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 co-regulation and for emotional self-regulation: a facial, voice, physiological or multimodal ML detector that scores EQ score, emotion regulation index, dysregulation alerts, affective compliance or calming percentile; a GenAI of emotion scripts, co-regulation lesson packs, feeling cards banks or calm-down recipes by prompt; a dashboard of dysregulated minutes/hour or emotion events/hour; a chatbot, robot, wearable or affective agent presented as a substitute that «co-regulates» or «calms» without adult mediation; and a ranking of teacher emotional sensitivity derived from affective computing without situated judgment. The artifacts make it possible to display that the center «already does co-regulation with AI». The leap —from score, script, 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 is adult presence, affective attunement, emotional naming in the episode, scaffolding of strategies and gradual transition to self-regulation, with co-presence and pedagogical interpretation, not an «emotion score».

The thesis is restrictive: those artifacts do not constitute support for co-regulation or for emotional self-regulation in early childhood. In early childhood education that practice develops when an adult and a child participate in an interactive process in which the adult holds, names and scaffolds affect until the child can regulate with growing autonomy (Kostøl, 2026, stage transfer; Hannan et al., 2026; Vasseleu et al., 2025; Denham et al., 2022; Lüken et al., 2025); not an emotion detector, a GenAI of scripts, a dashboard of minutes, a robot that «calms» nor a ranking of teacher sensitivity. AI may support the adult's preparation, documentation and reflection in a subordinate way (Saraç et al., 2025; Vasseleu et al., 2025; Lee, Lee y Yoon, 2025); it must not substitute the co-regulating relationship nor convert affect into a surveillance metric. Kostøl (2026) shows by video that co-regulation is an interactive process of affect, behavior and cognition, not an auto-score. Vasseleu et al. (2025) evaluate an intervention on educators' beliefs and practice: the adult's craft, not an EQ score. Denham et al. (2022) observe emotion socialization in two countries: situated reactions, not emotion events/hour. Hannan et al. (2026) introduce the emotional ecology of the center, not a calming percentile. That authorizes asking what was measured: an emotion regulation index —not the practice when an adult attunes, names and scaffolds, and a professional interprets the episode.

The problem is aggravated by ten category confusions. First: it is not generic SEL —25 August covered the CASEL package; here the axis is co-regulating affect and handing over toward self-regulation, not a SEC auto-score—. Second: it is not ToM (9 September). Third: it is not prosocial helping/sharing (8 September). Fourth: it is not conflict negotiation / turn-taking (7 September): the axis is attunement and scaffolding, not the turn. Fifth: it is not SST (10 September). Sixth: it is not executive functions as axis —Saraç et al. (2025) include an FE module and Konishi y Chesney (2024) measure HTKS; only a peripheral contrast—. Seventh: it is not CLASS nor generic interactions. Eighth: it is not joint attention. Ninth: it is not dialogic reading nor orality —Altınışık y Ünlüer (2026) are read for ER, not as dialogic reading—. Tenth: it is not teacher well-being, inclusion, participation/voice, documentation nor formative assessment. Mark-making, literacy, sand/water/playdough/loose parts/outdoor, STEM, motor skills, creativity, free play and symbolic play are not recycled: Guedes et al. (2026) are read for self-regulation in play; Andic y Basbug (2026) for dyadic co-regulation, not as loose parts. The artifacts fit on a slide; the episode in which an adult names «you are angry because…» and offers a strategy, 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 co-regulation / emotional self-regulation in early childhood education» usually mixes. The first is the construct from 3 to 6 years as situated relational practice: an interactive process in which the adult co-regulates with the child —presence, affective attunement, emotional naming, scaffolding of strategies— and hands over gradually toward self-regulation; flexibility of strategies matters more than a single recipe (Kostøl, 2026, transfer; Hannan et al., 2026; Lüken et al., 2025; Denham et al., 2022; Vasseleu et al., 2025). The second is the craft that cultivates it —an adult who interprets the episode; names without reducing to correctness; holds the body and the voice; co-presence that interprets, not only «there was an emotion event»— (Saraç et al., 2025; Vasseleu et al., 2025; Hannan et al., 2026; Loomis, 2021). The third is the evidence of AI affordances in ECE, robots, teacher perceptions, GenAI and emotion recognition/assessment systems —without equating them to situated pedagogy— (Su y Yang, 2022; Su y Zhong, 2022; Lee et al., 2025; Lee, Lee y Yoon, 2025; Ormanci, 2026; Giray, 2026; Radočaj y Martinović, 2025; Li y Li, 2026). The fourth is the framework of human measure, which treats the infant as a subject who regulates with another, not as a vector of dysregulated minutes/hour nor as the addressee of a robot that «calms» (Perry y Dollar, 2021; Lüken et al., 2025; Giray, 2026). Inference: an EQ score does not attune «you are angry because they took your turn»; an adult who co-regulates does. Su y Yang (2022) map AI without equating it to situated co-regulation. Lee et al. (2025) propose incorporating robots, not substituting attunement. 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: teacher attention in emotional situations (Kostøl, 2026, transfer), emotional ecology (Hannan et al., 2026), educators' practice (Vasseleu et al., 2025; Saraç et al., 2025), emotion socialization (Denham et al., 2022), flexibility of strategies (Lüken et al., 2025), ER mediated by reading and journaling (Altınışık y Ünlüer, 2026), dyadic CODE framework (Andic y Basbug, 2026) and self-regulation in play (Guedes et al., 2026). Inference: an emotion regulation index does not scaffold the strategy that the episode requires; an adult who interprets the situation does. Perry y Dollar (2021) anchor human measure. Li y Li (2026) and Radočaj y Martinović (2025, clinical transfer) saturate the ceiling of the detector and of multimodal assessment.

