1. Introduction and problem

In the 3-to-6-year span —kindergarten, preschool, CENDI, infant school— a package of three artefacts has been installed that claim to count as child participation. The first is a “feedback” button: a face, a thumb or an icon the child taps, exhibited as proof that the setting “listens to children”. The second is an avatar that “consults”: a conversational agent, a voice assistant or an on-screen character that asks questions, captures answers and, with that dialogue, declares that children’s voice is already in place. The third is a model that summarises opinion: a system that transcribes, clusters or paraphrases what the child said or tapped and delivers a summary to the adult or to coordination as if that condensate were the decision. All three make it possible to exhibit that the setting “already does participation with artificial intelligence”. The enunciative leap is enormous: one moves from capturing an input to affirming that there is participation. That leap is authorised neither by the right to be heard nor by evidence from early childhood classrooms.

The thesis of this article is restrictive. A feedback button, an avatar that consults the child or a model that summarises the child’s opinion do not constitute child participation. Participation in early childhood education is agency, situated listening and shared decision-making with responsible adults; not an input captured by an interface. Ward and Lundy recall that “voice” is not enough: the right requires space to express views, an audience that genuinely listens, and influence on what is decided, and in early childhood those four elements are relational, not a technical channel (Ward & Lundy, 2024). Lundy, Murray, Smith and Ward warn against a rights inflation that turns the young child into an “expert on their world” or into a subject who must opine on everything, and recall that the right to be heard is read alongside the adult’s duty to guide, protect and decide with responsibility (Lundy et al., 2024). An algorithm that first extracts an utterance and then summarises it is not making the child participate. It is administering a signal the adult did not hear in situ.

The problem is worsened by a professional reason that product sheets do not mention. Participation is not a capture event, nor a satisfaction survey, nor a dialogue with a character. It is a process in which children influence, with the adult as duty-bearer, matters that affect them in the everyday life of the setting. Correia, Aguiar and the PARTICIPA consortium apply to ECEC the model of space, voice, audience and influence and stress that participation is most meaningful when rooted in daily life, not when it is detached from it as a module (Correia et al., 2022). Confusing the product —click, speaking turn with an avatar, summary— with the process is the category error this paper names. Inference, marked as such: the market for “AI for children’s voice” inherits the error and automates it. It turns agency —a matter of professional practice and of rights— into an input to be extracted.

This paper does not recycle the axes already treated in this series. The question is what counts as participation when an early childhood setting “does AI”. The contributions are three: reconstruct the state of the art that separates the capture of an input from agency and shared decision-making; examine three families of cases; and offer four tests for deciding when a kindergarten may affirm that there is participation, and not only an input.

2. State of the art: from captured voice to the right to participate

Four strata that the market for “AI for children’s voice” tends to mix should be kept apart. The first is participation as a right and as a process: space, voice, audience and influence, with the adult as duty-bearer, not as the operator of a form (Ward & Lundy, 2024; Correia et al., 2022; Lundy et al., 2024). The second is the evidence of how that participation is built —or blocked— in classrooms for 3- to 7-year-olds (Hicban, Theobald, & Mascadri, 2024; Chicken & Tyrie, 2023; Waters-Davies et al., 2024; Holla Sivertsen & Bjørgen, 2025). The third is AI that speaks with the child or summarises what the child says (Xu et al., 2022; Kim et al., 2022; Xu, Branham, Deng, Collins, & Warschauer, 2021; Chen, 2024; Su & Yang, 2022; Ljungcrantz, 2026). The fourth is the framework of rights in digital environments and of high risk, which treats the 3–6-year-old as a subject of human oversight, of the best interests of the child and, in European territory, of high-risk systems when AI determines access, admission, assignment, assessment of learning or educational level.

In the rights stratum, Ward and Lundy (2024) revisit for early childhood the four elements: space is relational and emotional, not a channel; the voice of young children is seen and heard in multimodal ways, not only as text; the audience has a duty of active listening; influence requires that the adult be open to children’s perspectives carrying weight. Lundy et al. (2024) mark three frequent misunderstandings: that all children can, should and want to opine on all matters; that they should always be allowed to learn “in freedom”; and that they are the experts on their world. Status: rights framework, not an interface trial. Inference, marked as such: a button, an avatar or an automated summary can simulate the first misunderstanding —extract an “opinion” from everyone, about everything, in a click— and the third —treat the child’s output as a verdict. They do not cover audience or influence. The Committee on the Rights of the Child requires that children’s views in the digital environment be given due weight and that such participation not translate into undue monitoring or data collection that violates privacy, thought or opinion (Committee on the Rights of the Child, 2021). Inference: capturing in order to summarise can be exactly that monitoring by another name.

