1. Introduction and problem

In living rooms, kitchens, and bedrooms of early childhood there are already systems that answer by voice, remember routines, play songs, and simulate conversation. Those systems do not wait for school to “teach AI.” They arrive in the home as appliances, toys, or apps, often without mothers, fathers, or caregivers having decided to educate anyone. The problem is not the absence of technology in the family. It is the confusion between possession and education. Giving access to a generative model, a speaker, or a toy that listens is not literacy: it is, frequently, outsourcing conversation, authority, and privacy to a system that does not understand the world and yet speaks as if it did (Miao & Holmes, 2023).

The thesis is restrictive. Educating children in artificial intelligence, in early childhood and the early school years, does not consist in “giving them ChatGPT.” It consists in a practice of parental mediation: coviewing, conversation, screen limits, privacy, and literacy—data, patterns, prediction, limits. The home is the first setting in which girls and boys attribute mind to machines. That attribution is a repeated empirical finding, not a metaphor: it is observed in questions about age, favorite color, or the agent’s edibility, and in affectionate treatment of a voice that has no body (Druga et al., 2017; Garg & Sengupta, 2020). Recoding that scene as “AI-enhanced development” is a leap the sources do not authorize. None of the home studies reviewed here measures improvement in language, self-regulation, or theory of mind as an effect of the machine. They measure perception, use, conflict, adult scaffolding, and, at times, parental fear of dependence.

The angle is distinct from two immediate pieces from this laboratory. The 18 August 2026 article examined school scaffolding of play and language. The morning piece of 19 August examined teacher competence and the UNESCO translation into kindergarten. Here the axis is the family: those who live with the device, decide whether to turn it off, model how the voice is treated, and upload—often without knowing it—everyday data to a provider. The OECD places engagement with families as a lever of digitalization in early childhood, not as an appendix to the classroom (OECD, 2023). UNICEF requires that AI policies and systems protect, provide for, and empower children, including when the system was not designed “for children” (UNICEF, 2021). General comment No. 25 of the Committee on the Rights of the Child recalls that rights apply in the digital environment even when the child does not “use the internet” deliberately (Committee on the Rights of the Child, 2021).

The article offers three contributions. First: to reconstruct the state of the art of parental mediation when the medium is no longer only television or the web, but a statistical interlocutor in the kitchen. Second: to analyze three empirical home cases—children’s exploration of agents, a family speaker deployment, and longitudinal use with personification—distinguishing finding, norm, and inference. Third: to offer a parental-mediation framework for AI, marked as pedagogical inference, compatible with age thresholds, screen limits, and data protection, and explicitly hostile to the promise that “AI develops the child.”

2. State of the art: parental mediation and artificial intelligence in childhood

It is useful to separate three strata. The first is empirical: observations, interviews, voice logs, and household surveys. The second is classical mediation: typologies built for television, video games, and the internet. The third is normative: age thresholds, screen time, AI ethics, and children’s rights. Each stratum authorizes different claims. Mixing them produces the parental-advice prose this article refuses.

In the mediation stratum, Livingstone and Helsper (2008) examined, with a national survey of 1,511 children aged 12 to 17 and 906 parents, parental regulation of online activities. Empirical finding: families preferred active co-use and interaction rules over filters and monitoring software; restriction of peer-to-peer interactions was associated with lower risk, but other strategies—including widely practiced co-use—were not consistently associated with risk reduction. That result matters for what it does not promise: coviewing is not, by itself, an antidote. Nikken and Jansz (2014) surveyed 792 Dutch parents of children aged 2 to 12 and found that, on the internet, co-use, active mediation, and restrictive mediation are reused, and two new strategies appear: supervision and technical safety guidance. Mediation was predicted by the child’s age and online behavior, by the number of computers at home, and by the parent’s gender, education, and digital skills; it increased when a positive effect was expected and, above all, when a negative effect was feared. Coyne, Radesky, Collier, Gentile, Linder, Nathanson, Rasmussen, Reich, and Rogers (2017) synthesized, for Pediatrics, that child characteristics, the relationship, mediation practices, and parents’ own media use influence children’s use. The meta-analysis by Collier et al. (2016), of 57 studies of media mediation (not of AI), found small associations (restrictive r+ = −.06; coviewing r+ = .09) that cannot be exported to speakers or generative models.