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 co-regulation classroom, is primary, home/parents, ASD clinic, HRI/robot as framework, AI curriculum, or psychometric measure without equating to automatic pedagogy; (b) emotional co-regulation, emotional self-regulation, teacher emotion socialization, flexibility of ER strategies, teacher education to support self-regulation, emotional ecology of the center, or AI in ECE / robots / GenAI / emotion recognition or assessment with artifact/craft relevance; (c) kindergarten, preschool, CENDI or 3–6, or explicit transfer; (d) peer-reviewed journal with DOI; (e) verifiable DOI or publisher page. Excluded as central object were axes already used in this series —generic SEL, ToM, prosocial, peer-conflict, SST, participation/voice, CLASS, joint attention, inclusion, teacher well-being, formative assessment, documentation, mark-making, dialogic reading, literacy, orality, sand/water/playdough/loose parts/outdoor, free play, symbolic play, STEM, motor skills, creativity and inquiry, and FE as axis—. Altınışık y Ünlüer (2026) are read for ER; Guedes et al. (2026) for self-regulation in play; Andic y Basbug (2026) for dyadic co-regulation; Denham et al. (2022) for emotion socialization, not as generic SEL. Perry y Dollar (2021) = measure ≠ automatic pedagogy. Kostøl (2026) = primary: transfer. Radočaj y Martinović (2025) = FER in ASD: transfer. Ștefan y Dănilă (2025) = home: transfer. Saraç et al. (2025) and Konishi y Chesney (2024) = FE only peripheral.

The search was executed on 11 September 2026 (slot 09:02 America/Mexico_City, weekday routine) on DOI pages, Crossref, Springer, Elsevier, Wiley, Taylor & Francis, MDPI, Nature 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 co-regulation. 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 and, in some cases, emotional control as a reported outcome), not that an emotion regulation index cultivates situated attunement. 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 co-regulation. Inference: «high EQ score = there was support» is a ceiling of analytics. A calming percentile can coexist with an absence of attunement. The grammar of display inverts the sequence: first the detector is activated; then it is declared that «there is already co-regulation 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 «co-regulates» or «calms». 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 chatbot, a wearable or an affective agent that «co-regulates» without mediation. Ormanci (2026) —thirty-seven articles— finds an increase of interest in AI and early childhood education after 2023, with ChatGPT as the most frequent tool, followed by robotic tools. Inference: that frequency does not authorize emotion scripts by prompt nor a substitute-robot of attunement. 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 an affective agent as a regulation partner in 3–6 confuses product with craft.