In the classroom stratum, Hicban, Theobald and Mascadri (2024) show that participation rights are co-constructed turn by turn: adults mobilise or obstruct the opportunity for the child to have a say. Chicken and Tyrie (2023) find, with three teachers of 3- to 7-year-olds in Wales, that voice is bounded by place, time and adult expectations. Waters-Davies et al. (2024) situate the construction of the capable child as a threshold. Holla Sivertsen and Bjørgen (2025) hear sixteen five-year-olds who perceive circle time as a passive position. Status: process finding, not AI. Inference: when participation “works”, it is not an input. It is a relationship in which the adult listens in situ and shares the decision.

In the AI stratum, Xu et al. (2022) show that a conversational agent can replicate benefits of dialogic reading with 117 children aged 3 to 6: more story-relevant vocalisations, fewer irrelevant ones, better comprehension. Kim et al. (2022) find that a commercial assistant responds appropriately to the content of children’s speech only half the time, among 28 participants aged 5 to 10. Xu, Branham, Deng, Collins and Warschauer (2021) evaluate 143 voice apps for children: most do not sustain open-ended dialogue or extended back-and-forth. Chen (2024) maps 18 articles, 11 countries and 15,081 children aged 2 to 8; Su and Yang (2022) review 17 studies up to 2021; Ljungcrantz (2026) updates 39 studies from 2020–2024, with a peak in 2024 and a focus on ages 4 to 6. Status: AI field finding, not participation. Inference: eliciting vocalisations, transcribing a request or personalising a pathway is not, because it is “dialogue”, participation in Lundy’s sense. It is capture.

In the rights-and-systems stratum, the Recommendation on the ethics of AI requires human oversight and particular attention when children are involved (UNESCO, 2021). Miao and Holmes (2023) set a threshold of 13 years for independent conversations with generative platforms and require pedagogical validation. UNICEF (2021) prioritises the best interests of the child and treats participation as one of the lenses of child-centred AI. Regulation (EU) 2024/1689 treats as high-risk, in Annex III, systems intended to determine access, admission, assignment, assessment of learning outcomes or educational level (European Union, 2024). The European Commission (2022) and the U.S. Department of Education (2023) agree that professional judgement must not be replaced. OECD (2021, 2023, 2025) anchors quality in process interactions, documents digitalisation in ECEC and situates staff practices with children, not a “consultation” channel to the child. NAEYC (2022) requires developmentally appropriate practice and that decisions serve the child’s education, not a dashboard. Inference: a five-year-old is not the user of a feedback button. The child is the subject of a right that a duty-bearing adult exercises in the activity, with the group in view.

3. Review method

A critical narrative review was conducted, not a meta-analysis. The purpose was to articulate an argument of pedagogical and rights category with verified sources, not to estimate a homogeneous effect size. Inclusion criteria: (a) 2021–2026, with the participation model revisited for early childhood (Ward & Lundy, 2024; Lundy et al., 2024); (b) child participation as agency, listening and influence, or AI that interacts with 3–6-year-olds or summarises their speech; (c) relevance to kindergarten, preschool, CENDI or ages 3–6; (d) peer-reviewed journal, DOI, or report from NAEYC, UNESCO, OECD, UNICEF, the Committee on the Rights of the Child, the European Union, the European Commission or a department of education; (e) verifiable DOI or publisher page. Axes already used in this series were excluded as central objects. Privacy is cited only as a rights boundary.

The search was executed on 26 August 2026 on DOI pages, Springer, Elsevier, Wiley, SAGE, ACM, Brill, OECD iLibrary, UNESDOC, UNICEF, OHCHR, EUR-Lex, ERIC, NAEYC and publisher sites. Each source was verified against at least one of those pages. The analysis distinguished three enunciative statuses. Empirical finding: what was observed or measured in the sample. Conceptual or normative framework: what a framework or a regulation prescribes. Pedagogical inference: the translation to kindergartens and CENDI centres, marked as such.