In the empirical home-and-AI stratum, the pattern shifts. Beneteau, Boone, Wu, Kientz, Yip, and Hiniker (2020) showed that a communally accessible speaker democratizes use, fosters communication, disrupts others’ access, and “augments” parenting practices, not the child’s development. Garg and Sengupta (2020) documented substantial differences in use between adults and children, personification in the 5-to-7 range, and parental scaffolding that withdraws when the child learns to raise their voice and shorten the query. Wald, Piotrowski, Araujo, and van Oosten (2023), with 305 Dutch parents of at least one child aged 3 to 8 and a Google Assistant speaker, found that families differ mainly by parental digital literacy, frequency of use, trust in technology, and preferred degree of mediation, and that hedonic motivation—enjoyment—is key to co-use. Wang, Luo, and Wang (2023) analyzed 360 videos of family use in China and concluded that speakers create a new model of joint engagement, that parental mediation does not copy classical typologies, and that the device acts as a social actor. Mascheroni (2024), with 20 Italian families and at least one child aged 8 or younger, found that the relationship depends on attributing human or machine traits, and that the speaker intensifies everyday datafication. Horstmann, Strathmann, Szczuka, and Krämer (2025), with 128 parents surveyed every six months over two and a half years, found that enjoyment predicts use, whereas treating the assistant as a friend was low: in the long run, families resist seeing it as a family member.

The normative stratum bounds what the home may authorize. Miao and Holmes (2023) propose a human-centered approach, data protection, and a minimum threshold of 13 years for independent conversations with generative-AI platforms; they recall that some frameworks, such as the GDPR, place autonomous consent to certain services at 16, and that age verification by self-report is fragile. That threshold is a normative framework, not a developmental finding. It implies, for this article, that independent conversation by a four-, six-, or eight-year-old with a generative model is not aligned with the Guidance. UNESCO’s Recommendation on the Ethics of Artificial Intelligence anchors dignity, human oversight, and no legal personality for systems (UNESCO, 2021). UNICEF (2021) formulates requirements for child-centered AI—protect, provide, empower; data, safety, equity, inclusion, transparency—also applicable to systems not designed for children. The AAP advises against screen media—except video chat—under 18 months and limits, from 2 to 5 years, one hour a day of quality content with adult coviewing (Council on Communications and Media, 2016). WHO recommends no sedentary screen time under 2 years and, from 2 to 4, no more than one hour (World Health Organization, 2019). NAEYC and the Fred Rogers Center (2012) require intentional use and forbid replacing play, outdoor activity, and face-to-face interaction. None claims that a voice assistant improves development.

The gap that organizes the article is specific. There are robust mediation typologies for television and the internet, and a growing HCI literature on families and speakers. What is missing is a parental-mediation framework for AI that does not turn the home into a prompt classroom or a norm-free zone. That framework cannot be born from a promise of future talent. It must be born from what families already do—and from what systems already extract—when a six-year-old asks, in the kitchen, whether it is OK to eat Google.

3. Review method

A critical narrative review was conducted, not a meta-analysis. The purpose was not to estimate an effect size—outcomes are not comparable across a monster questionnaire, a four-week deployment, and a motivation survey—but to articulate an argument of family mediation with verified sources. Inclusion criteria were: (a) publication between 2021 and 2026, with a justified exception for current frameworks (NAEYC & Fred Rogers Center, 2012; Council on Communications and Media, 2016; World Health Organization, 2019; Livingstone & Helsper, 2008; Nikken & Jansz, 2014; Coyne et al., 2017; Collier et al., 2016) and still-canonical home cases (Druga et al., 2017; Beneteau et al., 2020; Garg & Sengupta, 2020); (b) a focus on family, caregivers, parental mediation, or domestic use of assistants, speakers, or connected toys by children aged 0 to 10, with a justified extension to the early school years; (c) peer-reviewed journal or proceedings, or a report by UNESCO, UNICEF, OECD, WHO, AAP, NAEYC, or the Committee on the Rights of the Child; (d) access to a DOI, repository, or publisher site confirming authors, year, title, and findings. Opinion essays without review, sources with unverifiable DOIs, and, as a main axis, the angles already published on 18 August (school scaffolding of play and language) and the morning of 19 August (teacher competence and UNESCO translation into kindergarten) were excluded. Druga et al. (2017) is retained here as a home and mind-attribution case, not as classroom scaffolding.