Radočaj y Martinović (2025) saturate the facial-detector anchor and must be read with marked transfer. They compare CNN and transformer for facial emotion recognition in autistic children (natural, anger, fear, joy); transformers outperformed CNNs. Status: technical FER finding. Transfer: clinical ASD sample, not a CENDI 3–6 classroom; frame classification ≠ co-regulation; this article is not converted into inclusion. Inference: a system that «reads emotions» in the face does not authorize a detector as pedagogical assessment. Li y Li (2026) saturate the multimodal-assessment anchor: a system that integrates facial expression, voice emotion and behavioral movement to assess socioemotional competence in preschoolers. Status: finding of an assessment system. Restrictive inference: assessing by face, voice and movement does not equal presence, naming and scaffolding nor authorize an emotion regulation index or a dashboard of emotion events/hour. Detector ≠ adult who attunes; GenAI of feeling cards ≠ open episode; dashboard of dysregulated minutes/hour ≠ observation; chatbot/robot/wearable that «calms» ≠ co-presence; ranking of teacher sensitivity ≠ situated judgment. 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 affect; the GenAI closes; the dashboard substitutes observation; the robot erases mediation; 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 adult presence; affective attunement; emotional naming in the episode; scaffolding of strategies adjusted to the situational demand; gradual transition to self-regulation; co-presence that interprets the episode —not only «there was an emotion event»—; and, when there is uncertainty, the adult holds the body and the voice instead of delegating to an agent that «calms» (Kostøl, 2026, transfer; Hannan et al., 2026; Vasseleu et al., 2025; Denham et al., 2022; Lüken et al., 2025; Saraç et al., 2025). Inference: there is support when that practice is protected; not when an EQ score, a GenAI of scripts or a sensitivity ranking is displayed. The craft is recognized in the episode: the child overflows; the adult approaches, names («you are angry because…»), offers a strategy that the episode admits and hands over when the child can take it; the robot or the GenAI, if they appear, are subordinate means to the adult, not substitutes.

Kostøl (2026) saturates the video-recorded-practice anchor and must be read with stage transfer. It analyzes 103 hours of video from five primary classes, following eight teachers, in situations of visible emotion; it identifies foci of attention on emotions, behavior and academic matters and insists on the joint importance of affect, behavior and cognition (ABC) in co-regulation. Status: empirical finding of video analysis. Transfer: primary, not CENDI 3–6. Marked inference: co-regulation is situated attention, not a dysregulation alert; the ABC focus does not authorize a detector nor a ranking of teacher sensitivity. Hannan, Pascal y Bertram (2026) —three centers in the southwest of the United Kingdom; observations, documentary analysis and interviews with leadership— find that emotional environments are complex and that there is ambiguity, particularly in the framing of affect; they propose the Emotional Ecology of ECEC centers. Status: qualitative finding and framework. Inference: a complex emotional environment cannot be reduced to a calming percentile.

Vasseleu, Neilsen-Hewett y Howard (2025) saturate the intervention-on-the-adult-craft anchor: mixed study of the Preschool Situational Self-Regulation Toolkit (PRSIST) program on educators' beliefs and self-reported changes of knowledge and practice. Empirical finding: significant improvement of confidence in knowledge of self-regulation; qualitative findings suggest a positive perceived change in knowledge and practice. Marked inference: educating the adult ≠ GenAI of co-regulation lesson packs; confidence is not a sensitivity ranking by affective computing. Saraç, Gülay Ogelman, Yurtseven y Akgün (2025) —68 preschool teachers from 26 cities in Türkiye; Understanding by Design and blended learning; five modules (executive functions, metacognition, emotional and motivational regulation); nested mixed design— find a significant increase in the use of metacognition and emotional-motivational regulation practices. Marked distinction: the FE module is peripheral; the axis is educating the adult to support self-regulation. Inference: training is preparation of the craft, not an emotion script by prompt.

Denham, Mortari y Silva (2022) —117 teachers and 381 preschoolers in classrooms in the United States or Italy; observation of emotions and reactions in small group— find cultural differences and that teacher emotion socialization contributed to children's emotions and reactions to teacher emotion, as well as to SEC with peers. Marked distinction: it is cited for reactions to affect in the episode, not as generic SEL nor as free play. Inference: emotion socialization is a craft of contingent reactions, not an EQ score. Lüken, Silkenbeumer, Holodynski y Kärtner (2025) —60 preschoolers (M = 59,67 months); frustrating persistence and challenging waiting, with opposite effects of attentional deployment— find that the effectiveness of strategies changed as predicted and that adapting them to the task was positively associated with age and with overall effectiveness. Status: empirical finding of ER. Restrictive inference: «one size fits all» does not hold; a calm-down recipe by prompt misreads situational flexibility; it is not converted into an FE axis.

