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 cooperative play and peer collaboration: a GenAI of cooperative play scripts, peer collaboration lesson packs, shared-goal activity banks or role-assignment checklists by prompt; a classifier, an ML system or a camera that assigns a cooperation quality score, peer collaboration fidelity, shared-goal compliance or cooperative minutes/hour; a dashboard of cooperative minutes or group task completion rate; a robot presented as a «cooperative peer» that replaces negotiation among peers without contingent adult mediation; and a ranking of collaborative effectiveness by vision/NLP without situated judgment. The artifacts make it possible to display that the center «already does cooperative play with AI». The leap —from script, score, count, substitute-robot or ranking to claiming support— is not authorized by the mappings of AI in early childhood education nor by pedagogy when there are girls and boys who negotiate shared goals, roles, contribution turns and joint resolution, with contingent adult mediation when needed, not a «cooperation quality score».

The thesis is restrictive: those artifacts do not constitute support for cooperative play in early childhood. In early childhood education that practice develops when two or more children work together toward a goal that is not reached alone —they communicate, exchange knowledge, deploy prosocial skills of collaboration and produce something joint—, and an adult observes, groups, assigns roles or guides contingently without hijacking negotiation among peers (Burns, Yu et al., 2024; Burns, Brathwaite et al., 2025; Lin et al., 2024; Evaldsson, 2022); not a GenAI of scripts, a quality score, a dashboard of minutes, a «cooperative peer» robot nor a ranking of collaborative effectiveness. AI may support the adult's preparation, documentation and reflection in a subordinate way (Lee, Lee y Yoon, 2025; Edwards, 2023; Giray, 2026); it must not substitute co-construction among peers nor convert cooperation into a surveillance metric. Burns, Yu et al. (2024) find mixed results in eighteen interventions, with limited attention to educators. Burns, Brathwaite et al. (2025) show that there is no shared definition nor observation of peer collaboration. Lin et al. (2024) evaluate twelve weeks of cooperative constructive play with rural 5–6-year-old children: situated practice, not an activity bank. That authorizes asking what was measured: a cooperation quality score —not the practice when peers negotiate the goal and a professional interprets the episode.

The problem is aggravated by ten category confusions. First: it is not free play —25 August covered that axis; here there is a shared goal and, when needed, adult mediation—. Second: it is not guided play (14 September): the axis is negotiation among peers, not the curricular guidance of the free–guided–directed continuum—. Third: it is not outdoor, loose parts, sand, water, playdough nor risky play. Fourth: it is not symbolic play nor spatial/block play —Lin et al. (2024) use construction and roles as a context of cooperation—. Fifth: it is not co-regulation (11 September) nor SST (10 September). Sixth: it is not ToM (9 September). Seventh: it is not prosocial helping/sharing (8 September) —Burns, Brathwaite et al. (2025) include prosocial skills as a domain of the definition; the axis is not helping or sharing as an already published article—. Eighth: it is not conflict negotiation / turn-taking (7 September): there may be contribution turns IN a cooperative episode; the axis is not conflict nor the turn—. Ninth: it is not inclusion as exclusive axis —Tan y Perren (2023), Kilincci et al. (2026) and Fedewa, Watkins, Carden et al. (2025) are read for the mediation of collaboration among peers—. Tenth: it is not generic SEL, CLASS, joint attention, mark-making, dialogic reading, literacy, orality, executive functions, numeracy, well-being, formative assessment nor documentation. The artifacts fit on a slide; the episode in which two children agree «we make the truck together» and repair the plan, does not. Contributions: separating craft and artifacts; examining marked cases; and offering four tests of support.

