1. Introduction and problem

In the 3-to-6 age band —kindergarten, preschool, CENDI, early childhood school— a package of five artifacts has been installed as if it counted as support for water play: a sensor or wearable that “scores pouring accuracy” or a “pouring accuracy score” / “STEM water score” for the child; a computer-vision / water-table camera system that labels “mess” or “off-task splashing” or triggers “mess / off-task splashing” alerts; a chatbot/GenAI that generates “water experiments” or “sink-or-float plans” or “water play lesson plans” by prompt (for the child or as a recipe the adult reads); a dashboard of “minutes at the water table” / “inquiry score” / “exploration minutes at water table” for administrative surveillance; and an AR layer that “guides” pouring or predicts sink/float. The five allow centers to display that they “already do water play with AI.” The leap —from pouring score, mess alert, generated experiment, minutes metric or AR guidance to claiming support— is not authorized by AI-in-ECE mappings nor by the pedagogy of water play when there is real water (flow, capacity, sink-float, mixtures), child agency to explore without an expected product, adult co-presence that names properties and sustains conversation, and tolerance of mess as a condition of the medium, not as error.

The thesis of this article is restrictive. A sensor or wearable that “scores pouring accuracy” or a “pouring accuracy score” / “STEM water score” for the child, a computer-vision / water-table camera system that labels “mess” or “off-task splashing” or triggers “mess / off-task splashing” alerts, a chatbot/GenAI that generates “water experiments” or “sink-or-float plans” or “water play lesson plans” by prompt, a dashboard of “minutes at the water table” / “inquiry score” for administrative surveillance, or an AR layer that “guides” pouring or predicts sink/float do not constitute support for water play in the early years. In early childhood water play develops with real water, child agency to explore without an expected product, adult co-presence that names properties and sustains conversation (Lange, 2022; Counsell, 2024; precursor sink-float model of Canedo Ibarra & Gómez Galindo, 2022), and tolerance of mess; not a pouring score, a mess classifier, an experiment generator that closes uncertainty, or a minutes dashboard. Lange (2022) states, from NSTA early-years craft, that water play is science: the liquid medium puts real physical properties into play. Counsell (2024) asks where the math is in water play and answers with capacity, volume and comparison at the water table mediated by the adult, not by a STEM water score. Amalia and Harahap (2022) —N = 15; one PAUD Khairul Ummah center— report a water-playing method and science in Group B: marked empirical finding with sample limit. Park and Park (2021) describe sinking or floating as a mediated inquiry-based STEM activity, not as an AR prediction app. That authorizes asking what was measured: pouring accuracy, mess label, lesson plan by prompt, inquiry score or AR prediction —not the practice when a child pours, observes flow, tests sink-float, mixes, splashes and receives adult co-presence that names properties without closing uncertainty.

The problem is worsened by six category confusions. First: water play is not outdoor/nature play as axis —water may be indoor at a table; the axis is the liquid, not outdoor minutes—. Second: it is not curricular scientific inquiry as axis —there may be science IN the play (Lange, 2022; Park & Park, 2021); the axis is the water medium—. Third: it is not STEM/robotics —Counsell (2024) and Park and Park (2021) are craft of the medium, not a STEM score—. Fourth: it is not loose parts —Sando et al. (2023) anchor objects-in-play (water/sand); here the medium is water—. Fifth: it is not free play as axis. Sixth: it is not curricular motor skills. This work does not recycle outdoor play, free play, loose parts, risky play, block play/spatial, symbolic play, motor skills, STEM/robotics, artistic creativity, SEL, executive functions, environmental education, scientific inquiry as axis, literacy, orality, UDL, documentation, formative assessment, participation, planning, joint attention, adult–child interactions as generic axis, teacher education, music or numeracy. Maraisane et al. (2024) —Grade R teachers’ conceptual knowledge of floating/sinking— is a peripheral teacher contrast, not a continuing-education axis already published. The question is what counts as support for water play when a center “does AI and water play.”

There is, moreover, an economy of coordination: pouring-accuracy sensor, mess camera, GenAI experiments, minutes dashboard and predictive AR fit on a slide; the prolonged episode with real water, child agency, tolerated mess and co-presence that names flow, capacity or sink-float does not. NAEYC (2022) and OECD (2021, 2023) require meaningful interactions and subordinated digitalization. Inference: support is not fulfilled by scoring pouring or generating “water experiments” by prompt. The contributions are three: separate craft from artifacts; examine cases —Amalia and Harahap (2022) and Park and Park (2021) as marked empirical/inferential; Maraisane et al. (2024) as peripheral teacher contrast—; and offer four tests of support for water play.