Altınışık y Ünlüer (2026) —67 children, mothers and four teachers; shared reading of emotion books, journaling, both, and a control— find positive effects on emotion regulation, stronger when implemented together, with persistence at three weeks. Marked distinction: it is read for mediated ER, not as dialogic reading. Inference: reading and journaling with an adult ≠ GenAI of emotion scripts. Andic y Basbug (2026) introduce the CODE model —Co-regulation, Object-mediation, Dyadic Exchange— as a framework of object-mediated dyadic co-regulation (Trigger–Intervention–Resolution cycle; regulation distributed in materials that anchor scripts). Status: conceptual framework, not a trial. Marked distinction: it is read for dyadic co-regulation, not as play or loose parts. Inference: an object can anchor strategies under adult mediation; it does not authorize a wearable or a robot that «co-regulates» without it. Guedes, Lerkkanen y Cadima (2026) observe children's self-regulation across play situations and peer effects in preschool classrooms. Status: empirical classroom finding. Marked distinction: it is read for self-regulation in the episode, not as free play. Inference: observing self-regulation with peers ≠ dashboard of emotion events/hour.

Loomis (2021) —126 preschoolers, predominantly Latino and Black; cumulative adversity; self-regulation and student–teacher relationship— finds that mid-year self-regulation problems mediated the relationship between adversities and student–teacher conflict at the end of the year. Marked inference: the adult–child relationship is the tissue of co-regulation; a dysregulation alert does not repair it. Ștefan y Dănilă (2025) —330 preschoolers; paths from parental ER, emotion socialization and parenting practices— find that parental reappraisal was directly associated with adaptive ER, with indirect effects via emotion coaching and positive parenting. Transfer: home, not a CENDI classroom. Perry y Dollar (2021) —N = 362 caregivers; children aged 2–6 years; partial transfer by including 2 years— present the EERBQ as a measure of behavioral ER strategies. Contrast inference: EERBQ is a human measure, not a multimodal auto-score. Konishi y Chesney (2024) —eight education students; four weeks; kindergarten Project Construct versus control; HTKS— find growth of constructivist knowledge and of HTKS in both classrooms, with no significant differences between conditions. Transfer and peripheral contrast: HTKS is not classroom emotional co-regulation nor an emotion score; this article is not converted into an FE axis.

Taken together: there are systems that analyze teacher co-regulation by video, describe the emotional ecology of the center, train educators to support self-regulation, observe emotion socialization, measure flexibility of strategies, intervene with reading and journaling, propose a dyadic framework, observe self-regulation in play, link adversity and relationship, trace parental paths or publish a human measure. None —by itself— signs a facial detector, a GenAI of scripts, a dashboard of minutes, a substitute-robot or a teacher ranking. The evidence is real in its domain; «score / script / robot / ranking = support for co-regulation» is an illegitimate category inference.

6. Case 3. Category boundaries

Ten boundaries protect the axis. First: generic SEL —Denham et al. (2022) measure emotion socialization; 25 August covered the CASEL package; here the axis is presence, attunement, naming and scaffolding, not a SEC auto-score—. Inference: «we work SEL = there is co-regulation» does not sign. Second: ToM (9 September). Third: prosocial helping/sharing (8 September). Fourth: conflict negotiation / turn-taking (7 September). Fifth: SST (10 September). Sixth: executive functions —Saraç et al. (2025) cover FE in a module; Konishi y Chesney (2024) measure HTKS; Lüken et al. (2025) measure ER flexibility, not FE as axis; only peripheral—. Seventh: CLASS / generic interactions —Hannan et al. (2026) and Loomis (2021) anchor environment and relationship; not an interaction quality percentile—. Eighth: joint attention. Ninth: dialogic reading / orality —Altınışık y Ünlüer (2026) are read for mediated ER, not as dialogic reading—. Tenth: teacher well-being, inclusion, participation/voice, documentation and formative assessment —Radočaj y Martinović (2025) are read as a detector ceiling with ASD transfer, not as inclusion; Vasseleu et al. (2025) and Saraç et al. (2025) are read for craft, not as teacher well-being—. Giray (2026) and Lee, Lee y Yoon (2025) require human mediation. Inference: coherent AI in 3–6 remains on the side of the adult (preparation, documentation, reflection), not as a detector, a GenAI of lesson packs, a dashboard, a robot that «calms» nor a ranking without judgment. A center that «works SEL» or «scores CLASS» has not demonstrated by that label the craft of attuning, naming and scaffolding. Guedes et al. (2026) are read for self-regulation, not as free play; Andic y Basbug (2026) for dyadic co-regulation, not as loose parts; Perry y Dollar (2021) as a measure, not as automatic formative assessment.

