1. Introduction and problem

In the 3-to-6-year span —kindergarten, preschool, CENDI, infant school— a package of artifacts has been installed that claim to count as support for prosocial behaviors (helping / sharing / consoling): a detector/CNN/multimodal system that scores helping count / kindness index / empathy score / sharing compliance as pedagogical assessment; a GenAI of kindness scripts, sharing worksheets or peace corner kindness cards by prompt; a dashboard of helps/hour or kindness compliance; a VR/app as a closed treatment that «trains prosociality»; a robot/app that substitutes scaffolding with a correctness score or automated praise; and a ranking of prosocial developmental level by proximity, synchrony or facial emotion without mediation. The artifacts make it possible to display that the center «already does helping / sharing / consoling with AI». The leap —from score, script, count, VR-treatment or substitute-robot to claiming support— is not authorized by the mappings of AI in ECE nor by pedagogy when there are real needs of the other, shared materials, scaffolding that names the need and models language of offering/helping/consoling, and co-presence that reads prosociality as relational practice, not as a prize for a «score de kindness».

The thesis is restrictive: those artifacts do not constitute support for the development of prosocial behaviors in early childhood. In early childhood education that practice develops with real needs of the other, shared materials, adult co-presence that interprets meaning and process, language of offering/helping/consoling and children's agency to act (NAEYC, 2022; OECD, 2021); not a kindness detector, a GenAI of worksheets, a dashboard of counts, a closed VR-treatment or a robot that substitutes scaffolding. Shoshani (2023) and Shoshani, Nelke y Girtler (2022) contribute helping/sharing trends in laboratory tasks after VR or a tactile app: a finding of experimental transfer, not of classroom pedagogy that substitutes scaffolding. Nijssen, Müller, Bosse y Paulus (2021) —n = 120; 4–9 years; partial transfer— situate sharing with a robot; Kim, Hwang, Lim, Cho y Lee (2024) measure friendship behaviors with a robot; Torpegaard, Knudsen, Linnet, Skov y Merritt (2022) observe a cardboard robot as a social interface; van Straten, Peter, Kühne y Barco (2022) —7–10 years; marked transfer— examine the child–robot relationship. None of these equals adult scaffolding of helping/sharing/consoling among peers. Gungordu, Hernandez-Reif, Walker y Wind (2025), Christner y Paulus (2022), Shi, Zhang y Zhu (2024), Chang y Kim (2025, partial transfer because it includes 7-year-olds) and Takamatsu (2022) anchor predictors, normative sharing, ToM, helping framing and group empathy. Arifiyanti, Dwiningrum y Syamsudin (2025) contrast real teaching strategies. That authorizes asking what was measured: helping count, kindness index, trends after VR/app or sharing with a robot —not the practice when a child helps a peer in difficulty and an adult names the need and sustains agency to act.

The problem is aggravated by six category confusions. First: it is not conflict negotiation / turn-taking as the axis —there may be a dispute IN a materials episode; the axis is helping, sharing and consoling—. Second: it is not generic SEL —there may be regulation IN consoling; the axis is not a SEC auto-score—. Third: it is not child participation/voice as the axis. Fourth: it is not CLASS —Longobardi, Lin y Fabris (2022) show that teacher–child relationship quality mediates prosociality; they do not authorize reducing helping to a CLASS score—. Fifth: it is not joint attention nor clinical emotion monitoring. Sixth: it is not a surveillance friendship-score —Kim et al. (2024) measure friendship behaviors; they do not authorize friendship ranking as the axis—. Mark-making, dialogic reading, literacy, orality, sand/water/playdough/loose parts/outdoor, inclusion, well-being, formative assessment, documentation, STEM, motor skills, creativity as axis and free/symbolic play are not recycled. There is an economy of coordination: the artifacts fit on a slide; the episode of a child who consoles or helps and an adult who names the other's need does not. NAEYC (2022) and OECD (2021) require developmentally appropriate practice and meaningful interactions, with digitization subordinated. Contributions: separating craft and artifacts; examining marked cases; and offering four tests of support.

