1. Introduction and problem

On the stretch from kindergarten, preschool or a CENDI centre into primary school, a package of three artefacts has been installed that claim to count as transition. The first is a score: a “school-readiness” or “healthy and ready to learn” rating that classifies the child as on track, needs support or at risk. The second is a model: a random-forest, extra-trees or boosting classifier that, with household, survey or administrative variables, predicts who will “arrive prepared”. The third is an early-warning system: a dashboard that, with attendance, demographics or behavioural indicators, flags who is “at risk” before a threshold is crossed. All three are visible, auditable and cheap in coordination time. They allow a setting to exhibit, to supervisors, networks or families, that it “already does the transition with artificial intelligence”. The enunciative leap is enormous: one moves from classifying, predicting or alerting to asserting that a pedagogical transition exists. That leap is not authorised by the evidence on ECEC–primary transitions or by the evidence on developmental prediction.

The thesis of this article is restrictive. A score or a model that “predicts” preparedness for primary school is not a pedagogical transition. Transition is a relational process of children, families and teachers; it is not an algorithmic threshold. The global report on early childhood care and education states the institutional point clearly: the right to a strong foundation is not exhausted by an “on-track” indicator; it requires environments, educators and families that sustain learning and well-being (UNESCO & UNICEF, 2024). Starting Strong VI anchors ECEC quality in process quality: everyday interactions with adults, peers, materials and space are the most proximal engine of development (OECD, 2021). A score extracts a signal, assigns a category and returns a report. It can produce a datum. It does not produce pedagogical continuity.

The problem is sharpened by a professional reason that product sheets omit. SDG 4.2 asks that, by 2030, all girls and boys have access to quality early childhood development, care and pre-primary education so that they are “ready for primary education”. Indicator 4.2.1 is operationalised, at population scale, with the Early Childhood Development Index 2030 (ECDI2030): twenty caregiver-reported items for children aged 24–59 months (Halpin, de Castro, Petrowski & Cappa, 2024; UNICEF, 2023). That instrument is a population monitor. It is not an individual diagnosis and not a rite of passage. When a setting, a district or a vendor converts the same gesture — threshold, label, ranking — into a model that “predicts each child’s readiness”, it displaces the educational act toward classification. In 2022, 63.6% of children aged 3–5 in the United States were considered “healthy and ready to learn”; one million, 9.0%, needed support in multiple domains (Ghandour et al., 2024). Those numbers describe a distribution. They do not describe a transition.

This article does not recycle axes already treated in this series. The question is one of pedagogical category: what counts as transition when an early childhood setting “does AI”. The contributions are three: to reconstruct the state of the art that separates school readiness as a child attribute from the relational process of passage; to examine three families of cases — prediction of development and readiness, early warning of absenteeism from prekindergarten, and ECEC–primary transitions observed as practice; and to offer four tests for deciding when a kindergarten may claim that it transitions, and not merely that it scores, predicts or alerts.

2. State of the art: from the ready child to the receiving school

Four strata that the “AI for transition” market usually mixes should be kept apart. The first is school readiness as a multidimensional construct of the child: cognitive, linguistic, socioemotional and motor. The second is the ECEC–primary transition as a relational, temporal and situated process involving children, families, ECEC educators and primary teachers. The third is prediction: models that, with administrative or survey data, classify “on track” or “at risk”. The fourth is the rights and high-risk framework, which treats automated evaluation of learning and assignment to educational institutions as practices that cannot evade human oversight or the best interests of the child.

In the readiness stratum, Garon-Carrier, Mavungu-Blouin, Letarte, Gobeil-Bourdeau and Fitzpatrick (2024) synthesise eight longitudinal person-centred studies. They identify fifteen school-readiness profiles before school entry, seven of which are associated with worse later academic and social outcomes. The most vulnerable — low-balanced — groups, in normative samples, between 7% and 13% of preschoolers. At-risk profiles share features: linguistic or ethnic minorities, low socioeconomic status, less preschool attendance and, more often, boys. A binary threshold — ready / not ready — does not describe that heterogeneity. Inference, marked as such: a single score compresses mixed profiles into a number. The number is not the profile, and the profile is not the transition.