4. Case 1. Capturing an input is not participating: the button, the avatar and the summary

Xu, Aubele, Vigil, Bustamante, Kim and Warschauer (2022) publish in Child Development the trial that best illustrates the second artefact when it is presented as “consulting” the child. With a 2 × 2 factorial design (dialogic reading or not × agent or human), they randomly assign 117 children aged 3 to 6. The agent replicates benefits of dialogic reading: more story-relevant vocalisations, fewer irrelevant ones and better comprehension. Status. Empirical finding of story comprehension, in the laboratory, not of participation in kindergarten decisions. The authors do not claim that eliciting vocalisations is enabling the child to participate in matters affecting them. Pedagogical inference, marked as such: this is the gesture a setting copies when it “does participation with an avatar”. A speaking turn or a click is taken, assigned the status of “voice”, and the right is declared fulfilled. What there is is an input. Participation, in Ward and Lundy (2024) and in Correia et al. (2022), requires audience and influence. An agent that extracts vocalisations does not move the group’s decision. It condenses it as data.

Kim et al. (2022) saturate the portrait of the assistant that “hears” the child. In the laboratory, 28 participants aged 5 to 10 interact with a commercial assistant. The device transcribes children’s speech more than 84% of the time, but responds appropriately to content only half the time. In free conversation, the most common goal is to build a relationship with the device. Status: finding for ages 5 to 10, not isolated to 3–6; not a participation finding. Inference, marked as such: if the device responds wrongly half the time, there is no audience. There is speech recognition with a hit rate. A kindergarten that declares “the avatar consults the child” cannot, with this evidence, affirm that there is listening. It can affirm that there is a channel. Xu, Branham, Deng, Collins and Warschauer (2021) evaluate 143 voice apps for children on Google Assistant and Amazon Alexa: most do not sustain open-ended dialogue or extended exchanges. Inference: the market for “child consultation” already fails at sustained dialogue. It fails, a fortiori, at influence.

Chen (2024) publishes in the Journal of Artificial Intelligence Research the scoping study that best names the affordances the market translates as “voice”. With 18 articles from 11 countries and 16 journals, it covers 15,081 children aged 2 to 8 between 2005 and 2023; 14 of the 18 texts are from 2020–2023. Thematic analysis extracts four affordances, among them AI as a tangible or intangible tool for interactive learning and information retrieval, and AI as an object that adapts and personalises. Status: field finding, not participation. Inference: interacting and personalising are exactly what a feedback button and a summarising model sell as “the child opined and the system took it into account”. Taking into account, in Lundy, is influence. Personalising a pathway is not influence: it is an adjustment of the system upon the child, not a shared decision. Su and Yang (2022) and Ljungcrantz (2026) saturate the growth: seventeen studies up to 2021 and thirty-nine in 2020–2024, with a peak in 2024 and a focus on ages 4 to 6. Status: growth finding, not equivalence. Inference: the 2024 surge does not authorise translating “more AI that talks to the child” as “more participation”. It authorises asking what the AI does: whether it serves the listening of a responsible adult or captures an input to exhibit a summary.

The third artefact —the summarising model— has, in this corpus, no kindergarten trial that measures it as participation. That is inference, marked as such. A system that transcribes and paraphrases “the group’s opinion” inherits the limits of Kim et al. (2022) and of Xu, Branham et al. (2021), and the misunderstanding Lundy et al. (2024) name: treating the child as an expert whose output suffices. The Committee on the Rights of the Child (2021) requires due weight and that digital participation not become monitoring. Miao and Holmes (2023) exclude under-13s. Inference: a CENDI that delivers “the summary of what the children opined” has not made them participate. It has made a product. Participation is verified in whether that perspective carried weight, with responsible adults, in a decision. It is not verified in the existence of the summary.