2. State of the art: situated practice vs artifact

It is useful to separate four strata that the market of «AI for cooperative play in early childhood education» usually mixes. The first is the construct from 3 to 6 years as situated relational practice: peer collaboration as a complex skill, without a single definition, observed when children pursue a larger goal, communicate, exchange knowledge and deploy collaborative skills —not when a score declares «there was cooperation»— (Burns, Yu et al., 2024; Burns, Brathwaite et al., 2025; Lin et al., 2024; Evaldsson, 2022). The second is the craft —an adult who observes and guides contingently; groups; assigns roles without closing the negotiation; forms pairs; co-presence that interprets the episode, not only «there were cooperative minutes»— (Fedewa, Watkins, Carden et al., 2025; Fedewa, Watkins, Barber et al., 2024; Tan y Perren, 2023; Kilincci et al., 2026). The third is the evidence of AI affordances, robots, teacher perceptions, postdigital critique and technology during collaboration —without equating them to cooperative pedagogy nor to a robot that substitutes the peer— (Su y Yang, 2022; Su y Zhong, 2022; Lee et al., 2025; Lee, Lee y Yoon, 2025; Edwards, 2023; Giray, 2026; Burns, Hoan et al., 2024). The fourth is the framework of peer culture and of network: the young child chooses with whom to play, attributes wisdom or dominance and negotiates microdramas, and is not a vector of cooperative minutes/hour nor the addressee of a «cooperative peer» agent (Evaldsson, 2022; Bombjaková y Mikuškovičová, 2026; Bier et al., 2026). Inference: a cooperation quality score does not negotiate «you put on the wheels; I the steering wheel»; a peer who co-constructs the goal does. Su y Yang (2022) map AI without equating it to situated collaboration. Lee et al. (2025) propose incorporating robots, not substituting the peer. Giray (2026) establishes that the authentic relationship is not replicated by a chatbot. Lee, Lee y Yoon (2025) find that teachers emphasize human interactions.

The verified systems measure objects distinct from the craft: mixed interventions on five components (Burns, Yu et al., 2024), domains of definition and of observation (Burns, Brathwaite et al., 2025), hardware and software during collaboration (Burns, Hoan et al., 2024), Anji-based CCP (Lin et al., 2024; Li, 2026, framework), microdramas (Evaldsson, 2022), teacher-facilitated PMI (Fedewa, Watkins, Carden et al., 2025; Watkins et al., 2023, transfer), teacher strategies (Tan y Perren, 2023; Kilincci et al., 2026), wisdom and dominance (Bombjaková y Mikuškovičová, 2026) and homophily at four years (Bier et al., 2026). Inference: a peer collaboration fidelity does not repair the plan that the episode requires; an adult who interprets and peers who renegotiate the role do. Skehill y O’Donovan (2025) anchor transfer: collaboration among centers in the transition, not among peers in play. Edwards (2023) anchors the postdigital framework: AI is not a neutral add-on to the kindergarten.

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 cooperative-play classroom, is primary or school/after-school, is institutional transition, is HRI/robot as framework, is an AI curriculum, or is a measure/network without equating to automatic pedagogy; (b) cooperative play, peer collaboration, peer-mediated social play, teacher strategies of peer interaction, peer culture, or AI in ECE / robots / GenAI / postdigital / technology during collaboration; (c) kindergarten, preschool, CENDI or 3–6, or explicit transfer; (d) peer-reviewed journal with DOI. Excluded as central object were axes already used in this series —free play, guided play, outdoor/loose parts, symbolic, spatial/block, co-regulation, SST, ToM, prosocial helping/sharing, peer-conflict/turn-taking, mark-making, dialogic reading, SEL, CLASS, joint attention, inclusion as exclusive axis, well-being, formative assessment, documentation, FE and numeracy—. Tan y Perren (2023), Kilincci et al. (2026) and Fedewa, Watkins, Carden et al. (2025) are read for mediation of collaboration, not as inclusion. Lin et al. (2024) are read for cooperation, not as block play. Li (2026) = Anji framework, not an AI trial. Skehill y O’Donovan (2025) = collaboration among centers: construct transfer. Watkins et al. (2023) = upper primary: stage transfer. Evaldsson (2022) = school/after-school: setting transfer. Burns, Yu et al. (2024) and Burns, Brathwaite et al. (2025) cover 0–6: age transfer toward 0–3.

The search was executed on 15 September 2026 (slot 09:02 America/Mexico_City, weekday routine) on DOI pages, Crossref, Springer, Elsevier, Wiley, Taylor & Francis, MDPI, SAGE 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. Twenty-one sources from the verified corpus were used.