In the transition stratum, González-Moreira, Ferreira and Vidal (2024) gather more than three hundred variables of the move into primary school into twenty-nine groups, academic and non-academic; 67% of articles consider both. Research has polarised toward literacy and logico-mathematical competence because, in many countries, that step inaugurates compulsory schooling. Degli Esposti and Cigala (2025), with 23 articles from 2008 to 2025, shift the focus to children’s experience: prior expectations, later perceptions and learning processes; a gradual process and supportive environments emerge as essential. Boylan et al. (2024) translate that into practice: effective transitions require schools to connect and empower children and families, not to classify them at the door. González-Moreira, Ferreira and Vidal (2023) also document an adult-centred bias: 5- to 7-year-olds are seldom included. Inference: a model that predicts them without them does not repair that gap. It deepens it.

In the prediction stratum, Benson, Chelangat, Brink, Mwangala, Waljee, Moyer and Abubakar (2025) map 27 machine-learning studies in early childhood development (0–8 years). Seventy-eight per cent come from high-income countries; none from sub-Saharan Africa. 74.1% are limited to prediction; only 7.4% validated out of sample; 41% explained the model; one reached a clinical trial. Tebeje et al. (2025) operate the gesture at scale: 12,860 children aged 24–59 months in Kenya, Mozambique and Tanzania; an ensemble of random forest and extreme gradient boosting; 66% accuracy and AUC 0.71 in predicting whether the child is “on track” on the ECDI2030. Halpin et al. (2024) recall what that index was designed for: comparable population monitoring, not individual classification of access to primary school. Inference: predicting the ECDI2030 is not transitioning. It is reusing a policy instrument as a child label.

In the rights stratum, Regulation (EU) 2024/1689 treats as high-risk, in Annex III, AI systems intended to determine access or admission to education and vocational training institutions, to assign persons to those institutions, to evaluate learning outcomes, or to determine the appropriate educational level a person will receive or to which they will have access (European Union, 2024). A readiness score used to decide whether a child “passes” or “waits” is not an analytic toy: in that territory it is a high-risk system. The Recommendation on the ethics of AI requires human oversight, proportionality and particular attention when children are involved (UNESCO, 2021). UNICEF (2021) requires that the best interests of the child be prioritised and development supported. The 2022 ethical guidelines and the U.S. Department of Education report agree that professional judgement and the teacher must not be replaced (European Commission, 2022; U.S. Department of Education, 2023). Miao and Holmes (2023) set an age threshold for independent conversations with generative platforms and require pedagogical validation. Inference: a five-year-old is not the user of a prediction dashboard. The child is the subject of a relation that the dashboard does not sustain.

3. Review method

A critical narrative review was conducted, not a meta-analysis. The purpose was not to estimate a homogeneous effect size across a survey model, a district early-warning system and a redesign of passage practices, but to articulate an argument of pedagogical category with verified sources. Inclusion criteria: (a) 2021–2026; (b) school readiness, prediction of child development or school readiness, early-warning systems in early childhood, or ECEC–primary transitions; (c) relevance to kindergarten, preschool, CENDI or ages 3–6; (d) peer-reviewed journal, DOI, or UNESCO, OECD, UNICEF, European Union, European Commission or ministry report; (e) verifiable DOI or publisher page. Axes already used in this series were excluded, including SEL/emotion chatbots, centre leadership and adaptive tutors as object.

The search was executed on 25 August 2026 on DOI pages, Springer, Elsevier, PLOS, BMJ, Frontiers, MDPI, Taylor & Francis, Annenberg/EdWorkingPapers, OECD iLibrary, UNESDOC, UNICEF, EUR-Lex, ERIC and PubMed/PMC. Each source was checked against at least one of those pages. The corpus was organised into prediction of readiness and ECD; early warning in the first grades; and evidence of relational ECEC–primary transitions.