2. State of the art: from the situated practice of helping/sharing/consoling to the artifact that is displayed

It is useful to separate four strata that the market of «AI for prosociality in early childhood education» usually mixes. The first is the construct of prosocial behaviors from 3 to 6 years as situated relational practice: acts of helping a peer in difficulty, sharing materials, consoling, offering and cooperating in tasks with a real need of the other (NAEYC, 2022; OECD, 2021; Christner y Paulus, 2022; Arifiyanti et al., 2025). The second is the craft that cultivates it —shared materials; an adult who names the other's need, models language of offering/helping/consoling and sustains children's agency to act; co-presence that interprets prosociality as practice, not as a «score de kindness»— (NAEYC, 2022; OECD, 2021; Arifiyanti et al., 2025; Longobardi et al., 2022, peripheral on the teacher–child relationship). The third is the evidence of AI affordances in ECE and of VR/tablet systems, HRI of sharing/caring/friendship, empathy predictors and XR reviews —without equating them to situated helping/sharing/consoling pedagogy— (Chen, 2024; Ljungcrantz, 2026; Nikolopoulou, 2025; Shoshani, 2023; Shoshani et al., 2022; Nijssen et al., 2021; Kim et al., 2024; Torpegaard et al., 2022; van Straten et al., 2022; Yousefi et al., 2024). The fourth is the framework of rights and of developmentally appropriate practice, which treats the young child as a subject who helps, shares and consoles with agency, not as a helping-count vector nor as the exclusive addressee of GenAI kindness cards, closed VR-treatments or robots that praise «correctness» (UNESCO, 2021; Miao y Holmes, 2023).

NAEYC (2022) and OECD (2021) anchor developmentally appropriate practice and meaningful interactions: framework. The verified systems measure objects distinct from the craft —laboratory trends after VR/tablet (Shoshani, 2023; Shoshani et al., 2022), sharing or relationship with a robot with age transfers (Nijssen et al., 2021; Kim et al., 2024; Torpegaard et al., 2022; van Straten et al., 2022), adult/general XR/EVA (Yousefi et al., 2024)—. Inference: a helping count does not name the peer's need nor model «do you want me to help you?»; an adult who sustains agency to act does. Chen (2024) and Ljungcrantz (2026) map AI in ECE without equating it to situated prosociality. Nikolopoulou (2025) and Miao y Holmes (2023) set GenAI caution. UNESCO (2021) requires human supervision.

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 artifacts, but to articulate an argument of pedagogical category with verified sources. Inclusion criteria: (a) 2021–2026, with marked transfer when the sample does not equal 3–6 in a classroom of helping/sharing/consoling, is HRI of 7–10 years or 4–9, XR of adults/general, or the object is friendship/child–robot relationship without equating to scaffolding of helping/sharing/consoling; (b) prosocial helping, sharing, consoling, cooperation-helping, VR/tablet prosocial, robotic mediation, normative understanding of sharing, helping framing, group empathy, teaching strategies, or AI in ECE with artifact/craft relevance; (c) kindergarten, preschool, CENDI or 3–6, or explicit transfer; (d) peer-reviewed journal, DOI or NAEYC/UNESCO/OECD report; (e) verifiable DOI or publisher page. Excluded as central object were axes already used in this series —conflict negotiation/turn-taking, generic SEL, participation/voice, CLASS, joint attention, free play, symbolic play, mark-making, dialogic reading, literacy, orality, sand/water/playdough/loose parts/outdoor/risky/block/spatial, inclusion, well-being, formative assessment, documentation, STEM, environmental, motor skills, creativity as axis and inquiry—. Longobardi et al. (2022) only as a peripheral contrast of the teacher–child relationship. Yousefi et al. (2024) = XR/EVA, adult/general transfer. van Straten et al. (2022) = 7–10 years. Nijssen et al. (2021) = 4–9, partial transfer. Chang y Kim (2025) = 5–7, partial transfer. Kim et al. (2024) are read for friendship CRI, not as a surveillance friendship-score. Torpegaard et al. (2022) are read for HRI, not as free play. Gungordu et al. (2025) are read for predictors, not as a creativity axis.

The search was executed on 8 September 2026 (slot 09:02 America/Mexico_City, weekday routine) on DOI pages, Crossref, Springer, Elsevier, Wiley, Taylor & Francis, MDPI, Frontiers, JAIR, OECD iLibrary, UNESDOC, NAEYC and publisher sites. Each source was verified against fuentes.md and crossref_check.json. Empirical finding, framework and pedagogical inference were distinguished. No N, d, r, AUC or DOI was invented: when an artifact lacks a verified trial as a display package in CENDI 3–6, it is discussed as a category ceiling supported by mappings of AI in ECE (Chen, 2024; Ljungcrantz, 2026) and by the verified VR/tablet/HRI/XR systems (Shoshani, 2023; Shoshani et al., 2022; Nijssen et al., 2021; Kim et al., 2024; Torpegaard et al., 2022; van Straten et al., 2022; Yousefi et al., 2024). Twenty-one sources from the verified corpus were used.