5. Case 2. What the kindergarten does when there is participation: situated listening and agency in interaction

Hicban, Theobald and Mascadri (2024) publish in The International Journal of Children’s Rights the instrument that best names, in this corpus, participation as a fact of interaction, not of capture. Listening to Voices in Interaction for Participation (LViIP) combines the space–voice–audience–influence model with conversation analysis to identify how the enactment of the right unfolds turn by turn and how adults mobilise or gatekeep the opportunity for the child to have a say. A worked example shows practices that co-construct the right in early childhood education. The doctoral thesis from which the instrument derives records 75 hours of video of educator–child and peer interactions, in a Queensland setting, with 15 educators and 86 children aged 15 months to 5 years (Hicban, 2024). Status of the evidence. Empirical and methodological process finding: participation is seen in the sequence, not in a log. It is not an AI finding, nor a finding that an avatar reproduces that craft. The context is an Australian setting; it is not generalised to a Latin American CENDI. Pedagogical inference, marked as such: this is the object an early childhood setting may properly call participation. It is agency and situated listening. A button does not give audience. An avatar is not a duty-bearer. A summarising model does not co-construct the next turn.

Chicken and Tyrie (2023) saturate the portrait with three teachers of 3- to 7-year-olds in Wales. Observation, interviews and documentary analysis show that voice is bounded by place, time and adult expectations; children often take back control in subtle ways. Status: N = 3, not AI. Inference: a button that standardises “feedback” erases precisely that micro-agency. Waters-Davies, Murphy, Chicken, Tyrie and Clement (2024) collect, in a bilingual survey of teachers of 3- to 7-year-olds in Wales, that the construction of the capable child operates as a threshold. Status: perceptions, not classroom observation. Inference: a kindergarten does not implement participation by buying an interface. An avatar that “consults” can coexist with a construction of the incapable child: the system asks; the adult does not share the decision.

Barros et al. (2024) add twenty-five teachers and twenty-five ECEC coordinators in Belgium, Greece, Poland and Portugal, referred as implementers of participatory practices. Focus groups, inspired by Lundy’s model, bring forth the four dimensions: participatory space, children’s voice, audience for children’s perspectives, and influence. Status: perceptions of adults already said to be participatory, not of children and not of AI. Inference: when early childhood staff name participation, they name the four dimensions, not a channel. Correia et al. (2022) had illustrated those dimensions with examples from the same four countries and underlined the need for organisational support. Holla Sivertsen and Bjørgen (2025) hear sixteen five-year-olds in two Norwegian institutions. They perceive circle time as a passive position —they do not feel heard or influential— and the outdoors as space, voice, audience and influence. Status: N = 16, not AI. This article does not convert it into a study of free play: the object retained is the position of influence. Inference: a feedback button in circle time can aggravate the passivity children already name.

6. Case 3. Influencing with responsible adults is not summarising an opinion

Barblett et al. (2024) publish in the International Journal of Early Years Education the case that best illustrates influence when it is placed on the right side of the craft. In the update of the Australian early years learning framework (from birth to age 5), a three-stage project over fifteen months sought children’s perspectives at each stage. Three hundred and forty-six children took part (105, 92 and 149). Familiar educators applied dialogic drawing, talking circles and photo elicitation, with children’s assent. The analysis showed the importance of relationships with educators, with friends, and between educators and families. In February 2023 the framework updates were published. A companion text records a four-year-old’s phrase: “We’re not useless, we know stuff!” (Barblett, Bobongie-Harris, Cartmel, Hadley, Harrison, Irvine, & Lavina, 2023). Status of the evidence. Empirical finding of a policy process: children’s perspectives entered, mediated by responsible adults and by situated methods, a national framework. It is not an AI finding, nor a finding that a summarising model produces the same effect. Pedagogical inference, marked as such: this is the only place in which “gathering voice” approaches participation without contradicting the framework. There is audience —a team that listens and analyses— and influence —the framework is updated. An automated summary of “what they opined this week” does not update a framework or move a circle time. It extracts a product.

Hildebrandt, Macha, Lonnemann and Hildebrandt (2026) saturate the institutional portrait. In three German centres implementing the Kita-Beirat —an ECEC council introduced in 2021 in Rhineland-Palatinate, with a legal duty to consider children’s perspectives in conceptual decisions— they combine participant observation, group interviews and child-led methods inspired by the Mosaic Approach. Seventy-eight children aged 3 to 6 take part. The analysis shows that interaction quality —and with it agency— emerges from three dimensions: power dynamics, relationships and space–time. The council policy is valuable, but its effectiveness depends on staff being able to inscribe relational and reflexive pedagogies in daily practice. Status: ethnography of three centres, not AI and not a finding that a formal body suffices. Inference: a children’s council without situated listening is the institutional analogue of the button. An avatar that “consults” without power, relationship and classroom time moving is the technical analogue of the empty council. NAEYC (2022) requires purpose. OECD (2021, 2025) situates the engine in process interactions and in staff practices. Inference: evidence of participation is verified in whether the perspective carried weight, with responsible adults, in a decision. It is not verified in the artefact that captured it.