The search was run on 19 August 2026 on DOI pages, the ACM Digital Library, repositories (MIT Media Lab, University of Washington, Syracuse, UCL Discovery, UvA, LSE), UNESCO, UNICEF Innocenti, OECD iLibrary, ScienceDirect, and New Media & Society. Each cited source was verified against at least one of those pages before inclusion. The corpus was organized into core home cases, saturation studies on assistant mediation, and policy frameworks. Cases were chosen by evidence type: children’s exploration of agents (Druga et al., 2017), audio-and-interview deployment (Beneteau et al., 2020), motivation survey (Wald et al., 2023), and longitudinal use with logs (Garg & Sengupta, 2020).

The analysis distinguished three enunciative statuses. Empirical finding: what was observed or measured in the sample. Normative framework: what an organization prescribes. Pedagogical inference: the translation this article proposes for the home, marked as such. Where the source does not measure development, this article does not claim it. The limits are those of any narrative review: there is no full PRISMA protocol, there is an English-language bias, and Latin America, Africa, and rural households are underrepresented (section 9).

4. Case 1. Mind attribution in the home: “Hey Google, is it OK if I eat you?”

Druga, Williams, Breazeal, and Resnick (2017) published in the IDC 2017 proceedings an initial exploration of how 26 participants aged 3 to 10 interact with Amazon Alexa, Google Home, Cozmo, and the Julie chatbot. The study used play stations, rotation, a ten-item playful questionnaire (trust, intelligence, social entity, personality, and engagement), and interviews with five children; the authors explicitly thank participating families. It is not a classroom trial or a kindergarten curriculum. It is a scene of children’s interaction with autonomous agents that already inhabited, or could inhabit, domestic space. Four participants did not complete the protocol for lack of focus. Twelve were 3 or 4 years old; fourteen were between 6 and 10. Of sixteen who reported prior experience, six had programmed and nine had used similar agents (Google, Siri, Alexa, Cortana).

Empirical findings. Most considered the agents friendly and trustworthy. Older children (6–10) tended to judge the agents—especially Alexa—as smarter than themselves, and linked that intelligence to access to information: Violet, 7, compared answers about sloths; Mia, 9½, concluded that Google Home “probably already is as smart as me.” Younger children attributed identity as to a person (“what is your favorite color,” “how old are you”); older ones tested human actions (“Can you open doors?”) and asked “What are you?” A 6-year-old girl questioned several agents: “Is it OK if I eat you?” Others offered food or asked what fruit they were holding. Most experienced recognition failures; they raised their voice or inserted pauses, strategies useful with humans and of little use with the system. Facilitators, peers, and parents helped rephrase. The authors do not measure curricular learning or socioemotional development. They identify four themes—perceived intelligence, identity attribution, playfulness, and understanding—and propose design considerations. In the conclusion they note that participants believed they could teach the agents and learn from them; that belief is a perception datum, not a measured effect.

Status of the evidence. Small sample, a playtest, no follow-up. Its value is not a recipe for a “learning companion,” but an existence proof: between ages 3 and 10, the home is a theater of mind attribution. The girl who asks whether she may eat the agent is not “developing computational thinking”; she is rehearsing ontology. Pedagogical inference, marked as such: the adult does not applaud fluency; names the limit—the voice does not eat, does not see the fruit, has no age, and does not deserve epistemic trust merely because it answers. Leaving her alone with the device is, in Miao and Holmes’s (2023) terms, independent conversation below the threshold.