2. State of the art: from situated water-play practice to the artifact on display

It is useful to separate four strata that the “AI for water play in ECE” market often blends. The first is the construct of water play ages 3–6 as open exploration of a liquid material with real physical properties —pouring, flow, capacity/volume, sink-float, mixtures, splashing— and mess as a condition of the medium (Lange, 2022; Counsell, 2024; Canedo Ibarra & Gómez Galindo, 2022; Amalia & Harahap, 2022; Park & Park, 2021; Sando et al., 2023). The second is the pedagogical craft that cultivates it —offer of real water at table or tray, child agency without expected product, adult co-presence that names properties and converses, tolerance of mess, not substituting exploration with a pouring score or a mess classifier— (Lange, 2022; Counsell, 2024; Canedo Ibarra & Gómez Galindo, 2022; Maraisane et al., 2024, only as teacher-knowledge contrast; NAEYC, 2022; OECD, 2021, 2023). The third is evidence on AI affordances in ECE and child-centered GenAI —without equating them to situated water play with real liquid— (Chen, 2024; Su & Yang, 2022; Su & Zhong, 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 who explores the liquid, not as a vector of signals for a pouring-accuracy pipeline nor as a passive recipient of GenAI experiments or AR predictions that close uncertainty (UNESCO, 2021; Miao & Holmes, 2023; U.S. Department of Education, 2023).

In construct and craft, Lange (2022) articulates water play is science: NSTA framework. Counsell (2024) situates capacity, volume and comparison at the table: craft framework. Canedo Ibarra and Gómez Galindo (2022) document social interaction and a floating/sinking precursor model in preschool. Amalia and Harahap (2022) —N = 15; PAUD Khairul Ummah; pretest-posttest; pouring, color, “walking water”— provide an empirical finding with sample and single-center limits. Park and Park (2021) describe sinking or floating as a mediated inquiry-based STEM activity, not predictive AR. Sando et al. (2023) —N = 928 two-minute video observations— associate objects (water/sand) with deep-level learning: objects-in-play finding, not loose parts. Maraisane et al. (2024) explore floating/sinking among Grade R teachers: peripheral contrast. Inference: a pouring accuracy score does not name liquid properties; a child who pours and an adult who converses do.

In the artifact stratum, Chen (2024), Su and Yang (2022), Su and Zhong (2022) and Ljungcrantz (2026) map affordances, curricula and the state of the art of AI in ECE without equating them to situated water play. Nikolopoulou (2025) balances promises and challenges of child-centered GenAI under teacher mediation: caution framework. In the systems stratum, UNESCO (2021) requires human oversight. Miao and Holmes (2023) set pedagogical validation and age thresholds for generative AI —direct framework for generators of “water experiments” or sink-or-float lesson plans by prompt—. U.S. Department of Education (2023) requires AI to support, not replace, professional judgment. OECD (2021, 2023) anchor meaningful interactions and subordinated digitalization; NAEYC (2022) anchors developmentally appropriate practice.

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 the five artifacts, but to articulate a pedagogical-category argument with verified sources. Inclusion criteria: (a) 2021–2026, with marked transfer when the sample is not equivalent to ages 3–6; (b) water play, water table, sink-float, pouring/flow/capacity, or AI in ECE with artifact/craft relevance; (c) kindergarten, preschool, CENDI or ages 3–6; (d) peer-reviewed journal, DOI or NAEYC/UNESCO/OECD report; (e) verifiable DOI or editorial page. Axes already used in this series were excluded as central objects —generic outdoor play, free play, loose parts, risky play, block play/spatial, symbolic play, STEM/robotics, artistic creativity, motor skills as axis, SEL, executive functions, environmental education, curricular scientific inquiry as axis, literacy, orality, UDL, documentation, formative assessment, participation, planning, joint attention, adult–child interactions as generic axis, teacher education, music and numeracy—. Maraisane et al. (2024) only as peripheral Grade R teacher-knowledge contrast; the article is not converted into continuing education already published. Sando et al. (2023) are read for objects-in-play (water/sand), not as a loose-parts axis. Park and Park (2021) and Counsell/Lange are read as craft of the water medium (science/math IN the play), not as a scored lab nor as curricular inquiry already published.