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 co-regulation or for emotional self-regulation in early childhood.

7.1. Test of situated practice (presence, affective attunement, emotional naming, scaffolding of strategies, transition to self-regulation), not of the facial/voice/physiological/ML detector nor of the chatbot/robot/wearable that «co-regulates» or «calms». Kostøl (2026, transfer), Hannan et al. (2026), Denham et al. (2022), Lüken et al. (2025) and Vasseleu et al. (2025) define the craft/artifact contrast. Su y Yang (2022), Lee et al. (2025), Radočaj y Martinović (2025, transfer) and Li y Li (2026) map AI, robots, FER and multimodal assessment as affordance or engineering, not as situated co-regulation. Inference: if the «evidence» is emotion score, EQ score, emotion regulation index, dysregulation alerts, affective compliance, calming percentile or an agent that «calms» without an adult, the center did analytics or HRI, not support.

7.2. Test of adult co-presence that names and scaffolds —not only «there was an emotion event»—, not of the GenAI of emotion scripts / co-regulation lesson packs / feeling cards banks / calm-down recipes. Saraç et al. (2025), Vasseleu et al. (2025), Denham et al. (2022) and Altınışık y Ünlüer (2026) situate practice and mediation. Lee, Lee y Yoon (2025) and Giray (2026) require human interactions. Andic y Basbug (2026) admit an object as an anchor subordinate to the dyad. Inference: a script by prompt does not sign the episode in which the adult names affect and hands over a strategy that the child can take. Lüken et al. (2025) show that the strategy is not unique: a calm-down recipe closes what flexibility opens.

7.3. Test of human mediation and of the interpretation of co-regulation as relational practice, not of the dashboard of dysregulated minutes/hour / emotion events/hour nor of the ranking of teacher emotional sensitivity by affective computing. Perry y Dollar (2021) anchor human measure. Hannan et al. (2026) anchor the complexity of the emotional environment. Loomis (2021) anchors the student–teacher relationship, not the surveillance of counts. Li y Li (2026) saturate the ceiling of multimodal assessment. Inference: high dysregulated minutes/hour do not by themselves raise scaffolding. Subordinate preparation of the adult is legitimate; the scorer that converts the child or the teacher into a compliance vector is not.

