang, H. (2023). Behavior analysis with integrated visual-motor tracking for developmental coordination disorder. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 31, 2164–2173. https://doi.org/10.1109/tnsre.2023.3270287
  9. Ljungcrantz, L. (2026). The interaction of AI and early childhood education. A state-of-the-art review 2020–2024. Early Childhood Education Journal, 54, 3565–3581. https://doi.org/10.1007/s10643-025-02079-3
  10. Miao, F., y Holmes, W. (2023). Guidance for generative AI in education and research. UNESCO. https://doi.org/10.54675/EWZM9535
  11. NAEYC. (2022). Developmentally appropriate practice in early childhood programs serving children from birth through age 8 (4.ª ed.). NAEYC. https://www.naeyc.org/resources/pubs/books/dap-fourth-edition
  12. Nikolopoulou, K. (2025). Child-centered integration of generative AI in early learning: Balancing promises and challenges. AI, Brain and Child. https://doi.org/10.1007/s44436-025-00023-1
  13. OECD. (2021). Starting Strong VI: Supporting meaningful interactions in early childhood education and care. OECD Publishing. https://doi.org/10.1787/f47a06ae-en
  14. OECD. (2023). Empowering young children in the digital age (Starting Strong). OECD Publishing. https://doi.org/10.1787/50967622-en
  15. Su, J., y Yang, W. (2022). Artificial intelligence in early childhood education: A scoping review. Computers and Education: Artificial Intelligence, 3, 100049. https://doi.org/10.1016/j.caeai.2022.100049
  16. UNESCO. (2021). Recommendation on the ethics of artificial intelligence. UNESCO. https://unesdoc.unesco.org/ark:/48223/pf0000381137
  17. U.S. Department of Education, Office of Educational Technology. (2023). Artificial intelligence and the future of teaching and learning: Insights and recommendations. U.S. Department of Education. https://www.ed.gov/sites/ed/files/documents/ai-report/ai-report.pdf
  18. Zhang, X., Tang, C., Geng, M., Li, K., Liu, C., y Cai, Y. (2025). The effects of active play interventions on children’s fundamental movement skills: A systematic review. BMC Pediatrics, 25, 40. https://doi.org/10.1186/s12887-024-05385-8
  19. Zheng, Y., Ye, W., Korivi, M., Liu, Y., y Hong, F. (2022). Gender differences in fundamental motor skills proficiency in children aged 3–6 years: A systematic review and meta-analysis. International Journal of Environmental Research and Public Health, 19(14), 8318. https://doi.org/10.3390/ijerph19148318
  20. Zhu, Y., Wang, J., Ding, Y., Qian, Y., Korivi, M., Chen, Q., y Ye, W. (2025). Assessing the measurement properties of the Test of Gross Motor Development-3 using the COSMIN methodology—A systematic review. Behavioral Sciences, 15(1), 62. https://doi.org/10.3390/bs15010062