4. Case 1. CV/multimodal detector of prosocial acts / prosocial rate / helping count / kindness index / empathy score / sharing compliance, a GenAI of kindness scripts/worksheets/lesson packs, a dashboard of helps/hour, a VR/app as a closed treatment of «training prosociality», a robot/app that substitutes scaffolding with a score or praise, or proximity/synchrony→prosocial developmental ranking do not constitute support for prosocial behaviors

Chen (2024) and Ljungcrantz (2026) saturate the ceiling of artifacts when AI in ECE is presented as if it were support for helping/sharing/consoling. Chen (2024) maps affordances —tutoring, analytics, content generation, agents—, not that a helping count cultivates language of offering. Ljungcrantz (2026) reviews AI–ECE interaction 2020–2024: state of the art, not situated prosociality. Inference: «high helping count = there was support» is a ceiling of analytics. A prosocial rate score can coexist with an absence of scaffolding. The grammar of display inverts the sequence: first the detector is activated; then it is declared that «there is already prosociality with AI».

Nikolopoulou (2025) and Miao y Holmes (2023) name the risk of a GenAI of kindness scripts / sharing worksheets / peace corner kindness cards by prompt that closes the episode: a framework of caution, not a trial versus an adult who models «shall I offer them one?» (NAEYC, 2022; OECD, 2021). Inference: producing a worksheet by prompt may be subordinate preparation; support begins with a real need of the peer and an adult who names without reducing to correctness. UNESCO (2021) requires human supervision. When the script speaks for the adult and the score speaks for the child, the threshold is not met even if the dashboard looks «green».

Shoshani (2023) and Shoshani et al. (2022) are the empirical anchor of the VR/app artifact and must be read with category precision: more helping/sharing in laboratory tasks after an experimental session does not equal «the VR/app is a closed treatment that trains prosociality and can substitute adult mediation and co-presence among peers in kindergarten». Yousefi et al. (2024) —fifteen XR/EVA studies; adult/general transfer— find potential of virtual agents and call for more design research; they do not authorize importing XR as a 3–6 classroom treatment. Inference: selling «we train kindness with VR» is inverted pedagogy. Nijssen et al. (2021), Kim et al. (2024), Torpegaard et al. (2022) and van Straten et al. (2022) saturate the ceiling of the robot/app: HRI of sharing, friendship, social interface or child–robot relationship, with age transfers where they apply, not the craft of sharing materials between two children with an adult who names need. Restrictive inference: detector ≠ adult who names; GenAI of kindness cards ≠ open episode; dashboard ≠ observation; closed VR-treatment ≠ co-presence among peers; praise/correctness robot ≠ agency to help; ranking by proximity ≠ mediation. The artifacts lack, in the verified corpus, pedagogical-equivalence trials with the craft in CENDI 3–6; no d, r or AUC of «support» is invented. They share a grammar of substitution: the score speaks for the act; the GenAI closes; the dashboard substitutes observation; the VR is sold as treatment; the robot erases scaffolding.

5. Case 2. What the kindergarten does do when there is support for helping/sharing/consoling: situated relational craft and what Shoshani VR, Shoshani tablet, Nijssen, Kim, Torpegaard, van Straten, Gungordu, Christner, Shi, Chang, Takamatsu and Arifiyanti measure —with Longobardi peripheral and Yousefi with XR transfer

The floor of the craft is not a score: it is real needs of the other; shared materials; acts of helping a peer in difficulty, sharing, consoling, offering and cooperating in situated tasks; adult scaffolding that names the need, models language of offering/helping/consoling and sustains children's agency to act; co-presence that interprets prosociality as relational practice, not as a «score de kindness» (NAEYC, 2022; OECD, 2021; Arifiyanti et al., 2025). Inference: there is support when that practice is protected; not when a helping count, a GenAI of worksheets or a VR-treatment is displayed. The craft is recognized in the episode: a peer offers, consoles or helps; the adult names («it fell; shall we help them?») and sustains that they act; the robot or the VR, if they appear, are subordinate means, not substitutes.