The analysis distinguished three enunciative statuses. Empirical finding: what was observed or measured in the sample. Conceptual or normative framework: what a framework, guide or regulation prescribes. Pedagogical inference: the translation to kindergartens, preschools, CENDI and infant schools, marked as such. The limits are those of a narrative review (section 9).

4. Case 1. Predicting “developmentally on track” is not transitioning into primary school

Tebeje, Tesfaye, Sisay, Seboka, Tesfa, Sisay and colleagues (2025) publish in PLOS One the study that best illustrates the leap the thesis rejects. With the ECDI2030 module of the Demographic and Health Surveys of Kenya, Mozambique and Tanzania, they analyse 12,860 children aged 24–59 months. 57.4% (95% CI: 56.5–58.3) were developmentally on track in health, learning and psychosocial well-being. They train seven supervised algorithms and an ensemble of the two best — extreme gradient boosting and random forest — to predict that label. The ensemble reaches 66% accuracy, 73% sensitivity, specificity 0.56, weighted F1 0.64 and AUC 0.71. The three predictors with the largest mean absolute SHAP values are child age (+0.17), maternal media exposure (+0.12) and maternal education (+0.10). The beeswarm shows that ages 24–35 months, mothers not exposed to media, and mothers with at least secondary education are associated with a higher probability of being “on track”. The authors conclude with policy recommendations: prioritise pre-primary education, women’s education and appropriate media use.

Status of the evidence. Empirical finding of population classification: given certain survey covariates, an ensemble reproduces the ECDI2030 label in two cases out of three. It is not a finding of pedagogical transition. The model does not observe the first day of primary school, the bond with the new teacher, continuity of play, or conversation between the kindergarten and the school. It is not a finding of individual validity: the ECDI2030 was designed for population monitoring (Halpin et al., 2024; UNICEF, 2023). An AUC of 0.71, which the authors themselves compare with 0.67 for an analogous model in Bangladesh, does not authorise an admission decision. Pedagogical inference, marked as such: the gesture a kindergarten copies when it “predicts school readiness” is exactly this. A milestone threshold is taken, a probability is assigned, and management of the transition is declared. What exists is a development file. SDG 4.2 asks for preparedness for primary school as the outcome of a care and education system. A SHAP dashboard is not that system.

Benson et al. (2025) saturate the portrait. Of 759 records, 27 studies meet criteria. Supervised classifiers (40.7%) and deep learning (22.2%) predominate. The most frequent outcome is cognitive development (33.3%); the most used measure, Bayley-III (33.3%). Data are mainly image, video and sensors. 74.1% remain at prediction. Only one reached clinical implementation. 59% do not explain the model. Inference: the “AI for child development” literature produces classifiers, not transitions. Confusing a laboratory F1 with a process of passage between institutions is the category error this article names.

Ghandour et al. (2021, 2024) show what happens when the score becomes a national indicator. In 2016, with a pilot measure on 7,565 children aged 3–5, 42.2% were considered “healthy and ready to learn”; the lowest domain was early learning skills (58.4% on-track). In 2022, with 11,121 children and a revised measure — on-track in four or five domains and none that “needs support” — the figure rises to 63.6%; one million, 9.0%, needed support in multiple domains. Status: prevalence finding, not transition. Inference: if changing the cut moves a country’s “readiness” by fifteen points, the threshold is not a fact about the child. It is a measurement convention. Halpin et al. (2024), with samples from Mexico (n = 1,641) and Palestine (n = 1,099), support the ECDI2030 for population monitoring. The index is not offered as a gate to primary school.