The search was executed on 3 September 2026 (slot 13:02 America/Mexico_City) on DOI pages, Crossref, Springer, Elsevier, Taylor & Francis, MDPI, JAIR, OECD iLibrary, UNESDOC, NAEYC, NSTA and editorial sites. Each source was verified. Empirical finding, framework and pedagogical inference were distinguished. No N, d, r, AUC or DOI were invented: when an artifact lacks a verified study with pouring-accuracy scoring, mess/off-task splashing alerts, inquiry-score dashboard or AR sink/float prediction metrics in ages 3–6, it is discussed as a category ceiling supported by AI-in-ECE mappings (Chen, 2024; Su & Yang, 2022; Ljungcrantz, 2026) and by the teacher-knowledge contrast (Maraisane et al., 2024). The eighteen verified corpus sources were used.

4. Case 1. Sensor/pouring accuracy score / STEM water score, CV/water-table camera with mess or off-task splashing alerts, GenAI of “water experiments” / sink-or-float lesson plans, minutes/inquiry-score dashboard or AR that guides pouring or predicts sink/float do not constitute support for water play

Chen (2024), Su and Yang (2022) and Ljungcrantz (2026) saturate the portrait of the artifact ceiling when AI in ECE is presented as if it were support for water play. Chen (2024) maps global AI affordances in early childhood education: scoping finding on emerging uses —tutoring, analytics, content generation—, not a finding that a pouring-accuracy wearable cultivates open liquid exploration with flow, capacity or sink-float. Su and Yang (2022) review the AI-in-ECE field: synthesis finding on trends, not on tolerated mess or co-presence that names properties. Ljungcrantz (2026) reviews AI–ECE interaction 2020–2024: state-of-the-art finding, not situated water play. Pedagogical inference, marked as such: the gesture “the child had high pouring accuracy = there was support for water play” is a precision-analytics ceiling. Lange (2022), Counsell (2024) and Canedo Ibarra and Gómez Galindo (2022) ask for real water, conversation about properties and social construction of a precursor model; a score useful for administration can coexist with absence of agency, tolerated mess and co-presence that names without closing.

Nikolopoulou (2025) and Miao and Holmes (2023) name the risk of the generator of “water experiments” or “sink-or-float plans” / “water play lesson plans” from a prompt that closes uncertainty. Nikolopoulou (2025) balances promises and challenges of child-centered GenAI under teacher mediation: caution framework. Miao and Holmes (2023) require pedagogical validation and age thresholds for generative AI. Status: normative and review framework, not a trial of lesson plans by prompt versus a water-play episode with real water and co-presence that sustains conversation (Lange, 2022; Counsell, 2024). Inference: producing a “water experiments” plan by prompt may be subordinated teacher preparation; support for water play begins when there is real liquid, child agency to pour, test sink-float or mix without an expected product, mess as a condition of the medium, and an adult who names properties without turning the episode into a recipe. Su and Zhong (2022) propose AI curriculum design in ECE as a future direction —AI literacy curriculum, not situated water play—. Limit inference: an AI curriculum does not sign liquid exploration.

The computer-vision / water-table camera that labels “mess” or “off-task splashing” or triggers “mess / off-task splashing” alerts, the “minutes at the water table” / “inquiry score” dashboard and the AR layer that “guides” pouring or predicts sink/float lack, in the verified corpus, trials with reported N, d, r or AUC for pouring-accuracy scoring, mess/off-task splashing classification, inquiry-score dashboard or AR sink/float prediction in preschool ages 3–6; figures are not invented. Chen (2024), Su and Yang (2022) and Ljungcrantz (2026) map analytics, tutoring, agents and generation as an AI-in-ECE trend, without equivalence to mediated water play. Amalia and Harahap (2022) measure a water-playing method with N = 15: situated practice finding in one center, not an autonomous mess classifier. Sando et al. (2023) measure human video observations (N = 928): objects-in-play and deep-level learning finding, not a minutes dashboard. Maraisane et al. (2024) document teachers’ conceptual knowledge of floating and sinking: empirical contrast of legitimate peripheral use only when the adult’s human knowledge sustains the craft. Restrictive inference: Grade R teacher knowledge (Maraisane et al., 2024) ≠ pouring-accuracy sensor on the child; GenAI experiments ≠ Lange/Counsell who open properties without closing; teacher empowerment ≠ administrative inquiry score; AR that predicts sink/float ≠ precursor model built in social interaction (Canedo Ibarra & Gómez Galindo, 2022). UNESCO (2021) and U.S. Department of Education (2023) require human oversight. The five artifacts share one substitution grammar: the pouring accuracy score speaks for the child’s exploration; the mess alert substitutes pedagogical tolerance of the medium; GenAI closes uncertainty; the dashboard substitutes pedagogical observation; AR directs or predicts what craft leaves open.