Shoshani (2023) saturates the VR anchor: two experiments with 4–6-year-olds (N = 166 helping toward the experimenter; N = 173 sharing with peers in a dictator game); more helping and sharing after prosocial VR play than after violent, positive-affect or neutral conditions; mediation by positive affect, not by competence. Status: empirical finding of laboratory trends. Inference: it does not equal classroom pedagogy nor authorize VR as a closed treatment that substitutes scaffolding and co-presence. Shoshani et al. (2022) —3–6 years; N = 186 and N = 189— saturate the tablet anchor: helping toward the experimenter and behaviors toward real and hypothetical peers after a prosocial tactile app; moderation of low ToM and high empathy. Inference: app as experimental platform ≠ GenAI sharing worksheets nor a dashboard of minutes of sharing. Marked distinction: they are cited for the VR/app artifact and for what they do not equal as craft; they are not converted into a SEL axis.

Nijssen et al. (2021) —n = 120; 4–9 years; partial transfer— show more sharing with the robot to which feelings are attributed, with no effect of visual anthropomorphism. Inference: HRI of sharing with a robot «that feels» ≠ adult scaffolding of sharing among peers; it does not authorize a substitute-robot nor praise of «correct sharing». Kim et al. (2024) —ten 5–6-year-old children; four triadic sessions in after-school in the USA— observe liking, togetherness, parity, agreement and co-construction under robotic mediation. Marked inference: friendship behaviors ≠ helping/sharing/consoling; it does not authorize substituting the adult nor converting friendship into a surveillance score. Torpegaard et al. (2022) —214 children; 164 observations at a convention playground— describe a cardboard robot treated as a social agent and as a material object. Inference: social interface ≠ pedagogy of consoling; it is cited for HRI, not as a free-play axis. van Straten et al. (2022) —293 children aged 7–10 years; museum; Nao; marked transfer— find that asking increases trust and belief in cognitive perspective-taking, while personal self-disclosure decreases the perception of affective perspective-taking. Inference: child–robot relationship at 7–10 ≠ helping/sharing/consoling in kindergarten 3–6.

Gungordu et al. (2025) —122 dyads; 3–5 years; picture tasks and parental questionnaires— find that age, creativity and cognitive empathy play roles in the decision and in prosocial creativity. Marked inference: predictors ≠ pedagogical kindness index; creativity is not recycled as an axis. Christner y Paulus (2022) —N = 90; 4–6 years— organize normative expressions of sharing into representation and enforcement; moral self-concept was related to actual sharing. Inference: normative understanding is a pedagogical base, not a sharing compliance dashboard. Shi et al. (2024) —experiment 1, N = 95; 4–6 years— find that anticipated and consequent emotions influence sharing; ToM influences sharing and is related to allocation equity. Inference: ToM and emotion ≠ auto-score or ranking. Chang y Kim (2025) —90 children aged 5, 6 and 7 years; partial transfer— show that, with low issue relevance, loss frames were more effective at age 5 and gain frames at ages 6–7. Inference: framing may inform the adult language that names need; it does not authorize a GenAI of kindness scripts as a closed recipe. Takamatsu (2022) —N = 50; 4–6 years; Japanese preschool— finds less empathic concern toward the «odd one out» among friends than toward the majority. Inference: empathy is group-based and situated; a uniform empathy score erases whom one consoles; the craft names the need of the excluded child.

Arifiyanti et al. (2025) —49 kindergarten educators in Indonesia; open interviews; thematic analysis— saturate the non-AI pedagogical contrast: teachers report difficulties of sharing and cooperation; they implement direct engagement, collaborative activities, fostering of values, classroom rules and modeling; collaborative-learning frameworks predominate. Status: qualitative finding of teaching strategies. Inference: the craft is modeling, collaboration and direct engagement, not a detector nor a worksheet by prompt. Longobardi et al. (2022) —936 children aged 3–6 years and 60 teachers in Italy; peripheral— show that daytime sleepiness predicts closeness, conflict and dependence in the teacher–child relationship, and that those dimensions predict prosocial behaviors. Peripheral inference: the pathway is relationship quality, not a helping-count; this article is not converted into a CLASS axis nor a well-being/sleep axis. Yousefi et al. (2024) —fifteen studies; adult/general XR transfer— confirm potential and limits of EVA in XR for triggering prosociality, without equating to a 3–6 classroom. Coherent mediation (NAEYC, 2022; OECD, 2021): real need of the other, observation of process, language of offering/helping/consoling and agency to act without a correctness script.