The distribution of roles is confirmed by contrast. Xu et al. (2022) elicit vocalisations. Kim et al. (2022) measure a device that responds wrongly half the time. Lundy et al. (2024) refuse to treat the child as an expert whose output suffices. Inference: AI enters early childhood participation, if at all, on the side of the adult who listens and shares the decision, subject to pedagogical validation. It does not enter as the child’s interlocutor or as notary of the child’s opinion. Miao and Holmes (2023) exclude under-13s. UNICEF (2021) requires the best interests of the child and participation as a lens, not as extraction. Annex III of Regulation (EU) 2024/1689 includes systems that determine access, admission, assignment, assessment or level (European Union, 2024). Inference: a system that, in a CENDI, summarises “the child’s voice” in order to decide groupings or pathways is not a participatory innovation. In that territory it is a high-risk practice if it determines assignment or level. Outside that territory, it remains a category error.

7. Inferential framework: four tests for affirming that there is participation, not an input

The framework that follows is this article’s pedagogical inference, anchored in the cases and in the verified instruments. It is not a new international standard. It distinguishes four tests. If a kindergarten, preschool, CENDI or infant school does not pass them, it cannot declare that a feedback button, an avatar that consults the child or a model that summarises the child’s opinion constitute child participation.

7.1. Test of agency, not of input. Ward and Lundy (2024) situate voice in multimodal and relational modes, not in a channel. Hicban, Theobald and Mascadri (2024) show turn-by-turn co-construction. Chicken and Tyrie (2023) find micro-agency that takes back control when practices constrain it. Xu et al. (2022) elicit story vocalisations. Chen (2024) maps interaction and personalisation. Inference: evidence of participation is verified in whether the child can influence, resist, propose and be taken seriously in the activity. If the “evidence” that the setting makes children participate is a click log, a turn with an avatar or a summary, the setting has made input, not agency.

7.2. Test of situated listening, not of automated consultation. Barros et al. (2024) hear adults who name audience. Holla Sivertsen and Bjørgen (2025) hear children who do not feel heard in circle time. Kim et al. (2022) measure an assistant that responds appropriately only half the time. Xu, Branham et al. (2021) find 143 apps that do not sustain open-ended dialogue. Miao and Holmes (2023) exclude under-13s as independent interlocutors. Inference: the 3–6-year-old is not the user of a model that “consults them”. The child is the subject of a listening that a duty-bearing adult exercises in situ. If AI enters, it enters as a workshop or as support for that listening, subject to pedagogical validation. It does not enter as consulting avatar or as the child’s stenographer.

7.3. Test of shared decision-making with responsible adults, not of the opinion summary. Barblett et al. (2024) show children’s perspectives that carry weight, mediated by adults, in a national framework. Hildebrandt et al. (2026) show that a formal council does not suffice without power, relationship and time. Lundy et al. (2024) reject rights inflation and the child as sole expert. Correia et al. (2022) require the four dimensions together. Inference: a summary of “what they opined” is not a shared decision. Shared decision-making requires that the responsible adult give due weight and account for what moved. In ages 3–6, that is played out in the next move of the setting’s life —changing a material, reopening a gathering, holding a proposal, refusing a demand with reasons— not in a generated paragraph.

7.4. Test of rights and of high risk, not of the product dashboard. The Committee on the Rights of the Child (2021) requires due weight and forbids digital participation from translating into undue monitoring. UNESCO (2021) requires human oversight. UNICEF (2021) requires the best interests of the child. Annex III of Regulation (EU) 2024/1689 includes access, admission, assignment, assessment and educational level (European Union, 2024). The European Commission (2022) and the U.S. Department of Education (2023) require that the teacher not be replaced. Inference: an early childhood setting cannot treat the child as the object of an automated verdict of “their opinion”. In European territory, a system that assigns or assesses with AI is, save for the regulation’s safeguards, a high-risk practice. Outside that territory, it remains a category error: participation is not fulfilled by capturing better. It is fulfilled by exercising agency, situated listening and shared decision-making with responsible adults.