It is useful to specify the ceiling without inventing effects. A sensor can score pouring; Counsell (2024) articulates human capacity, not AUC. A camera can label “off-task splashing”; Lange (2022) treats water play as science of the medium. A GenAI can print experiments; Miao and Holmes (2023) require validation. An AR can predict sink/float; Canedo Ibarra and Gómez Galindo (2022) document the social precursor model. A dashboard can count minutes; NAEYC (2022) and OECD (2021) anchor practice. Inference: “high pouring accuracy” or “lesson plan according to the model” does not sign support.

5. Case 2. What the kindergarten does when there is support for water play: liquid properties, Lange, Counsell, Amalia and Harahap, Park and Park, Canedo Ibarra and Gómez Galindo —with Maraisane et al. as peripheral contrast

Lange (2022) saturates the craft floor: water play is science —flow, sink-float, mixtures with adult co-presence, not a STEM water score—. Status: NSTA framework. Counsell (2024) answers “Where’s the math?” with capacity, volume and comparison at the table: mathematical framework IN the play, not a scored lab. Pedagogical inference: a center may call it support for water play when it protects real water, agency and conversation about properties —not when it scores pouring—. Water play is not a pouring accuracy score: it is pouring, observing flow, comparing capacity, testing sink-float, mixing, splashing and sustaining mess as a condition of the medium, while the adult names without a fixed expected product.

Amalia and Harahap (2022) provide the marked empirical case: N = 15 Group B children at PAUD Khairul Ummah (Medan Johor; 2021/2022); water-playing method to improve science; pouring, color and “walking water” activities; pretest-posttest design. Status: empirical finding. Marked limit: small sample, single center; cautious transfer toward Latin American CENDI/kindergarten; not a trial of a mess camera or GenAI lesson plans. Inference: the measured object is water-playing practice with human mediation, not a precision score. Park and Park (2021) describe sinking or floating as an inquiry-based STEM activity for children: the adult organizes inquiry with real materials; here it is not converted into a curricular scientific-inquiry axis already published nor into STEM/robotics as product, but into evidence of mediated craft on sink-float. Canedo Ibarra and Gómez Galindo (2022) document social interaction in the construction of a floating and sinking precursor model during preschool education: the model is built in conversation and interaction, not “predicted” by AR. Status: precursor-model and social-interaction framework/chapter. Sando et al. (2023) —N = 928 video observations— anchor the role of objects (including water/sand) in deep-level learning mediated by play: water as object-in-play, not as a loose-parts collection. NAEYC (2022) and OECD (2021, 2023) anchor meaningful interactions and subordinated digitalization.

The peripheral good-use contrast is Maraisane, Jita and Jita (2024). They explore notions of floating and sinking in Grade R teachers’ conceptual knowledge in South Africa. Status: empirical finding of TEACHER knowledge. Marked transfer: Grade R / South African context; read toward the adult, not as a trial of AI on the child. Inference: subordinated digital —an adult who deepens human knowledge of sink-float to sustain conversation at the water table— can be coherent with water play; it does not authorize declaring that a pouring-accuracy sensor, a mess camera, a GenAI of experiments, a minutes dashboard or predictive AR constitute the craft. Maraisane et al. are not converted here into a continuing-education article already published: they are cited only as peripheral contrast (human knowledge→adult). Mediation —NAEYC (2022), OECD (2021); Lange (2022); Counsell (2024); Canedo Ibarra and Gómez Galindo (2022)— is adult co-presence that offers real water, tolerates mess, names properties, sustains conversation, observes child agency without expected product and expands without imposing a digital script that closes uncertainty. Inference: pedagogical observation of water play is professional reading of the episode with the liquid; the dashboard counts minutes. A dashboard does not build the precursor model; a child and an adult who converse about sink-float do.