Taken together: there are systems that measure trends after VR or tablet, sharing or relationship with a robot, empathy predictors, normative sharing and ToM, helping framing, group empathy, teaching strategies, teacher–child relationship or XR/EVA with transfer. None —by itself— signs a real need of the other, language of offering/helping/consoling, co-presence that interprets process and children's agency to act in the classroom. The technical evidence is real in its domain; «score / VR-treatment / robot = support for prosocial behaviors» is an illegitimate category inference.

6. Case 3. Prosocial helping/sharing/consoling ≠ conflict negotiation/turn-taking, SEL as axis, participation-voice, CLASS, joint attention, clinical emotion monitoring or a surveillance friendship-score

Six borders protect the axis. First: peer conflict negotiation / turn-taking —there may be a dispute over a material IN a sharing episode; the axis is helping, offering, sharing and consoling, not negotiating a turn nor repairing a conflict with «say you're sorry»—. Inference: «we work on turn-taking = there is prosociality» does not sign; nor is a sharing compliance score negotiation or pedagogical helping. This article does not recycle yesterday's axis. Second: generic SEL —Gungordu et al. (2025) predict via cognitive empathy; Shi et al. (2024) link ToM and emotion to sharing; there may be regulation IN consoling; the axis is the relational practice of offering help, not a SEC auto-score nor a kindness index as SEL assessment—. Third: child participation/voice —there is speech of offering and agency to act; this article is not converted into participation/voice already published—. Fourth: CLASS / adult–child interactions as axis —Longobardi et al. (2022) show mediation of teacher–child relational quality; they anchor that the relationship matters; they do not authorize reducing helping/sharing/consoling to an already published adult–child interaction score—. Fifth: joint attention —the co-presence that names the other's need is not joint attention as axis; the article is not rewritten as joint attention—. Sixth: clinical emotion monitoring and a surveillance friendship-score —Takamatsu (2022) situates group empathy; Kim et al. (2024) measure friendship behaviors; neither a ranking of facial empathy nor a surveillance friendship score equals consoling a peer in difficulty—. UNESCO (2021), Miao y Holmes (2023) and Nikolopoulou (2025) require human supervision. Inference: coherent AI at 3–6 remains on the adult's side (subordinate preparation; Arifiyanti et al., 2025; Longobardi et al., 2022), not as a detector, GenAI of kindness cards, dashboard of helps/hour, closed VR-treatment, praise robot or ranking without mediation.

These borders protect the axis. Helping/sharing/consoling may touch, in a lateral way, regulation, speech, interaction quality, friendship or emotion; it is not allowed to be reduced to conflict negotiation, SEL, voice, CLASS, joint attention, clinical monitoring or a friendship-score. A center that «works on SEL», «negotiates turns» or «scores friendship» has not demonstrated by that label alone the craft of helping, sharing and consoling. The distinction is one of pedagogical category, not a denial of lateral overlaps. Gungordu et al. (2025) are read for predictors, not as a creativity axis; Torpegaard et al. (2022) are read for HRI, not as a free-play axis.

7. Inferential framework: four tests for claiming that there is support for prosocial behaviors (helping/sharing/consoling), not an artifact

The framework that follows is a 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 the artifacts constitute support for the development of prosocial behaviors in early childhood.

7.1. Test of situated practice (real need of the other; shared materials; acts of helping, sharing and consoling with agency), not of the CV/multimodal detector nor of the correctness or praise robot/app. NAEYC (2022), OECD (2021) and the VR/HRI/teacher cases (Shoshani, 2023; Shoshani et al., 2022; Nijssen et al., 2021; Kim et al., 2024; Torpegaard et al., 2022; van Straten et al., 2022; Arifiyanti et al., 2025) define the craft/artifact contrast. Chen (2024), Ljungcrantz (2026) and Nikolopoulou (2025) map analytics as an affordance, not as situated helping. Inference: if the «evidence» is helping count, kindness index, sharing compliance or automated praise, the center did analytics or behavioral HRI, not support.