The framework admits co-construction in interaction, the four dimensions, and processes in which perspective carries weight with responsible adults. It rejects declaring participation by a button, an avatar or a summary (Xu et al., 2022; Kim et al., 2022; Chen, 2024; Miao & Holmes, 2023).

8. Discussion

Three tensions organise the discussion. The first is between capturing and participating. It is a finding that an agent can elicit story vocalisations (Xu et al., 2022); that an assistant responds appropriately only half the time (Kim et al., 2022); that 143 voice apps do not sustain open-ended dialogue (Xu, Branham et al., 2021); and that the field of AI in ECE grew to 39 studies in 2020–2024 (Ljungcrantz, 2026; Su & Yang, 2022; Chen, 2024). It is framework that participation requires space, voice, audience and influence, with duty-bearing adults (Ward & Lundy, 2024; Correia et al., 2022; Lundy et al., 2024). It is not a finding that a button, an avatar or a summary produce the craft that Hicban, Theobald and Mascadri (2024), Chicken and Tyrie (2023) and Barblett et al. (2024) observe. The three artefacts measure what engineering knows how to extract and declare what only situated listening would authorise.

The second is between the child as a data source and the child as a rights-holder. Chen (2024) maps personalisation. Xu et al. (2022) deliver story comprehension. Kim et al. (2022) collect children who seek a relationship with the device. Holla Sivertsen and Bjørgen (2025) collect children who do not feel heard in circle time. Inference: insisting that the child “have their feedback channel” while the adult stops listening in situ is an inverted pedagogy. The input blames the child for what the process did not interpret. OECD (2021) had anchored quality in interactions. The summarising model does the inverse operation: it extracts the child from the group and returns them as a paragraph.

The third is between the duty-bearing adult and the artificial interlocutor. Barblett et al. (2024) and Hildebrandt et al. (2026) show adults who mediated children’s perspectives towards decisions. Miao and Holmes (2023) exclude under-13s. Inference: the only use of AI that does not contradict participation at ages 3–6 is the one that remains on the side of the adult who listens and shares the decision, subject to pedagogical validation. An avatar that consults the child because the system needs an input is not that use.

9. Limits

This review is narrative. It does not apply PRISMA or estimate combined effects. Xu et al. (2022) measure story comprehension with an agent, not participation in kindergarten decisions; transfer to the thesis is inference. Kim et al. (2022) cover ages 5 to 10, not isolated to 3–6. Xu, Branham et al. (2021) evaluate voice apps, not early childhood settings. Chen (2024) maps AI affordances in ECE ages 2–8; Su and Yang (2022) cover 1995–2021; Ljungcrantz (2026), 2020–2024. Hicban, Theobald and Mascadri (2024) are an instrument; the densely observed N is in the thesis (Hicban, 2024). Chicken and Tyrie (2023) are three teachers. Waters-Davies et al. (2024) and Barros et al. (2024) are perceptions. Holla Sivertsen and Bjørgen (2025) are sixteen children; they are not used as a free-play article. Barblett et al. (2023, 2024) are a policy process, not an AI trial. Hildebrandt et al. (2026) are three centres. The texts by Lundy, Correia, UNESCO, UNICEF, the Committee, OECD and the European Union are framework. No Latin American trials of participation and AI in kindergarten were located that measure agency and shared decision-making against button, avatar or summary. The inferences in section 7 are hypotheses of pedagogical category, not implementation evidence.

10. Conclusions

A feedback button, an avatar that consults the child or a model that summarises the child’s opinion do not constitute child participation in an early childhood setting. The verified evidence does not authorise that declaration. An agent can elicit story vocalisations; that is reading dialogue, not participation (Xu et al., 2022). An assistant responds appropriately only half the time (Kim et al., 2022). One hundred and forty-three voice apps do not sustain open-ended dialogue (Xu, Branham et al., 2021). The reviews saturate the growth of the AI-in-ECE field without equating it to agency or influence (Chen, 2024; Su & Yang, 2022; Ljungcrantz, 2026). By contrast, when there is participation in early childhood, there is an adult who listens in situ and shares the decision: turn-by-turn co-construction (Hicban, Theobald, & Mascadri, 2024; Hicban, 2024), Welsh classrooms (Chicken & Tyrie, 2023; Waters-Davies et al., 2024), European staff who name the four dimensions (Barros et al., 2024; Correia et al., 2022), children who distinguish passive circle time from influence (Holla Sivertsen & Bjørgen, 2025), 346 children whose perspectives carried weight in a national framework (Barblett et al., 2024, 2023) and a centre council that does not suffice without power, relationship and time (Hildebrandt et al., 2026). Current law requires space, voice, audience and influence, due weight, human oversight, the best interests of the child, an age threshold for independent conversations with generative platforms and, in the European Union, treating as high-risk the AI that determines access, admission, assignment, assessment or educational level (Ward & Lundy, 2024; Lundy et al., 2024; Committee on the Rights of the Child, 2021; UNESCO, 2021; UNICEF, 2021; Miao & Holmes, 2023; European Union, 2024; European Commission, 2022).