6. Case 3. Water play ≠ outdoor/nature play as axis, curricular scientific inquiry as axis, STEM-robotics, loose parts, free play or motor skills

The first frontier is outdoor/nature play as axis: water play may be outdoor or, centrally here, indoor table; the axis is the liquid (flow, capacity, sink-float, mixtures, mess), not outdoor minutes. Inference: “we have a yard = there is water play” does not sign. A pouring-accuracy sensor is analytics, not outdoor. The second is curricular scientific inquiry: Lange (2022) and Park and Park (2021) show science IN water play; they do not authorize an inquiry-score dashboard nor the inquiry axis already published. The third is STEM/robotics: Counsell (2024) and Park and Park (2021) are craft of the medium, not a STEM score; a GenAI of “STEM water lesson plans” inverts the argument. The fourth is loose parts: Sando et al. (2023) anchor objects (water/sand); here the medium is the liquid, not Nicholson’s open collection. The fifth is free play as axis; a GenAI of experiments closes uncertainty. The sixth is curricular motor skills: pouring involves the body, not motor tracking or off-task splashing alerts. UNESCO (2021), Miao and Holmes (2023) and U.S. Department of Education (2023) require human oversight. Inference: coherent AI use at ages 3–6 remains on the adult side (Maraisane et al., 2024), contrasted with Lange/Counsell and the precursor model (Canedo Ibarra & Gómez Galindo, 2022). It does not enter as scorer, mess classifier, GenAI of experiments, dashboard or predictive AR.

7. Inferential framework: four tests to claim support for water play, 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 the five artifacts constitute support for water play.

7.1. Test of situated practice with real water —flow, capacity, sink-float, mixtures—, child agency and mess as a condition of the medium, not of the pouring-accuracy sensor or the AR layer that guides or predicts. Lange (2022), Counsell (2024), Canedo Ibarra and Gómez Galindo (2022), Amalia and Harahap (2022) and Park and Park (2021) define the craft as observable liquid exploration. Chen (2024), Su and Yang (2022) and Ljungcrantz (2026) map analytics and generation as AI affordances, not as water play. Inference: evidence of support for water play is verified in whether the child explored real water with agency and without an expected product. If the center’s “evidence” is a pouring accuracy score or an AR sink/float prediction, the center has done analytics or digital direction that closes, not support for water play.

7.2. Test of adult co-presence that names properties and sustains conversation, not of GenAI “water experiments” / sink-or-float lesson plans. Lange (2022), Counsell (2024), Canedo Ibarra and Gómez Galindo (2022) and NAEYC (2022) situate conversation, math/science IN the play and developmentally appropriate practice. Miao and Holmes (2023) require pedagogical validation of generative AI. Inference: producing experiments or lesson plans by prompt that close uncertainty does not show that there was co-presence that names flow, capacity or sink-float. A script may exist; it does not sign the water-play episode.

7.3. Test of human mediation —including the Maraisane et al. peripheral contrast—, not of the minutes/inquiry-score dashboard or the CV/camera of mess or off-task splashing alerts. OECD (2021, 2023) require meaningful interactions and subordinated digitalization. Maraisane et al. (2024) anchor teachers’ conceptual knowledge of floating and sinking. Sando et al. (2023) anchor objects-in-play and deep-level learning observed by humans. Inference: “high minutes at the table” or a “mess alert” may raise an administrative threshold without raising co-presence quality or child exploration. Pedagogical observation reads the episode with the liquid; the dashboard counts surveillance. Adult human knowledge is legitimate periphery; the wearable/CV that substitutes exploration or mess tolerance is not.

7.4. Test of category distinction and professional judgment, not of the product catalog. Water play ≠ outdoor/nature play as axis, curricular scientific inquiry as axis, STEM-robotics, loose parts, free play or motor skills. Sando et al. (2023) prevent reducing water objects to loose parts as axis. Park and Park (2021) and Lange (2022) prevent confusing science IN the play with curricular inquiry already published. Counsell (2024) prevents confusing table math with a STEM water score. UNESCO (2021), Miao and Holmes (2023), 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 infant as a pouring-accuracy vector nor as exclusive recipient of GenAI experiments. Support for water play is not fulfilled by better scoring “precision” or surveilling “mess.” It is fulfilled by practicing real water, child agency to explore without an expected product, adult co-presence that names properties and sustains conversation, and tolerance of mess.