5. Case 2. Flagging absenteeism from prekindergarten is not a pedagogy of passage

Wu and Weiland (2025) publish, in EdWorkingPapers of the Annenberg Institute, the case that best illustrates the second artefact: the early-warning system brought into early childhood. With a demographically diverse sample of 6,698 students followed from prekindergarten to third grade in Boston Public Schools, they compare the logistic regression typical of EWSs in use with contemporary algorithms — SMOTE and XGBoost — to predict who will become chronically absent in third grade. Chronic absenteeism is defined as missing 10% or more of school days, for any reason. The best model, XGBoost with SMOTE, achieves a 54-percentage-point improvement in recall over the logistic model closest to current EWSs. Models that exclude student demographic information maintain comparable predictive accuracy. The study also asks how many years of history are needed to decide an early intervention and how different probability thresholds can accommodate budget constraints.

Status of the evidence. Empirical finding of attendance prediction: in a large urban district, boosting with oversampling better identifies, from the first grades, those who will be chronically absent in third grade. It is not a finding of ECEC–primary transition. The object is absenteeism, not the relational process of passage. It is not a finding that alerting produces pedagogical continuity, joint play, family–school bonds or curricular alignment. The authors are explicit: many EWSs were designed to predict high-school dropout; taking them to prekindergarten is an analytic extension, not a kindergarten pedagogy. Pedagogical inference, marked as such: the gesture a setting copies when it “does the transition with an early warning” is this. An attendance signal is taken, a risk is assigned, and early warning of school readiness is declared. What exists is a forecast of absences. Missing school is a serious problem, associated with achievement, socioemotional skills and, later, dropout (Wu & Weiland, 2025). Intervening on attendance can be a just policy. That does not make the dashboard the transition. Transition occurs when an ECEC educator and a first-grade teacher coordinate around a concrete child, with the family, in a time that does not fit in a recall statistic.

Wu and Weiland also show that omitting demographics does not collapse prediction. Inference: if the model “works” without race or income, it does not follow that the alert is pedagogically neutral. It follows that the attendance signal already concentrates inequalities the district will not resolve with a threshold. A more accurate EWS may help allocate attendance supports. It does not make the primary school receive the child the kindergarten knew. OECD (2021) locates that work in process quality. UNESCO and UNICEF (2024) locate it in the right to a strong foundation and in the fall of net enrolment one year before primary — from 75% in 2020 to 72% in 2022. A risk-of-absence dashboard does not enrol, does not accompany and does not align pedagogies.

Kim (2024) offers a useful contrast and a limit. With the ECLS-K:2011 cohort (n = 18,174), she shows that teachers’ perceptions of approaches to learning, mathematical thinking and science, at kindergarten entry, predict later mathematics and science achievement through fifth grade. Status: finding that teacher judgement at entry is associated with trajectories. It is not an AI finding. It is not a finding that automating that judgement improves the transition. Inference: if the kindergarten teacher’s report already predicts, the added value of a model is often not to “discover” the child. It is to duplicate an adult perception in the vocabulary of accuracy. The transition is not that prediction. It is what happens between the report and the new classroom.

6. Case 3. What the kindergarten does when there is a transition: a relational process

Boylan, Barblett, Lavina and Ruscoe (2024) published in the European Early Childhood Education Research Journal (vol. 32, pp. 704–718) the study that, in this corpus, best anchors what a pedagogical transition actually is. With a funds-of-knowledge and funds-of-identity lens, children aged 3–6 and their teachers reimagined practices of passage into primary school. Through a design-based thinking process, the team collected qualitative data from four professional-learning days, individual coaching sessions, observations of the first days of school and stakeholder interviews. Case studies were developed in collaboration with each participating school. Three design principles, formulated by the teachers themselves, positively affected teachers’ and schools’ preparation for transitions. All stakeholders described the process as useful for transforming practices and for thinking about transitions from different perspectives (Boylan et al., 2024). ERIC indexes the work as empirical research with case study, coaching and family–school relationship in Australia (ERIC, 2024, EJ1431215).