4. Case 1. The artifacts do not constitute support

Su y Yang (2022) and Su y Zhong (2022) saturate the ceiling of artifacts when AI in ECE is presented as if it were support for cooperative play. Su y Yang (2022) —a scoping of seventeen studies, 1995–2021— map tools and impacts of AI for children (AI concepts, robotics, creativity, literacy, computational thinking), not that a cooperation quality score cultivates situated negotiation of goals among peers. Su y Zhong (2022) propose an AI curriculum for kindergarten and a social robot as a learning companion of AI principles: a curricular framework, not cooperative play. Inference: «high quality score = there was support» is a ceiling of analytics. A peer collaboration fidelity can coexist with an absence of co-construction among peers. The grammar of display inverts the sequence: first the classifier or the GenAI of scripts is activated; then it is declared that «there is already peer collaboration with AI».

Lee et al. (2025) explore a practical framework for incorporating AI humanoid robots in early childhood education —appropriate concepts, child–AI relationship, activities and challenges—: a framework, not a trial that the robot «be the cooperative peer» without adult mediation. Lee, Lee y Yoon (2025) —six early childhood teachers; 20–30-minute interviews— find that all six considered AI useful for learning and classroom management, but all emphasized human interactions and expressed concern about training and about AI making the teacher obsolete. Status: qualitative finding of perceptions. Inference: that voice prevents selling a «cooperative peer» agent that replaces negotiation among children. Edwards (2023) situates AI in ECEC in a postdigital framework: it is not a neutral technical add-on to the kindergarten. Giray (2026) argues that the role of AI in ECE for 0–7 years remains highly contested and that high-quality development requires authentic relationships that AI does not replicate: a critical framework. Restrictive inference: selling a robot as a cooperative peer in 3–6 confuses product with craft and erases precisely the negotiation among peers that the construct claims.

Burns, Hoan et al. (2024) saturate the anchor of technology during collaboration, with age transfer. A scoping of twenty-four articles with at least one child from 0 to 6 years; six types of hardware (iPads, computers, robots, Microsoft Kinect, multitouch tables and cameras) and seven of software. Exchange of ideas, turns, negotiation, sharing, joint understanding of goals or processes and joint action were observed. Transfer: 0–6 does not equal CENDI 3–6. Restrictive inference: robot and camera appear as hardware in collaboration among peers, not as a substitute peer nor as a scorer of cooperative minutes/hour; converting that map into a «cooperative peer» robot or into a dashboard of group task completion rate inverts the synthesis. 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 GenAI closes; the score speaks for the negotiation; the dashboard substitutes observation; the robot erases the peer; the ranking converts process into a percentile without judgment.

5. Case 2. The relational craft in the kindergarten

The floor of the craft is not a score: it is an episode in which two or more children negotiate a goal that is not reached alone, share out roles and contribution turns, repair the plan when it breaks and, when collaboration stalls, an adult observes and mediates contingently without substituting the peer (Burns, Yu et al., 2024; Burns, Brathwaite et al., 2025; Lin et al., 2024; Evaldsson, 2022; Fedewa, Watkins, Carden et al., 2025; Tan y Perren, 2023). Inference: there is support when that practice is protected; not when a quality score, a GenAI of scripts or a ranking is displayed. The craft is recognized in the episode: the children agree on a joint objective; the adult observes, groups or offers a role if needed and withdraws when negotiation among peers is sustained; the robot or the GenAI, if they appear, are subordinate means, not substitutes.

Burns, Yu et al. (2024) saturate the intervention anchor, with age transfer. A systematic review of eighteen studies to improve peer collaboration in children from 0 to 6 years. Five components: communication, shared goals, collaborative products, knowledge exchange and collaborative prosocial skills. Strategies: grouping with a criterion, evaluating products, assigning roles, adult observation and guidance. Limited attention to educators; mixed results. Transfer: 0–6 does not equal CENDI 3–6. Inference: heterogeneity and mixed results ≠ a cooperation quality score; assigning roles as a situated strategy ≠ a role-assignment checklist by prompt; the scarcity of attention to educators contradicts a dashboard of peer collaboration fidelity. Burns, Brathwaite et al. (2025) —a scoping and network analysis, Joanna Briggs methodology; one hundred and twenty-three articles, 0–6 years— find that there is no shared definition of peer collaboration nor a single way of observing it. Four domains of definition: larger goal, verbal communication, prosocial skills and knowledge exchange. Six of observation: interactive characteristics, communication, activity structure, assessment of performance, reciprocity and cognitive skills. Restrictive inference: without a shared definition nor observation, a classifier that assigns a cooperation quality score, shared-goal compliance or cooperative minutes/hour as assessment is illegitimate as a category; human «Assessment of Performance» ≠ a dashboard of group task completion rate.