7.2. Test of adult co-presence that names the other's need, models language of offering/helping/consoling and sustains that the children act —not only «be kind»—, not of the GenAI of kindness scripts/worksheets/lesson packs nor of the VR/app as a closed treatment. NAEYC (2022), OECD (2021), Arifiyanti et al. (2025) and Chang y Kim (2025, helping framing, not a recipe) situate practice and modeling. Miao y Holmes (2023) and Nikolopoulou (2025) require GenAI mediation. Shoshani (2023), Shoshani et al. (2022) and Yousefi et al. (2024, transfer) prevent selling VR/app/XR as a treatment that substitutes scaffolding. Inference: a script by prompt or a laboratory VR session does not sign the classroom episode.

7.3. Test of human mediation and of the interpretation of prosociality as relational practice, not of the dashboard of helps/hour / kindness compliance nor of ranking by proximity/synchrony/facial emotion. OECD (2021) requires meaningful interactions. Christner y Paulus (2022), Shi et al. (2024), Takamatsu (2022) and Gungordu et al. (2025) anchor normative sharing, ToM and group empathy, not surveillance of counts. Longobardi et al. (2022) anchor the teacher–child relationship without converting it into a ranking. Inference: high helps/hour do not by themselves raise scaffolding. Subordinate preparation of the adult is legitimate; the scorer that converts the child into a helping-count vector is not.

7.4. Test of category distinction and of professional judgment, not of the product catalogue. Helping/sharing/consoling ≠ conflict negotiation/turn-taking, SEL, participation-voice, CLASS, joint attention, clinical emotion monitoring or a surveillance friendship-score. Kim et al. (2024), Longobardi et al. (2022), Takamatsu (2022) and Gungordu et al. (2025) prevent those confusions. UNESCO (2021), Miao y Holmes (2023), NAEYC (2022) and OECD (2021) require human supervision. Inference: support is fulfilled with a real need of the other, co-presence that names/models and agency to act —not by scoring kindness nor by substituting scaffolding with a robot.

The framework admits digital subordinate to the adult's preparation/documentation (Arifiyanti et al., 2025; Chang y Kim, 2025, read by the adult; Shoshani, 2023, and Shoshani et al., 2022, as experimental evidence, not as classroom treatment). It rejects declaring support by the artifacts (Miao y Holmes, 2023; Nikolopoulou, 2025; UNESCO, 2021). The four tests are read together.

8. Discussion

Three tensions. First: displaying the artifacts versus exercising support. Shoshani (2023), Shoshani et al. (2022), Nijssen et al. (2021, 4–9 transfer), Kim et al. (2024), Torpegaard et al. (2022), van Straten et al. (2022, 7–10 transfer), Gungordu et al. (2025), Christner y Paulus (2022), Shi et al. (2024), Chang y Kim (2025, partial transfer), Takamatsu (2022), Arifiyanti et al. (2025), Longobardi et al. (2022, peripheral) and Yousefi et al. (2024, XR transfer) sustain the contrast; Chen (2024) and Ljungcrantz (2026) map AI without equivalence to situated helping/sharing/consoling. Second: automated assessment or VR-treatment versus relational pedagogy —declaring support by helping count, kindness index or by a laboratory VR session is inverted pedagogy—. Third: GenAI of kindness cards / praise robot / ranking by proximity versus craft (Miao y Holmes, 2023; Nikolopoulou, 2025; NAEYC, 2022; OECD, 2021): selling VR that «trains prosociality» or robotic friendship mediation without human scaffolding confuses product with practice. Arifiyanti et al. (2025) confirm teacher modeling without authorizing dashboards; Longobardi et al. (2022) confirm that the relationship mediates prosociality without authorizing ranking; UNESCO (2021) subordinates AI to professional judgment.

The four tests of section 7 read these tensions. The empirical contrast defines the floor that the artifacts do not reach on their own. Inference: support is eroded when they are treated as if they were the practice. Subordinate digital is legitimate toward the adult (preparation of language of offering; documentation; Arifiyanti et al., 2025 contrast); inverting the sequence is not (NAEYC, 2022; OECD, 2021). If the center shows real needs of the other, acts of helping/sharing/consoling, an adult who named and modeled, and children who acted with agency, it may speak of support; if it only shows detectors, GenAI scripts, helps/hour, VR-treatments, praise robots or rankings without mediation, it speaks of artifacts. Chen (2024) and Ljungcrantz (2026): AI affordances in ECE ≠ situated helping/sharing/consoling.