Status of the evidence. Empirical finding of practice redesign: transition, when taken seriously, appears as joint work of teachers, children and families, mediated by the knowledge the child brings and the identity the child puts into play. It is not an AI finding. It is not a finding that a score reproduces those three principles. Pedagogical inference, marked as such: this is the object an early childhood setting may properly call transition. It is a relational, temporal and situated process. An ECDI classifier does not ask about funds of knowledge. An attendance EWS does not observe the first day. A “ready / not ready” dashboard does not empower the family in the process: at best it informs them of a label.

Degli Esposti and Cigala (2025) saturate the portrait with 23 articles: expectations before the step, perceptions afterwards, learning processes, and a gradual process as an essential factor. González-Moreira et al. (2024) show why the score is not enough: more than three hundred variables, twenty-nine groups. Reducing that ecology to a milestone threshold — seven to fifteen depending on age, in the ECDI2030 (Tebeje et al., 2025) — is an operation of statistical policy, legitimate at country scale, illegitimate as a child’s rite of passage. González-Moreira et al. (2023) document an adult-centred view in the methods. Inference: a model trained on the mother’s or the teacher’s report inherits that adult-centrism and presents it as objectivity.

Garon-Carrier et al. (2024) close the contrast: fifteen profiles, seven at risk. The finding helps design supports. It does not authorise treating the profile as an entry verdict. The authors locate the response in quality care and family–school partnerships, not in a threshold. OECD (2021) aligns curriculum as a lever of continuity. TALIS Starting Strong 2024 locates weekly practices in observing interactions and in improving how children play together (OECD, 2025). Inference: evidence of transition is verified in those interactions and alliances, not in an AUC. If the “evidence” is a readiness ranking or a risk traffic-light, the setting has done product management, not pedagogy of passage.

7. Inferential framework: four tests for claiming there is a transition, not a threshold

The framework that follows is pedagogical inference of this article, anchored in the cases and in the verified instruments. It is not a new international standard. It distinguishes four tests. If a kindergarten, preschool, CENDI or infant school does not pass them, it cannot declare that a readiness score, a model that predicts preparedness or an early-warning system constitutes a pedagogical transition.

7.1. Test of the relational process, not of the threshold. Boylan et al. (2024) show transition as redesign with children, families and teachers. Degli Esposti and Cigala (2025) show gradual child experience. González-Moreira et al. (2024) show an ecology of more than three hundred variables. Inference: evidence of transition is verified in the bond, in shared time, in continuity of practices and in the child’s voice, not in a milestone cut. If the “evidence” that the setting transitions is an on-track score or a risk log, the setting has classified, not passed.

7.2. Test of pedagogical continuity, not of prediction. OECD (2021) locates curriculum as a lever of alignment across stages and process quality as the engine of development. OECD (2025) locates interactions and joint play as weekly practices. Tebeje et al. (2025) predict a survey label. Wu and Weiland (2025) predict absences. Inference: predicting a later outcome is not producing the continuity that outcome would require. A model may correctly identify who will be absent in third grade. It has not thereby aligned the kindergarten with the primary school. Kim (2024) shows that teacher judgement at entry is already associated with trajectories: automating it does not, by itself, add pedagogy of passage.

7.3. Test of the population instrument, not of the individual verdict. Halpin et al. (2024) and UNICEF (2023) design the ECDI2030 for population monitoring and for indicator 4.2.1. Ghandour et al. (2021, 2024) show that changing the HRTL cut moves national prevalence from 42.2% to 63.6%. Garon-Carrier et al. (2024) show profiles, not a bit. Benson et al. (2025) show that most ECD models are neither validated out of sample nor implemented. Inference: a survey or dashboard threshold is not a readiness diagnosis for deciding a child’s passage. Using it as such confuses the SDG with an admission rite.