Lin, Wu, Wu y Qin (2024) saturate the 5–6 classroom anchor. Quasi-experiment; sixty rural Chinese children; twelve weeks of Anji-based cooperative constructive play (CCP) versus regular instruction; a truck-race task (pre, mid, post, follow-up) and role-play video. Finding: significant improvements in cooperation in the experimental group relative to baseline, without a similar pattern in the control; follow-up exceeded pre and post. Distinction: it is cited for cooperation, not as block play, guided play nor free play. Li (2026) reconstructs Anji Play —from «no play» to teacher «fake play» and to children's «real play»— as indigenous innovation: a framework, not an AI trial. Inference: twelve weeks of CCP ≠ a shared-goal activity bank by prompt. Evaldsson (2022) —three video sets in suburban school and after-school in Sweden; setting transfer— shows microdramas that dramatize experiences, subvert hierarchies and comment on social life through collaborative performances. Inference: collaboration is situated performance, not cooperative minutes/hour; a GenAI of scripts closes the microdrama that the craft leaves open.

Fedewa, Watkins, Carden y Grbac (2025) saturate adult mediation in preschool. Multiple baseline across four dyads; peer-mediated Stay-Play-Talk, implemented by a teacher with cascade coaching; the teacher trained four peers. Finding: a functional relation with more interactive play and initiations and less solitary play in all dyads; generalization and maintenance when teacher support was withdrawn. Distinction: peer-mediated cooperative social play, not an inclusion article. Inference: training the adult and the peer ≠ a robot that substitutes the peer; withdrawing support is craft, not permanent fidelity. Fedewa, Watkins, Barber y Baggett (2024) offer special educators and speech-language pathologists an overview of strategies for play among preschoolers with and without autism: a craft framework, not a closed lesson pack. Fedewa, Watkins, Barnard-Brak y Akemoglu (2024) —twenty SCED (2000–2020) of play interventions with typically developing peers— find that the majority produced significant effects. Inference: the human peer is the mediator; effect sizes not recovered in the abstract are not invented. Watkins, Fedewa, Hu y Ledbetter-Cho (2023) —three upper-primary students and three peers; Title 1— find a functional relation with cooperative play and high human fidelity of the peers. Stage transfer: not CENDI 3–6. Inference: «fidelity» observed by the teacher ≠ a camera auto-score.

Tan y Perren (2023) —seven teachers, four inclusive classrooms in Shanghai; participant observation— find preventive and interventional strategies at five levels (community, environment, curriculum, activity, child). Distinction: mediation of interaction among peers, not an inclusion article. Inference: five situated levels ≠ a role-assignment checklist by prompt. Kilincci, Love y An (2026) —sixteen studies in inclusive preschool— find that explicit social-skills interventions, especially those mediated by peers, were the most common, mostly with 4-year-old White children with autism, and positioned typically developing peers as agents. Inference: the map does not authorize a quality score of a general classroom nor a GenAI of lesson packs. Bombjaková y Mikuškovičová (2026) —mixed methods in a Slovak kindergarten— find that children did not conflate dominance and wisdom: those perceived as wise tended to be prosocial, helpful, rule-respecting and protective. Inference: a ranking of collaborative effectiveness by vision/NLP may reward visible dominance and erase the one whom peers recognize as wise. Bier, Vaade y Witthuhn (2026) find homophily already at four years in a diverse Midwest district: strong by sex, weaker by race and socioeconomic status, moderate by special education. Inference: companion preferences are not a group task completion rate nor an activity bank that «assigns» the peer. Skehill y O’Donovan (2025) —action research between a preschool and a primary school in Ireland— treat collaboration among centers in the transition: construct transfer, not negotiation of goals among peers in play. Taken together, none of these systems signs a GenAI of scripts, a quality score, a dashboard of minutes, a substitute-robot or a ranking. The evidence is real in its domain; «score / script / robot / ranking = support for cooperative play» is an illegitimate category inference.