9. Limits

This review is narrative. It does not apply its own PRISMA nor estimate combined effects. Shoshani (2023) and Shoshani et al. (2022) are read as experimental transfer, not as classroom craft nor as authorization of a closed treatment; no d or r is invented. Nijssen et al. (2021) delimit sharing with a robot at 4–9 years: partial transfer. Kim et al. (2024) measure ten children: small sample; friendship ≠ helping/consoling. Torpegaard et al. (2022) observe a convention playground, not a CENDI classroom. van Straten et al. (2022) are 7–10 years in a museum: marked transfer. Yousefi et al. (2024) review XR/EVA with a non-3–6 predominance: transfer. Gungordu et al. (2025) measure predictors with picture tasks: creativity is not recycled. Christner y Paulus (2022) and Shi et al. (2024) delimit normative sharing and ToM in tasks. Chang y Kim (2025) include 7-year-olds and measure perceived intention: partial transfer. Takamatsu (2022) measures N = 50. Arifiyanti et al. (2025) are qualitative with 49 teachers. Longobardi et al. (2022) are peripheral. Chen (2024), Ljungcrantz (2026) and Nikolopoulou (2025) map AI in ECE, not dashboards nor a GenAI of kindness cards as a display package in a Latin American CENDI. No verified trials of the artifacts as a single package of «prosociality with AI» in a 3–6 classroom were located; they are discussed as a category ceiling. NAEYC, UNESCO and OECD are framework. The inferences of section 7 are category hypotheses, not implementation evidence.

10. Conclusions

The artifacts —a CV/multimodal detector or scores of helping count / kindness index / empathy score / sharing compliance; a GenAI of kindness scripts/worksheets/lesson packs; a dashboard of helps/hour; a VR/app as a closed treatment; a robot/app that substitutes scaffolding with a score or praise; a prosocial developmental ranking without mediation— do not constitute support for the development of prosocial behaviors (helping / sharing / consoling) in early childhood education. Chen (2024) and Ljungcrantz (2026) confirm affordances without equivalence to situated helping/sharing/consoling. Nikolopoulou (2025) and Miao y Holmes (2023) set GenAI limits. Shoshani (2023), Shoshani et al. (2022), Nijssen et al. (2021), Kim et al. (2024), Torpegaard et al. (2022), van Straten et al. (2022), Gungordu et al. (2025), Christner y Paulus (2022), Shi et al. (2024), Chang y Kim (2025), Takamatsu (2022), Arifiyanti et al. (2025), Longobardi et al. (2022) and Yousefi et al. (2024) require reading VR/tablet/HRI/predictors/normative sharing/framing/group empathy/teaching strategies/relationship/XR by what they measure —with transfers— and by what they do not equal as situated pedagogy. When there is support there are real needs of the other, shared materials, co-presence that names/models and children's agency to act (NAEYC, 2022; OECD, 2021). Helping/sharing/consoling is distinguished from conflict negotiation/turn-taking, SEL, participation-voice, CLASS, joint attention, clinical emotion monitoring and a surveillance friendship-score. The guidelines require human supervision (UNESCO, 2021; Miao y Holmes, 2023).

Where the sources do not measure a kindergarten, this article does not claim it. Where they measure mappings of AI, trends after VR/tablet, sharing or relationship with a robot, predictors, normative sharing, framing, group empathy, teaching strategies, relational quality or XR with transfer, it does not translate them into support via helping-count, kindness-index, worksheet by prompt, closed VR-treatment or substitute-robot. Accompanying girls and boys from three to six years in prosocial behaviors is to exercise real needs of the other, shared materials, acts of helping a peer in difficulty, sharing and consoling, scaffolding that names the need and models language of offering/helping/consoling, and co-presence that interprets prosociality as situated relational practice, not as a «score de kindness». The rest is a detector, a GenAI of worksheets, a dashboard of counts, a closed VR-treatment, a praise robot and a ranking. It is not support for prosocial behaviors in early childhood education, and it must not be presented as what it is not.

Laboratorio Editorial de NEXTECH.IA / Ingeniero Mitre.