7.4. Test of rights and high risk, not of the product dashboard. Annex III of Regulation (EU) 2024/1689 includes, among high-risk systems, those intended to determine access or admission to educational institutions, to assign persons to those institutions, to evaluate learning outcomes or to determine the appropriate educational level (European Union, 2024). UNESCO (2021) requires human oversight and particular attention to children. UNICEF (2021) requires the best interests of the child. European Commission (2022) and U.S. Department of Education (2023) require that the teacher not be replaced. Miao and Holmes (2023) require pedagogical validation and an age threshold for autonomous use of generative platforms. Inference: an early childhood setting cannot treat the 3- to 6-year-old as the object of an automated “ready for primary” verdict. The threshold is not “met” by lowering the age of the classified child. It is respected by not using prediction as a gate. In European territory, a system that assigns a place or a level with AI is not a transition innovation: unless the regulation’s safeguards apply, it is a high-risk practice.

The framework admits the ECDI2030 and HRTL as population monitors (Halpin et al., 2024; Ghandour et al., 2024; UNICEF, 2023); EWSs as attendance tools, not as pedagogy of passage (Wu & Weiland, 2025); readiness profiles as hypotheses for support, not as verdicts (Garon-Carrier et al., 2024); and teacher judgement as relational information, not as input to a classifier that replaces it (Kim, 2024). It rejects declaring transition by a readiness score, a model that predicts the ECDI or HRTL at child scale, an early-warning traffic-light presented as “transition management”, or the treatment of the early-years child as the user of an algorithmic threshold (Tebeje et al., 2025; Benson et al., 2025; European Union, 2024; Boylan et al., 2024).

8. Discussion

Three tensions organise the discussion. The first is between predicting and transitioning. It is a finding that an ensemble predicts the ECDI2030 with AUC 0.71 in 12,860 East African children (Tebeje et al., 2025); that 27 ML studies in ECD concentrate on prediction, with scarce external validation and almost no implementation (Benson et al., 2025); and that an XGBoost improves recall of chronic absenteeism by 54 points from prekindergarten (Wu & Weiland, 2025). It is framework that the ECDI2030 is a population monitor (Halpin et al., 2024) and that assigning level or access with AI is, in the European Union, high-risk (European Union, 2024). It is not a finding that predicting a label produces the process Boylan et al. (2024) observe. The politics of the three artefacts — score, model, alert — measure what engineering knows how to measure and declare what only relational transition would authorise.

The second is between the child as attribute and the system as receiver. Garon-Carrier et al. (2024) show that readiness is not a single trait. Ghandour et al. (2021, 2024) show that the indicator cut changes who counts as “ready”. González-Moreira et al. (2024) show that transition is researched as an ecology, not as a bit. UNESCO and UNICEF (2024) show that the world is off track for 4.2 and that enrolment one year before primary receded. Inference: insisting that the child “be ready” while the system does not align is an inverted pedagogy. The score blames the child for what curriculum, time and bond do not sustain. OECD (2021) had warned that the lever is process quality and continuity across stages. The readiness model performs the inverse operation: it extracts the child from the process and returns the child as a probability.

The third is between the dashboard and the voice. Degli Esposti and Cigala (2025) and González-Moreira et al. (2023) document that children aged 5–7 remain little heard in research on the passage. Boylan et al. (2024) show what happens when they are included: practices change, not the ranking. An early-warning system speaks of them in the third person. A readiness score speaks of them as a vector. Neither asks for the funds of knowledge Boylan places at the centre. Datafying that voice — turning it into a feature — does not repair the silence. It administers it.