6. Case 3. Category boundaries

Ten boundaries protect the axis. First: free play —Burns, Yu et al. (2024) and Lin et al. (2024) situate a shared goal; 25 August covered free play—. Second: guided play (14 September): the axis is negotiation among peers, not the free–guided–directed continuum. Third: outdoor / loose parts / sand / water / playdough / risky play. Fourth: symbolic / spatial-block play —Lin et al. (2024) and Evaldsson (2022) are read for cooperation, not as articles of symbolic or of block play—. Fifth: co-regulation (11 September) and SST (10 September). Sixth: ToM (9 September). Seventh: prosocial helping/sharing (8 September) —Burns, Brathwaite et al. (2025) include prosocial skills as a definition domain; the axis is not helping or sharing as an already published article—. Eighth: conflict negotiation / turn-taking (7 September) —Burns, Hoan et al. (2024) observe turns during collaboration; the axis is not conflict nor the turn—. Ninth: inclusion as exclusive axis —Tan y Perren (2023), Kilincci et al. (2026) and Fedewa, Watkins, Carden et al. (2025) are read for the mediation of collaboration among peers—. Tenth: SEL, CLASS, joint attention, mark-making, dialogic reading, literacy, FE, numeracy, well-being, formative assessment and documentation —Burns, Brathwaite et al. (2025) anchor human observation, not automatic assessment; Skehill y O’Donovan (2025) anchor collaboration among centers, not documentation as axis—. Giray (2026) and Lee, Lee y Yoon (2025) require human mediation. Inference: coherent AI in 3–6 remains on the adult's side (preparation, documentation, reflection), not as a GenAI of lesson packs, a quality score, a dashboard, a «cooperative peer» robot nor a ranking without judgment.

7. Four tests of support (not 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 cooperative play nor for peer collaboration in early childhood.

7.1. Test of situated practice (shared goals, roles, contribution turns and joint resolution among peers), not of the classifier/ML/camera nor of the «cooperative peer» robot. Burns, Yu et al. (2024), Burns, Brathwaite et al. (2025), Lin et al. (2024) and Evaldsson (2022) define the craft/artifact contrast. Su y Yang (2022), Lee et al. (2025) and Giray (2026) map AI and robots as affordance, not as situated collaboration. Burns, Hoan et al. (2024) show robots and cameras as hardware during collaboration, not as a substitute peer nor as a scorer. Inference: if the «evidence» is a cooperation quality score, peer collaboration fidelity, shared-goal compliance, cooperative minutes/hour or a «cooperative peer» agent without adult mediation, the center did analytics or HRI, not support.

7.2. Test of adult co-presence that observes and mediates contingently —not only «there were cooperative minutes»—, not of the GenAI of cooperative play scripts / lesson packs / shared-goal activity banks / role-assignment checklists. Fedewa, Watkins, Carden et al. (2025), Tan y Perren (2023), Kilincci et al. (2026) and Fedewa, Watkins, Barber et al. (2024) situate practice and mediation. Burns, Yu et al. (2024) show that assigning roles is a heterogeneous strategy with mixed results, not a recipe. Lin et al. (2024) situate twelve weeks of CCP, not an activity bank. Inference: a script by prompt does not sign the episode in which peers renegotiate the role and the adult yields.

7.3. Test of human mediation and of relational interpretation, not of the dashboard of cooperative minutes / group task completion rate nor of the ranking of collaborative effectiveness by vision/NLP. Burns, Brathwaite et al. (2025) anchor the absence of a shared definition and observation. Bombjaková y Mikuškovičová (2026) anchor that visible dominance ≠ peer wisdom. Bier et al. (2026) anchor homophily. Fedewa, Watkins, Barnard-Brak et al. (2024) and Watkins et al. (2023, transfer) anchor human fidelity, not an auto-score. Inference: high cooperative minutes/hour do not by themselves raise negotiation of the goal. Subordinate preparation of the adult is legitimate; the compliance scorer is not.