The framework admits digital subordinated to adult knowledge/preparation and validated human mediation (Maraisane et al., 2024; Chen, 2024; Su & Yang, 2022; Ljungcrantz, 2026). It rejects declaring support by the five artifacts (Miao & Holmes, 2023; Nikolopoulou, 2025; UNESCO, 2021). The four tests are read together.

8. Discussion

Three tensions organize the discussion. The first is between displaying the five artifacts and exercising support for water play. Lange, Counsell, Amalia and Harahap (N = 15), Park and Park, Canedo Ibarra and Gómez Galindo, Sando et al. (N = 928) and Maraisane et al. sustain the craft or its frontier; Chen (2024), Su and Yang (2022) and Ljungcrantz (2026) map AI without equivalence to situated water play. It is not a finding that the artifacts produce the craft.

The second is between automated assessment of pouring accuracy or mess and pedagogy of the liquid medium. Chen (2024), Su and Yang (2022) and Ljungcrantz (2026) objectify analytics; they do not report verified scoring or mess-alert trials in preschool ages 3–6 —metrics are not invented—. Inference: declaring support by a pouring score or a mess alert is inverted pedagogy.

The third is between GenAI experiments / predictive AR and integral craft. Miao and Holmes (2023) and Nikolopoulou (2025) set limits; Lange (2022), Counsell (2024) and Canedo Ibarra and Gómez Galindo (2022) set science/math IN the play and a social precursor model. Inference: selling prompt experiments or AR sink/float as “water play with AI” confuses product with liquid exploration. Amalia and Harahap (2022) confirm N = 15 practice without authorizing a score; UNESCO (2021), Miao and Holmes (2023) and U.S. Department of Education (2023) subordinate AI to professional judgment.

The four tests in section 7 read these tensions. The empirical contrast —Lange/Counsell; N = 15; Park and Park; precursor model; N = 928; Maraisane et al.— defines the floor the artifacts do not reach alone. Inference: support erodes when they are treated as if they were the practice. Subordinated digital is legitimate toward the adult (Maraisane et al., 2024); inverting the sequence is not (NAEYC, 2022; OECD, 2021, 2023; Lange, 2022; Counsell, 2024).

9. Limits

This review is narrative. It does not apply its own PRISMA nor estimate primary combined effects. Amalia and Harahap (2022) measure N = 15 in one PAUD center: sample limit and marked partial transfer. Lange (2022) and Counsell (2024) provide NSTA craft frameworks, not sensor trials. Park and Park (2021) describe mediated sink/float activity; Canedo Ibarra and Gómez Galindo (2022), a precursor-model and social-interaction chapter; Sando et al. (2023), N = 928 objects-in-play observations (water/sand), not a loose-parts axis; Maraisane et al. (2024), Grade R teacher knowledge as peripheral contrast, not a continuing-education axis. Chen (2024), Su and Yang (2022), Su and Zhong (2022), Ljungcrantz (2026) and Nikolopoulou (2025) map AI in ECE, not pouring-accuracy scoring or mess alerts. No verified trials of pouring-accuracy sensors, CV mess/off-task splashing, GenAI experiments/lesson plans, minutes/inquiry-score dashboards or AR guide/sink-float prediction in CENDI ages 3–6 were located; they are discussed as a category ceiling. NAEYC, UNESCO and OECD are framework sources. Section 7 inferences are category hypotheses, not implementation evidence.

10. Conclusions

A sensor or wearable that “scores pouring accuracy” or a “pouring accuracy score” / “STEM water score” for the child, a computer-vision / water-table camera system that labels “mess” or “off-task splashing” or triggers “mess / off-task splashing” alerts, a chatbot/GenAI that generates “water experiments” or “sink-or-float plans” or “water play lesson plans” by prompt, a dashboard of “minutes at the water table” / “inquiry score” for administrative surveillance, or an AR layer that “guides” pouring or predicts sink/float do not constitute support for water play in an early childhood education center. Chen (2024), Su and Yang (2022) and Ljungcrantz (2026) confirm AI affordances without equivalence to situated water play. Nikolopoulou (2025) and Miao and Holmes (2023) set GenAI limits. Maraisane et al. (2024) require reading Grade R teacher knowledge as peripheral contrast —adult ≠ sensor, mess CV, GenAI, dashboard or AR—. Lange (2022) and Counsell (2024) set science and math IN water play. When there is support, there is situated practice: water play is science (Lange, 2022); math at the table (Counsell, 2024); N = 15 Amalia and Harahap (2022) with marked limit; mediated sink/float (Park & Park, 2021); precursor model and social interaction (Canedo Ibarra & Gómez Galindo, 2022); objects-in-play N = 928 (Sando et al., 2023); DAP and interactions (NAEYC, 2022; OECD, 2021, 2023). Water play is distinguished from outdoor/nature play as axis, curricular scientific inquiry as axis, STEM-robotics, loose parts, free play and motor skills. The guidance documents require human oversight (UNESCO, 2021; Miao & Holmes, 2023; U.S. Department of Education, 2023; Su & Zhong, 2022).