7.4. Test of category distinction and of professional judgment, not of the product catalogue. Co-regulation / emotional self-regulation ≠ generic SEL, ToM, prosocial, peer-conflict, SST, FE as axis, CLASS, joint attention, dialogic reading, inclusion nor teacher well-being. Denham et al. (2022), Altınışık y Ünlüer (2026), Guedes et al. (2026), Saraç et al. (2025) and Konishi y Chesney (2024, peripheral) prevent those confusions. Giray (2026) and Lee, Lee y Yoon (2025) require human mediation. Inference: support is fulfilled with presence, attunement, naming and scaffolding of strategies with gradual handover —not by scoring emotions nor by substituting the adult with an affective agent. The framework admits digital subordinate to the adult's preparation, documentation and reflection (Vasseleu et al., 2025; Saraç et al., 2025; Lee, Lee y Yoon, 2025). It rejects declaring support by the artifacts (Giray, 2026; Su y Yang, 2022; Li y Li, 2026; Radočaj y Martinović, 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 Ormanci (2026) map AI, ChatGPT and robots without equivalence to situated co-regulation. Second: automated assessment or substitute-robot versus relational pedagogy —declaring support by EQ score or by a system that assesses socioemotional competence by face, voice and movement is inverted pedagogy; Perry y Dollar (2021) show that even a human measure of strategies is not the craft; Li y Li (2026) saturate the ceiling of that inversion—. Third: GenAI of feeling cards / a chatbot that «calms» / a sensitivity ranking versus craft (Giray, 2026; Lee, Lee y Yoon, 2025; Lee et al., 2025; Radočaj y Martinović, 2025): selling an affective agent as co-regulation in 3–6 confuses product with practice. Vasseleu et al. (2025) and Saraç et al. (2025) confirm that the intervention operates on the adult; Lüken et al. (2025) confirm that the strategy is not unique; Giray (2026) subordinates any use to the human relationship. Subordinate digital is legitimate toward the adult (preparation; EERBQ for the human observer; training); inverting the sequence is not. If the center shows presence, attunement, naming and scaffolding, it may speak of support; if it only shows detectors, GenAI scripts, dysregulated minutes/hour, robots that «calm» 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. Kostøl (2026) is video analysis of primary (103 hours; five classes; eight teachers): transfer. Hannan et al. (2026) are three centers and a framework. Vasseleu et al. (2025) measure educators' beliefs and practice, not a detector. Saraç et al. (2025) measure 68 teachers: the FE module is not converted into an axis. Denham et al. (2022) observe 117 teachers and 381 children: emotion socialization, not generic SEL. Lüken et al. (2025) measure 60 preschoolers in the laboratory. Altınışık y Ünlüer (2026) measure 67 children: they are not converted into dialogic reading. Andic y Basbug (2026) are a CODE framework, not a trial. Guedes et al. (2026) are read for self-regulation in play; N not recovered in the abstract are not invented. Loomis (2021) measures 126 preschoolers. Ștefan y Dănilă (2025) are parental paths: home transfer. Perry y Dollar (2021) validate EERBQ (362 caregivers; 2–6): measure, partial transfer. Konishi y Chesney (2024) are eight students and HTKS: peripheral of FE. Lee, Lee y Yoon (2025) measure six teachers. Lee et al. (2025), Su y Yang (2022), Su y Zhong (2022) and Ormanci (2026) map AI/robots/trends, not the display package in CENDI. Radočaj y Martinović (2025) are FER of ASD: transfer. Li y Li (2026) are multimodal assessment, not pedagogy. Giray (2026) is a critical perspective. No verified trials of the artifacts as a single package of «co-regulation 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 facial/voice/physiological/ML detector or scores of EQ / emotion regulation index / dysregulation alerts / affective compliance / calming percentile; a GenAI of emotion scripts/lesson packs/feeling cards/calm-down recipes; a dashboard of dysregulated minutes/hour; a chatbot/robot/wearable that «calms» without adult mediation; a ranking of teacher sensitivity by affective computing— do not constitute support for co-regulation or for emotional self-regulation in early childhood education. Su y Yang (2022) and Su y Zhong (2022) confirm AI affordances without equivalence to situated co-regulation. Lee et al. (2025), Lee, Lee y Yoon (2025), Ormanci (2026) and Giray (2026) set frameworks, perceptions, trends and critique, not a substitute-robot. Radočaj y Martinović (2025) and Li y Li (2026) are read as an artifact ceiling, not as the craft of 3–6. The empirical cases (sections 5–6) require reading video of co-regulation, emotional ecology, teacher education, emotion socialization, flexibility of strategies, reading and journaling, the CODE model, self-regulation in play, relationship and human measure by what they measure and by what they do not equal as situated pedagogy. When there is support there is presence, attunement, naming, scaffolding and gradual handover (Kostøl, 2026, transfer; Hannan et al., 2026; Vasseleu et al., 2025; Denham et al., 2022; Lüken et al., 2025). Co-regulation is distinguished from generic SEL, ToM, prosocial, peer-conflict, SST, FE as axis, CLASS, joint attention, dialogic reading, inclusion and teacher well-being.

Where the sources do not measure a kindergarten, this article does not claim it. Where they measure mappings of AI, robots, GenAI of trends, clinical FER, multimodal assessment, primary video, emotional ecology, educator training, emotion socialization, flexibility of strategies, reading and journaling, CODE framework, self-regulation in play, adversity and relationship, parental paths or human measure, it does not translate them into support via emotion score, script by prompt, dashboard of minutes, substitute-robot or teacher ranking. Accompanying girls and boys from three to six years in co-regulation and emotional self-regulation is to exercise presence, affective attunement, emotional naming in the episode, scaffolding of strategies adjusted to the situational demand, gradual transition to self-regulation, and co-presence that interprets affect as situated relational practice, not as an «emotion score». The rest is a detector, a GenAI of scripts, a dashboard of minutes, a robot that «calms» and a ranking. It is not support for co-regulation or for emotional self-regulation in early childhood education, and it must not be presented as what it is not.

Laboratorio Editorial de NEXTECH.IA / Ingeniero Mitre.