7.4. Test of category distinction and of professional judgment, not of the product catalog. Cooperative play ≠ free play, guided play, outdoor/loose parts, symbolic/spatial, co-regulation, SST, ToM, prosocial helping/sharing, peer-conflict/turn-taking, inclusion as axis, SEL, CLASS nor automatic formative assessment. Burns, Yu et al. (2024), Lin et al. (2024), Tan y Perren (2023) and Skehill y O’Donovan (2025, transfer) prevent those confusions. Giray (2026) and Lee, Lee y Yoon (2025) require human mediation. Inference: support is fulfilled with peers who negotiate goal, role and contribution, and with an adult who mediates contingently. The framework admits digital subordinate to the adult's preparation, documentation and reflection (Lee, Lee y Yoon, 2025; Edwards, 2023; Burns, Hoan et al., 2024). It rejects declaring support by the artifacts (Giray, 2026; Su y Yang, 2022; Lee et al., 2025). The four tests are read together.

8. Discussion

Three tensions. First: displaying the artifacts versus exercising support. The cases of sections 5–6 —with marked transfers— sustain the contrast; Su y Yang (2022), Su y Zhong (2022) and Lee et al. (2025) map AI and robots without equivalence to situated cooperative play. Second: automated assessment or substitute-robot versus relational pedagogy —declaring support by a cooperation quality score or by a system that «is the cooperative peer» is inverted pedagogy; Burns, Brathwaite et al. (2025) show that the literature does not share a definition nor an observation; Burns, Hoan et al. (2024) show that robot and camera appear during collaboration, not in its place; Bombjaková y Mikuškovičová (2026) show that visible dominance is not the wisdom that peers attribute—. Third: a GenAI of activity banks / a «cooperative peer» robot / a ranking versus craft (Giray, 2026; Lee, Lee y Yoon, 2025; Edwards, 2023): selling an agent as a peer in 3–6 confuses product with practice. Burns, Yu et al. (2024) confirm mixed results and little attention to educators; Lin et al. (2024) confirm that cooperation is cultivated in weeks of joint play, not in a prompt; Fedewa, Watkins, Carden et al. (2025) confirm that the teacher trains the peer and then yields; Giray (2026) subordinates any use to the human relationship. Subordinate digital is legitimate toward the adult (preparation; reflection on the homophilous network of Bier et al., 2026; Lee, Lee y Yoon, 2025, as training); inverting the sequence is not. If the center shows peers who negotiated the goal, the role and the contribution, and an adult who mediated contingently, it may speak of support; if it only shows GenAI scripts, quality scores, cooperative minutes, «cooperative peer» robots or rankings without judgment, it speaks of artifacts.

9. Limits

This review is narrative. It does not apply its own PRISMA nor estimate combined effects. No d, r or AUC is invented. Burns, Yu et al. (2024) cover eighteen studies and 0–6 years: age transfer; mixed results. Burns, Brathwaite et al. (2025) cover one hundred and twenty-three articles (0–6): definition, not a score. Burns, Hoan et al. (2024) cover twenty-four articles (0–6): technology during collaboration, not a substitute-robot. Lin et al. (2024) are sixty rural 5–6-year-old children: they are not generalized to an AI package. Li (2026) is a historical framework, not a trial. Evaldsson (2022) is school/after-school: setting transfer. Fedewa, Watkins, Carden et al. (2025) are four dyads. Fedewa, Watkins, Barber et al. (2024) are an overview of practice. Fedewa, Watkins, Barnard-Brak et al. (2024) cover twenty SCED (2000–2020): effect sizes not recovered in the abstract are not invented. Watkins et al. (2023) are upper primary: stage transfer. Tan y Perren (2023) are seven teachers. Kilincci et al. (2026) cover sixteen studies, with sample bias. Bombjaková y Mikuškovičová (2026) are one Slovak kindergarten. Bier et al. (2026) measure homophily, not a score. Skehill y O’Donovan (2025) are transition among centers: construct transfer. Lee, Lee y Yoon (2025) measure six teachers. Lee et al. (2025), Su y Yang (2022), Su y Zhong (2022), Edwards (2023) and Giray (2026) map AI/robots/curriculum/postdigitality/critique, not the display package in CENDI. No verified trials of the artifacts as a single package of «cooperative play with AI» in a 3–6 classroom were located; they are discussed as a category ceiling. The inferences of section 7 are category hypotheses, not implementation evidence.