References

  1. Arifiyanti, N., Dwiningrum, S. I. A., y Syamsudin, A. (2025). Strategies of Indonesian kindergarten teachers for developing children’s prosocial behaviour. Journal of Education Culture and Society, 2025(2), 57–75. https://doi.org/10.15503/jecs2025.2.57.75
  2. Chang, H., y Kim, H. (2025). Message framing effects on helping behavior as a function of age in early childhood and school-age children. Behavioral Sciences, 15(6), Article 758. https://doi.org/10.3390/bs15060758
  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. Christner, N., y Paulus, M. (2022). Varieties of normative understanding and their relation to sharing behavior in preschool children. Journal of Experimental Child Psychology, 222, Article 105498. https://doi.org/10.1016/j.jecp.2022.105498
  5. Gungordu, N., Hernandez-Reif, M., Walker, D. I., y Wind, S. A. (2025). Empathy and creativity as foundations and predictors of how prosocial behavior develops in preschool age children. International Journal of Child Care and Education Policy, 19, Article 6. https://doi.org/10.1186/s40723-025-00147-0
  6. Kim, Y., Hwang, J., Lim, S., Cho, M.-H., y Lee, S. (2024). Child–robot interaction: designing robot mediation to facilitate friendship behaviors. Interactive Learning Environments, 32(8), 4169–4182. https://doi.org/10.1080/10494820.2023.2194936
  7. 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
  8. Longobardi, C., Lin, S., y Fabris, M. A. (2022). Daytime sleepiness and prosocial behaviors in kindergarten: The mediating role of student-teacher relationships quality. Frontiers in Education, 7, Article 710557. https://doi.org/10.3389/feduc.2022.710557
  9. Miao, F., y Holmes, W. (2023). Guidance for generative AI in education and research. UNESCO. https://doi.org/10.54675/EWZM9535
  10. 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
  11. Nijssen, S. R. R., Müller, B. C. N., Bosse, T., y Paulus, M. (2021). You, robot? The role of anthropomorphic emotion attributions in children’s sharing with a robot. International Journal of Child-Computer Interaction, 30, Article 100319. https://doi.org/10.1016/j.ijcci.2021.100319
  12. 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
  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. Shi, Y., Zhang, M., y Zhu, L. (2024). Sharing and allocation in preschool children: The roles of theory of mind, anticipated emotions, and consequential emotions. Behavioral Sciences, 14(10), Article 931. https://doi.org/10.3390/bs14100931
  15. Shoshani, A. (2023). From virtual to prosocial reality: The effects of prosocial virtual reality games on preschool children’s prosocial tendencies in real life environments. Computers in Human Behavior, 139, Article 107546. https://doi.org/10.1016/j.chb.2022.107546
  16. Shoshani, A., Nelke, S., y Girtler, I. (2022). Tablet applications as socializing platforms: The effects of prosocial touch screen applications on young children’s prosocial behavior. Computers in Human Behavior, 127, Article 107077. https://doi.org/10.1016/j.chb.2021.107077
  17. Takamatsu, R. (2022). Empathy and group processes in Japanese preschool children: The odd one out among friends receives less empathic concern than out-groups. Journal of Experimental Child Psychology, 221, Article 105460. https://doi.org/10.1016/j.jecp.2022.105460
  18. Torpegaard, J., Knudsen, L. S., Linnet, M. P., Skov, M. B., y Merritt, T. (2022). Preschool children’s social and playful interactions with a play-facilitating cardboard robot. International Journal of Child-Computer Interaction, 31, Article 100435. https://doi.org/10.1016/j.ijcci.2021.100435
  19. UNESCO. (2021). Recommendation on the ethics of artificial intelligence. UNESCO. https://unesdoc.unesco.org/ark:/48223/pf0000381137
  20. van Straten, C. L., Peter, J., Kühne, R., y Barco, A. (2022). On sharing and caring: Investigating the effects of a robot’s self-disclosure and question-asking on children’s robot perceptions and child-robot relationship formation. Computers in Human Behavior, 129, Article 107135. https://doi.org/10.1016/j.chb.2021.107135
  21. Yousefi, M., Crowe, S. E., Hoermann, S., Sharifi, M., Romera, A., Shahi, A., y Piumsomboon, T. (2024). Advancing prosociality in extended reality: systematic review of the use of embodied virtual agents to trigger prosocial behaviour in extended reality. Frontiers in Virtual Reality, 5, Article 1386460. https://doi.org/10.3389/frvir.2024.1386460