Where sources do not measure a kindergarten, this article does not affirm it. Where they measure AI mappings, GenAI or Grade R teacher knowledge, it does not translate them into pedagogical water-play support via score or mess alert. Accompanying three-to-six-year-olds in water play is to exercise real water (flow, capacity, sink-float, mixtures), child agency to explore without an expected product, adult co-presence that names properties and sustains conversation, and tolerance of mess as a condition of the medium. The rest is a pouring-accuracy sensor, a mess classifier, an experiment generator that closes uncertainty, a minutes dashboard and an AR layer that guides or predicts. It is not support for water play in early childhood education, and it must not be presented as what it is not.

Editorial Laboratory of NEXTECH.IA / Ingeniero Mitre.

References

  1. Amalia, S., y Harahap, J. Y. (2022). Water playing method to improve science in Group B children at PAUD Khairul Ummah. International Journal of Educational Research Excellence, 1(2), 170–173. https://doi.org/10.55299/ijere.v1i2.206
  2. Canedo Ibarra, S. P., y Gómez Galindo, A. A. (2022). Social interaction in the construction of a floating and sinking precursor model during preschool education. En Precursor models for teaching and learning science during early childhood (pp. 53–73). Springer. https://doi.org/10.1007/978-3-031-08158-3_4
  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. Counsell, S. L. (2024). Water play: Where’s the math? Science and Children, 61(3), 11–12. https://doi.org/10.1080/00368148.2024.2340798
  5. Lange, A. (2022). The early years: Water play is science. Science and Children, 59(4), 10–11. https://doi.org/10.1080/00368148.2022.12291760
  6. Ljungcrantz, L. (2026). The interaction of AI and early childhood education. A state-of-the-art review 2020–2024. Early Childhood Education Journal, 54(5), 3565–3581. https://doi.org/10.1007/s10643-025-02079-3
  7. Maraisane, M. J. L., Jita, L. C., y Jita, T. (2024). The notions of floating and sinking: Exploring the conceptual knowledge of Grade R teachers. South African Journal of Childhood Education, 14(1), Article 1407. https://doi.org/10.4102/sajce.v14i1.1407
  8. Miao, F., y Holmes, W. (2023). Guidance for generative AI in education and research. UNESCO. https://doi.org/10.54675/EWZM9535
  9. 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
  10. Nikolopoulou, K. (2025). Child-centered integration of generative AI in early learning: Balancing promises and challenges. AI, Brain and Child, 1(1). https://doi.org/10.1007/s44436-025-00023-1
  11. OECD. (2021). Starting Strong VI: Supporting meaningful interactions in early childhood education and care. OECD Publishing. https://doi.org/10.1787/f47a06ae-en
  12. OECD. (2023). Empowering young children in the digital age (Starting Strong). OECD Publishing. https://doi.org/10.1787/50967622-en
  13. Park, D.-Y., y Park, M.-H. (2021). Sinking or floating: An inquiry-based STEM activity for children. Childhood Education, 97(6), 74–80. https://doi.org/10.1080/00094056.2021.2000310
  14. Sando, O. J., Sandseter, E. B. H., y Brussoni, M. (2023). The role of play and objects in children’s deep-level learning in early childhood education. Education Sciences, 13(7), 701. https://doi.org/10.3390/educsci13070701
  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. Su, J., y Zhong, Y. (2022). Artificial Intelligence (AI) in early childhood education: Curriculum design and future directions. Computers and Education: Artificial Intelligence, 3, 100072. https://doi.org/10.1016/j.caeai.2022.100072
  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. UNESCO. (2021). Recommendation on the ethics of artificial intelligence. UNESCO. https://unesdoc.unesco.org/ark:/48223/pf0000381137