References

  1. Altınışık, E., y Ünlüer, E. (2026). Enhancing preschoolers’ emotion regulation skills through shared book reading and journaling. Early Childhood Education Journal. https://doi.org/10.1007/s10643-026-02212-w
  2. Andic, S., y Basbug, F. (2026). The CODE model: Object-mediated play as a scalable framework for dyadic co-regulation in early childhood. Acta Psychologica, 267, Article 107062. https://doi.org/10.1016/j.actpsy.2026.107062
  3. Denham, S., Mortari, L., y Silva, R. (2022). Preschool teachers’ emotion socialization and child social-emotional behavior in two countries. Early Education and Development, 33(5), 806–831. https://doi.org/10.1080/10409289.2021.2015915
  4. Giray, L. (2026). A critical stand against artificial intelligence in early childhood education. Discover Artificial Intelligence, 6, Article 768. https://doi.org/10.1007/s44163-026-01510-x
  5. Guedes, C., Lerkkanen, M., y Cadima, J. (2026). Observing children’s self-regulation across play situations and peer effects on self-regulation during play in preschool classrooms. Early Childhood Research Quarterly. https://doi.org/10.1016/j.ecresq.2026.07.009
  6. Hannan, L., Pascal, C., y Bertram, T. (2026). An investigation into the emotional environment: introducing the emotional ecology of early childhood education and care settings. European Early Childhood Education Research Journal. https://doi.org/10.1080/1350293x.2026.2631608
  7. Konishi, H., y Chesney, R. (2024). How constructivism influences pre-service teachers’ beliefs and practices and child self-regulation. Journal of Early Childhood Teacher Education, 45(3), 337–351. https://doi.org/10.1080/10901027.2024.2345064
  8. Kostøl, E. (2026). Teachers’ co-regulation in classrooms: a video-based analysis of teachers’ foci of attention in emotional situations. Journal of Early Childhood Teacher Education, 175–189. https://doi.org/10.1080/10901027.2025.2455505
  9. Lee, J., Jo, J., Lee, J., y Kim, S. (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
  10. Lee, J., Lee, J., 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
  11. Li, J., y Li, X. (2026). A multimodal artificial intelligence system integrating facial expression voice emotion and behavioral movement analysis for assessing social-emotional competence in preschool children. Scientific Reports. https://doi.org/10.1038/s41598-026-61418-5
  12. Loomis, A. (2021). The influence of early adversity on self-regulation and student-teacher relationships in preschool. Early Childhood Research Quarterly, 54, 294–306. https://doi.org/10.1016/j.ecresq.2020.10.004
  13. Lüken, L., Silkenbeumer, J., Holodynski, M., y Kärtner, J. (2025). One size fits all?—The form and function of preschoolers’ strategy flexibility in emotion regulation. Social Development. https://doi.org/10.1111/sode.12791
  14. Ormanci, U. (2026). Artificial intelligence in early childhood education: a systematic review of trends, methods, and recommendations for quality education. European Early Childhood Education Research Journal. https://doi.org/10.1080/1350293x.2026.2651904
  15. Perry, N., y Dollar, J. (2021). Measurement of behavioral emotion regulation strategies in early childhood: The Early Emotion Regulation Behavior Questionnaire (EERBQ). Children, 8(9), Article 779. https://doi.org/10.3390/children8090779
  16. Radočaj, P., y Martinović, G. (2025). Emotion recognition in autistic children through facial expressions using advanced deep learning architectures. Applied Sciences, 15(17), Article 9555. https://doi.org/10.3390/app15179555
  17. Saraç, S., Gülay Ogelman, H., Yurtseven, N., y Akgün, E. (2025). Learning to support self-regulation: empowering preschool teachers through understanding by design and blended learning. Teacher Development, 29(3), 510–532. https://doi.org/10.1080/13664530.2024.2408373
  18. Ștefan, C., y Dănilă, I. (2025). Pathways from parenting emotion regulation, emotion socialization and parenting practices to preschoolers’ emotion regulation. Early Childhood Research Quarterly, 72, 273–283. https://doi.org/10.1016/j.ecresq.2025.04.006
  19. 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
  20. 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
  21. Vasseleu, E., Neilsen-Hewett, C., y Howard, S. (2025). An early start to self-regulation: Evaluating the effects of an early childhood self-regulation intervention on educator beliefs, knowledge, and practice. Journal of Research in Childhood Education, 643–663. https://doi.org/10.1080/02568543.2024.2396907