10. Conclusions

The artifacts —a GenAI of cooperative play scripts, lesson packs, shared-goal activity banks or role-assignment checklists; a classifier/ML/camera of cooperation quality score, peer collaboration fidelity, shared-goal compliance or cooperative minutes/hour; a dashboard of cooperative minutes or group task completion rate; a «cooperative peer» robot without contingent adult mediation; a ranking of collaborative effectiveness by vision/NLP— do not constitute support for cooperative play in early childhood education. Su y Yang (2022) and Su y Zhong (2022) confirm AI affordances without equivalence to situated collaboration. Lee et al. (2025), Lee, Lee y Yoon (2025), Edwards (2023) and Giray (2026) set frameworks, perceptions, postdigitality and critique, not a substitute-robot nor a GenAI of a recipe. Burns, Hoan et al. (2024) are read as a map of technology during collaboration, not as a license for an artificial peer. The empirical cases (sections 5–6) require reading interventions, definitions, CCP, microdramas, PMI, teacher strategies, peer culture and homophily by what they measure and by what they do not equal as situated pedagogy. When there is support there are peers who negotiate goal, role and contribution, and an adult who mediates contingently (Burns, Yu et al., 2024; Burns, Brathwaite et al., 2025; Lin et al., 2024; Fedewa, Watkins, Carden et al., 2025). Cooperative play is distinguished from free play, guided play, outdoor/loose parts, symbolic/spatial, co-regulation, SST, ToM, prosocial helping/sharing, peer-conflict/turn-taking, inclusion as axis, SEL, CLASS and automatic formative assessment.

Where the sources do not measure a kindergarten, this article does not claim it. Where they measure mappings of AI, reviews of collaboration, technology during collaboration, Anji-based CCP, microdramas, PMI, teacher strategies, peer culture, homophily or collaboration among centers, it does not translate them into support via quality score, script by prompt, dashboard of minutes, substitute-robot or ranking. Accompanying girls and boys from three to six years in cooperative play is to exercise situated practices in which peers negotiate shared goals, roles, contribution turns and joint resolution, with contingent adult mediation when needed, and co-presence that interprets collaboration as situated relational practice, not as a «cooperation quality score». The rest is a GenAI of scripts, a quality score, a dashboard of minutes, a «cooperative peer» robot and a ranking. It is not support for cooperative play in early childhood education, and it must not be presented as what it is not.

Laboratorio Editorial de NEXTECH.IA / Ingeniero Mitre.