9. Limits

This review is narrative. It does not apply PRISMA or estimate pooled effects. Tebeje et al. (2025) predict a survey label, not a primary-classroom outcome; transfer to the rite of passage is inference. Benson et al. (2025) cover ECD ages 0–8, not only the ECEC–primary transition, and exclude neurodevelopmental disorders. Wu and Weiland (2025) is a district working paper on absenteeism, not a trial of transition practices; the institutional Annenberg DOI is verified, and preprint status is declared. Boylan et al. (2024) is qualitative in independent schools in Australia. Degli Esposti and Cigala (2025) and González-Moreira et al. (2023, 2024) are reviews, not AI observations. Garon-Carrier et al. (2024) review profiles from 2005–2022; several base studies predate 2021. Ghandour et al. (2021, 2024) are U.S. parent surveys; the change of measure forbids reading the rise from 42.2% to 63.6% as a net improvement in development. Kim (2024) is teacher judgement, not a model. Halpin et al. (2024) validate a population index. UNESCO, UNICEF, the OECD, the European Union and the 2022–2023 guides are prescriptive. Latin American trials of AI school-readiness prediction at the kindergarten–primary step were not located with the same degree of DOI. The inferences in section 7 are hypotheses of pedagogical category, not implementation evidence.

10. Conclusions

A school-readiness score, a model that predicts preparedness for primary school or an early-warning system does not constitute a pedagogical transition in an early childhood setting. The verified evidence does not authorise that declaration. An ensemble that reproduces the ECDI2030 in 12,860 children classifies a survey label; that is population prediction, not passage (Tebeje et al., 2025). Twenty-seven machine-learning studies in child development remain, in their great majority, at prediction without external validation or implementation (Benson et al., 2025). An EWS that improves recall of chronic absenteeism from prekindergarten identifies absences; that is an attendance alert, not pedagogical continuity (Wu & Weiland, 2025). National HRTL moves when the indicator cut changes (Ghandour et al., 2021, 2024). The ECDI2030 is offered as a population monitor, not as a child verdict (Halpin et al., 2024; UNICEF, 2023). Fifteen readiness profiles refute the single threshold (Garon-Carrier et al., 2024). By contrast, when there is a transition, there is relational redesign with children aged 3–6, families and teachers (Boylan et al., 2024); there is gradual child experience (Degli Esposti & Cigala, 2025); there is an ecology of factors that does not fit in a bit (González-Moreira et al., 2024). Current law requires human oversight, the best interests of the child and, in the European Union, treating as high-risk the AI that determines access, admission, assignment or educational level (UNESCO, 2021; UNICEF, 2021; European Union, 2024; European Commission, 2022; U.S. Department of Education, 2023).

Where the sources do not measure a kindergarten, this article does not assert it. Where they measure classification or alert, it does not translate them into transition. Accompanying children aged three to six into primary school is sustaining a relational process of children, families and teachers. The rest is a threshold. It is not transition, and it must not be presented as what it is not.

Laboratorio Editorial de NEXTECH.IA / Ingeniero Mitre.