Where the sources do not measure a kindergarten, this article does not affirm it. Where they measure a button, an avatar or a summary, it does not translate them into participation. Accompanying children aged three to six with participation is to exercise agency, situated listening and shared decision-making with responsible adults. Anything else is an input captured by an interface. It is not child participation, and it must not be presented as what it is not.

Laboratorio Editorial de NEXTECH.IA / Ingeniero Mitre.

References

  1. Barblett, L., Bobongie-Harris, F., Cartmel, J., Hadley, F., Harrison, L., Irvine, S., y Lavina, L. (2023). “We’re not useless, we know stuff!” Gathering children’s voices to inform policy. Australasian Journal of Early Childhood, 48(2), 134–147. https://doi.org/10.1177/18369391221139912
  2. Barblett, L., Cartmel, J., Lavina, L., Hadley, F., Irvine, S., Harrison, L. J., y Bobongie-Harris, F. (2024). Prioritising children’s participation in research: Including children’s voices in updating a national early learning framework. International Journal of Early Years Education, 32(4), 852–870. https://doi.org/10.1080/09669760.2024.2405535
  3. Barros, S., Coelho, V., Wysłowska, O., Penderi, E., Taelman, H., Araújo, S. B., Correia, N., Markowska-Manista, U., Petrogiannis, K., Boderé, A., Pessanha, M., Guimarães, C., y Aguiar, C. (2024). A focus group study on participatory practices in early childhood education and care across four European countries. Early Education and Development, 35(6), 1292–1315. https://doi.org/10.1080/10409289.2024.2360868
  4. 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
  5. Chicken, S., y Tyrie, J. (2023). Can you hear me? Problematising the enactment of UNCRC Article 12 in Welsh early years classrooms: Exploring the challenges of “children’s voice”. The International Journal of Children’s Rights, 31(2), 301–325. https://doi.org/10.1163/15718182-31010001
  6. Comité de los Derechos del Niño. (2021). Observación general núm. 25 (2021) relativa a los derechos de los niños en relación con el entorno digital (CRC/C/GC/25). Naciones Unidas. https://digitallibrary.un.org/record/3906061
  7. Correia, N., Aguiar, C., Taelman, H., Boderé, A., Petrogiannis, K., Penderi, E., Karachristos, C., Stavropoulos, E., Wyslowska, O., Markowska-Manista, U., Mesquita, C., Ribeiro, L., Barros, S., Araújo, S., Pessanha, M., Guimarães, C., Coelho, V., Correia, I., Magalhães, E., Figueiredo, V., y Almeida, F. (2022). Children’s right to participate: The Lundy model applied to early childhood education and care. The International Journal of Children’s Rights, 30(2), 378–405. https://doi.org/10.1163/15718182-30020010
  8. European Commission. (2022). Ethical guidelines on the use of artificial intelligence (AI) and data in teaching and learning for educators. Publications Office of the European Union. https://doi.org/10.2766/153756
  9. Hicban, C. (2024). Hand-in-hand: The co-construction of children’s participation rights in an early childhood education setting [Tesis doctoral, Queensland University of Technology]. https://doi.org/10.5204/thesis.eprints.246679
  10. Hicban, C., Theobald, M., y Mascadri, J. (2024). A new approach for investigating children’s participation rights in early childhood education: Listening to Voices in Interaction for Participation (LViIP). The International Journal of Children’s Rights, 32(4), 910–944. https://doi.org/10.1163/15718182-32040003
  11. Hildebrandt, A., Macha, K., Lonnemann, J., y Hildebrandt, F. (2026). Beyond voice: High-quality interactions and children’s agency in early childhood education. Early Childhood Education Journal. Publicación anticipada en línea. https://doi.org/10.1007/s10643-026-02237-1