References

  1. Bier, R., Vaade, E., y Witthuhn, C. (2026). Peers in the early years: Peer network homophily among four-year-old preschool students. Early Childhood Research Quarterly, 76, 240–249. https://doi.org/10.1016/j.ecresq.2026.03.009
  2. Bombjaková, D., y Mikuškovičová, S. (2026). Wisdom and dominance in preschool peer culture: A mixed-methods exploration. Contemporary Issues in Early Childhood. https://doi.org/10.1177/14639491251404956
  3. Burns, S., Brathwaite, L., Hoan, E., Yu, E., Yasiniyan, S., White, L., Dhuey, E., y Perlman, M. (2025). A scoping review and network analysis of the characteristics of peer collaboration in early educational settings from studies using diverse methodologies. Educational Review, 77(5), 1649–1671. https://doi.org/10.1080/00131911.2025.2450228
  4. Burns, S., Hoan, E., Hreno, S., Yu, E., Brathwaite, L., Wiebe, J. P., White, L. A., Dhuey, E., y Perlman, M. (2024). Navigating technology in the classroom: a scoping review of technology use during peer collaboration in early educational settings. Educational Review, 76(7), 2063–2089. https://doi.org/10.1080/00131911.2024.2336970
  5. Burns, S., Yu, E., Brathwaite, L., Masum, M., White, L., Dhuey, E., y Perlman, M. (2024). Improving young children's peer collaboration in early educational settings: A systematic review. Review of Education, 12(2), Article e3484. https://doi.org/10.1002/rev3.3484
  6. Edwards, S. (2023). Early childhood education and care, artificial intelligence and the postdigital. Australasian Journal of Early Childhood, 48(3), 179–181. https://doi.org/10.1177/18369391231198910
  7. Evaldsson, A.-C. (2022). Examining children’s peer play-in-action: Micro dramas and collaborative play performances. IJREE – International Journal for Research on Extended Education, 9(1), 69–86. https://doi.org/10.3224/ijree.v9i1.07
  8. Fedewa, M., Watkins, L., Barber, A., y Baggett, J. (2024). Supporting social play of preschoolers with and without autism: A collaborative approach for special educators and speech language pathologists. Early Childhood Education Journal, 52(6), 1023–1033. https://doi.org/10.1007/s10643-023-01488-6
  9. Fedewa, M., Watkins, L., Barnard-Brak, L., y Akemoglu, Y. (2024). A systematic review and meta-analysis of single case experimental design play interventions for children with autism and their peers. Review Journal of Autism and Developmental Disorders, 11(2), 361–383. https://doi.org/10.1007/s40489-022-00343-5
  10. Fedewa, M. P., Watkins, L., Carden, K., y Grbac, G. (2025). Effects of a teacher-facilitated peer-mediated intervention on social play of preschoolers with autism. Journal of Autism and Developmental Disorders, 55(5), 1595–1614. https://doi.org/10.1007/s10803-024-06320-7
  11. Giray, L. (2026). A critical stand against artificial intelligence in early childhood education. Discover Artificial Intelligence, 6(1), Article 768. https://doi.org/10.1007/s44163-026-01510-x
  12. Kilincci, E., Love, H. R., y An, Z. G. (2026). Interventions to improve peer interaction in inclusive preschool classrooms: A systematic scoping review. Journal of Early Intervention. https://doi.org/10.1177/10538151261448160
  13. Lee, J., Jo, J., Lee, J. O., y Kim, S. H. (2025). Incorporating humanoid artificial intelligence (AI) robots into early childhood education. Early Childhood Education Journal, 53(8), 2849–2857. https://doi.org/10.1007/s10643-024-01690-0
  14. Lee, J., Lee, J. O., y Yoon, J. (2025). Exploring perceptions of early childhood teachers on the use of artificial intelligence in early childhood education. Journal of Early Childhood Teacher Education. https://doi.org/10.1080/10901027.2025.2600033
  15. Li, H. (2026). The historical development of Anji Play: a model for global education. International Journal of Play, 15(2), 245–264. https://doi.org/10.1080/21594937.2026.2650990
  16. Lin, X., Wu, Y., Wu, J., y Qin, L. (2024). Enhancing cooperation in 5–6-year-old rural Chinese children through cooperative constructive play based on Anji Play: A quasi-experimental study. Behavioral Sciences, 14(7), Article 533. https://doi.org/10.3390/bs14070533
  17. Skehill, S., y O’Donovan, P. (2025). Collaboration as foundational in the transition from preschool to primary school: a shared action research project between educational settings. European Early Childhood Education Research Journal, 33(4), 645–658. https://doi.org/10.1080/1350293x.2024.2418379
  18. Su, J., y Yang, W. (2022). Artificial intelligence in early childhood education: A scoping review. Computers and Education: Artificial Intelligence, 3, Article 100049. https://doi.org/10.1016/j.caeai.2022.100049
  19. Su, J., y Zhong, Y. (2022). Artificial intelligence (AI) in early childhood education: Curriculum design and future directions. Computers and Education: Artificial Intelligence, 3, Article 100072. https://doi.org/10.1016/j.caeai.2022.100072
  20. Tan, R., y Perren, S. (2023). Promoting peer interactions in an inclusive preschool in China: what are teachers’ strategies? International Journal of Inclusive Education, 27(9), 987–1003. https://doi.org/10.1080/13603116.2021.1879955
  21. Watkins, L., Fedewa, M., Hu, X., y Ledbetter-Cho, K. (2023). A teacher-facilitated peer-mediated intervention to support interaction between students with and without autism. Advances in Neurodevelopmental Disorders, 7(2), 252–267. https://doi.org/10.1007/s41252-022-00299-x