References

  1. Benson, F. N., Chelangat, D., Brink, W., Mwangala, P. N., Waljee, A. K., Moyer, C. A., & Abubakar, A. (2025). Application of machine learning in early childhood development research: A scoping review. BMJ Open, 15(8), e100358. https://doi.org/10.1136/bmjopen-2025-100358
  2. Boylan, F., Barblett, L., Lavina, L., & Ruscoe, A. (2024). Transforming transitions to primary school: Using children’s funds of knowledge and identity. European Early Childhood Education Research Journal, 32(4), 704–718. https://doi.org/10.1080/1350293X.2023.2291354
  3. Degli Esposti, C., & Cigala, A. (2025). Children’s experience of transition from preschool to primary school: A scoping literature review. Early Childhood Education Journal. https://doi.org/10.1007/s10643-025-02019-1
  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. European Union. (2024). Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence. Official Journal of the European Union, L 2024/1689. https://eur-lex.europa.eu/eli/reg/2024/1689/oj
  6. Garon-Carrier, G., Mavungu-Blouin, C., Letarte, M.-J., Gobeil-Bourdeau, J., & Fitzpatrick, C. (2024). School readiness among vulnerable children: A systematic review of studies using a person-centered approach. Psicologia: Reflexão e Crítica, 37, 16. https://doi.org/10.1186/s41155-024-00298-y
  7. Ghandour, R. M., Hirai, A. H., Moore, K. A., Robinson, L. R., Kaminski, J. W., Murphy, K., Lu, M. C., & Kogan, M. D. (2021). Healthy and ready to learn: Prevalence and correlates of school readiness among United States preschoolers. Academic Pediatrics, 21(5), 818–829. https://doi.org/10.1016/j.acap.2021.02.019
  8. Ghandour, R. M., Hirai, A. H., Moore, K. A., Paschall, K., LaForett, D. R., Reddington, E., & Kogan, M. D. (2024). School readiness among United States children: Results from the 2022 National Survey of Children’s Health. Academic Pediatrics, 24(7), 1049–1061. https://doi.org/10.1016/j.acap.2024.02.013
  9. González-Moreira, A., Ferreira, C., & Vidal, J. (2023). Review on research methods for studying transition from early childhood education to primary education. Education Sciences, 13(3), 254. https://doi.org/10.3390/educsci13030254
  10. González-Moreira, A., Ferreira, C., & Vidal, J. (2024). A journey to primary education: A systematic review of factors affecting the transition from early childhood education to primary education. Preventing School Failure, 69(1), 77–91. https://doi.org/10.1080/1045988X.2024.2335679
  11. Halpin, P. F., de Castro, E. F., Petrowski, N., & Cappa, C. (2024). Monitoring early childhood development at the population level: The ECDI2030. Early Childhood Research Quarterly, 67, 1–12. https://doi.org/10.1016/j.ecresq.2023.11.004
  12. Kim, J. (2024). The relationships between teachers’ evaluation of children’s academic readiness and children’s later outcomes. International Journal of Child Care and Education Policy, 18, 6. https://doi.org/10.1186/s40723-024-00131-0
  13. Miao, F., & Holmes, W. (2023). Guidance for generative AI in education and research. UNESCO. https://doi.org/10.54675/EWZM9535
  14. OECD. (2021). Starting Strong VI: Supporting meaningful interactions in early childhood education and care. OECD Publishing. https://doi.org/10.1787/f47a06ae-en
  15. OECD. (2025). Results from TALIS Starting Strong 2024: Strengthening early childhood education and care. OECD Publishing. https://doi.org/10.1787/20af08c0-en
  16. Tebeje, T. M., Tesfaye, S. H., Sisay, G., Seboka, B. T., Tesfa, G. A., Sisay, D., et al. (2025). Application of machine learning algorithms to predict early childhood development in children aged 24–59 months across three East African countries. PLOS One, 20(9), e0332023. https://doi.org/10.1371/journal.pone.0332023
  17. UNESCO. (2021). Recommendation on the ethics of artificial intelligence. UNESCO. https://unesdoc.unesco.org/ark:/48223/pf0000381137
  18. UNESCO & UNICEF. (2024). Global report on early childhood care and education: The right to a strong foundation. UNESCO. https://doi.org/10.54675/FWQA2113
  19. UNICEF. (2021). Policy guidance on AI for children 2.0. UNICEF Innocenti. https://www.unicef.org/innocenti/reports/policy-guidance-ai-children
  20. UNICEF. (2023). The Early Childhood Development Index 2030: A new measure of early childhood development. UNICEF. https://data.unicef.org/resources/early-childhood-development-index-2030-ecdi2030/
  21. U.S. Department of Education, Office of Educational Technology. (2023). Artificial intelligence and the future of teaching and learning: Insights and recommendations. https://www.ed.gov/sites/ed/files/documents/ai-report/ai-report.pdf
  22. Wu, T., & Weiland, C. (2025). Leveraging modern machine learning to improve early warning systems and reduce chronic absenteeism in early childhood (EdWorkingPaper 24-1081). Annenberg Institute at Brown University. https://doi.org/10.26300/xxvz-cv94