  12. Holla Sivertsen, A., y Bjørgen, K. (2025). Children’s participation positions in various activities within Norwegian early childhood education and care (ECEC) institutions. Early Child Development and Care, 195(4), 299–312. https://doi.org/10.1080/03004430.2025.2478889
  13. Kim, M. K., Druga, S., Esmaeili, S., Woodward, J., Shaw, A., Jain, A., Langham, J., Hollingshead, K., Lovato, S. B., Beneteau, E., Ruiz, J., Anthony, L., y Hiniker, A. (2022). Examining voice assistants in the context of children’s speech. International Journal of Child-Computer Interaction, 34, 100540. https://doi.org/10.1016/j.ijcci.2022.100540
  14. Ljungcrantz, L. (2026). The interaction of AI and early childhood education. A state-of-the-art review 2020–2024. Early Childhood Education Journal, 54, 3565–3581. https://doi.org/10.1007/s10643-025-02079-3
  15. Lundy, L., Murray, C., Smith, K., y Ward, C. (2024). Young children’s right to be heard on the quality of their education: Addressing potential misunderstandings in the context of early childhood education. British Educational Research Journal. Publicación anticipada en línea. https://doi.org/10.1002/berj.3968
  16. Miao, F., y Holmes, W. (2023). Guidance for generative AI in education and research. UNESCO. https://doi.org/10.54675/EWZM9535
  17. 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
  18. OECD. (2021). Starting Strong VI: Supporting meaningful interactions in early childhood education and care. OECD Publishing. https://doi.org/10.1787/f47a06ae-en
  19. OECD. (2023). Empowering young children in the digital age (Starting Strong). OECD Publishing. https://doi.org/10.1787/50967622-en
  20. OECD. (2025). Results from TALIS Starting Strong 2024: Strengthening early childhood education and care. OECD Publishing. https://doi.org/10.1787/20af08c0-en
  21. Su, J., y Yang, W. (2022). Artificial intelligence in early childhood education: A scoping review. Computers and Education: Artificial Intelligence, 3, 100049. https://doi.org/10.1016/j.caeai.2022.100049
  22. UNESCO. (2021). Recommendation on the ethics of artificial intelligence. UNESCO. https://unesdoc.unesco.org/ark:/48223/pf0000381137
  23. UNICEF. (2021). Policy guidance on AI for children 2.0. UNICEF Innocenti. https://www.unicef.org/innocenti/reports/policy-guidance-ai-children
  24. Unión Europea. (2024). Reglamento (UE) 2024/1689 del Parlamento Europeo y del Consejo, de 13 de junio de 2024, por el que se establecen normas armonizadas en materia de inteligencia artificial. Diario Oficial de la Unión Europea, L. https://eur-lex.europa.eu/eli/reg/2024/1689/oj
  25. U.S. Department of Education, Office of Educational Technology. (2023). Artificial intelligence and the future of teaching and learning: Insights and recommendations. U.S. Department of Education. https://www.ed.gov/sites/ed/files/documents/ai-report/ai-report.pdf
  26. Ward, C., y Lundy, L. (2024). Space, voice, audience, and influence: The Lundy model and early childhood. En L. Mahony, S. McLeod, A. Salamon y J. Dwyer (Eds.), Early childhood voices: Children, families, professionals (pp. 17–28). Springer. https://doi.org/10.1007/978-3-031-56484-0_2
  27. Waters-Davies, J., Murphy, A., Chicken, S., Tyrie, J., y Clement, J. (2024). Constructing child participation in early years classrooms: An exploration from Wales. Children & Society, 38(5), 1824–1841. https://doi.org/10.1111/chso.12848
  28. Xu, Y., Aubele, J., Vigil, V., Bustamante, A. S., Kim, Y.-S., y Warschauer, M. (2022). Dialogue with a conversational agent promotes children’s story comprehension via enhancing engagement. Child Development, 93(2), e149–e167. https://doi.org/10.1111/cdev.13708
  29. Xu, Y., Branham, S. M., Deng, X., Collins, P., y Warschauer, M. (2021). Are current voice interfaces designed to support children’s language development? En Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems (artículo 633). ACM. https://doi.org/10.1145/3411764.3445271