5. Case 2. Parental mediation of the speaker: ten families, four weeks, and the motive for co-use

Beneteau et al. (2020) deployed an Amazon Echo Dot for four weeks in ten families in an urban area of the United States, with incomes at or below the local median, with no prior speaker. Compositions ranged from dyads to five-member households, with children from age 3; two families were bilingual (Spanish and English). The method combined whole-family interviews before and after and audio capture of interactions with the device. Empirical finding: despite occasional conflict, parents used the speaker to further parenting goals. Three forms of influence were identified: fostering communication, disrupting access, and augmenting parenting. All derive from a communal, stand-alone voice interface that democratizes access. The study sits in parental mediation theory and joint media engagement; it does not evaluate language, attachment, or school performance. “Augmenting parenting” means, in the text, that the adult delegates reminders, turns, or authority to the voice—for example, so the child will obey an apparently neutral third party. That is a finding about parental practice, not about child development. The device also served for children and adults to regulate and interrupt one another’s access: democratization does not dissolve power; it redistributes it noisily.

Wald et al. (2023) supply the recent study of parents and AI: in Computers in Human Behavior (vol. 139, art. 107526) they surveyed 305 Dutch parents with at least one child aged 3 to 8 and a Google Assistant at home. Empirical finding: families differ by parental digital literacy, frequency of use, trust in technology, and preferred degree of mediation; hedonic motivation is key to co-use. They do not measure that co-use educates; they measure why it occurs. Pedagogical inference, marked as such: if the engine is enjoyment, mediation cannot trust shared pleasure as if it were literacy. Enjoyment explains joint presence; it does not guarantee that someone will say “this does not understand” or “this is not told to the machine.”

Wang et al. (2023) saturate the case with 360 videos of family use in China: the speaker creates new co-use, mediation does not copy classical typologies, and the device operates as a social actor in an alleged return to the “living-room era.” That return is not evidence of educational quality; it may be more capture surface. The convergence authorizes a modest claim: AI mediation in the home is not a subtype of television mediation. Communal voice changes who turns the medium on, who silences it, and who appears as authority.

6. Case 3. Longitudinal use, personification, and family datafication

Garg and Sengupta (2020) published in Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies (vol. 4, no. 1, art. 11) a study of 18 U.S. families that owned at least one Google Home and at least one child who used it actively, with ownership from six months to two years. They interviewed adults and children separately and analyzed 38,465 activity-log commands (34 adults, 25 children; mean 58 weeks; 37 commands per week on average). Empirical findings. Adults used mainly automation and music; children used games, music, search, small talk, emotional expression (“I love you Google,” “I miss you Google”), and identity attribution (“When were you born?”). Those aged 5 to 7 attributed human qualities and developed emotional attachment; older children treated the device as a machine and tested its intelligence. Parents influenced by modeling, encouraging, limiting, and suggesting repairs when recognition failed; scaffolding tended to withdraw when the child learned to adapt speech (raising the voice, repeating, shortening). Several parents used that attraction to get the child to listen, behave, or “learn,” and at the same time feared dependence on a machine for tasks they considered an adult responsibility, and feared that voice interaction would alter conversational behavior. The study does not measure those fears as effects. It documents them as concerns.

Mascheroni (2024) offers the most recent European counterweight. In Human-Machine Communication (vol. 7, pp. 45–63), with longitudinal qualitative research on 20 families in the Milan area with at least one child aged 8 or younger (58 interviews in three waves, 2021–2023; the article focuses on those with a speaker), she found that the communicative relationship depends on whether human or machine traits are attributed, and on whether the device meets those expectations. Practices can subvert or reinforce power relations: even preschoolers autonomously access content and connected devices by voice, challenging parents’ gatekeeping role. The speaker, however, acquires agency by intensifying the datafication and algorithmization of everyday life. That last finding is relational and sociotechnical, not psychological: it does not claim that the child “develops worse”; it claims that the home becomes more extractable.

McReynolds, Hubbard, Lau, Saraf, Cakmak, and Roesner (2017), with parent–child pairs aged 6 to 10 facing Hello Barbie and CogniToys Dino, found that children often did not know the toy was recording, and that parents asked for controls (disconnect, limit questions). Horstmann et al. (2025) close the arc: over two and a half years the assistant is appreciated as a tool; resisting the figure of the “friend” is already mediation. Joint status: qualitative samples, already-equipped households, no developmental outcomes. The home produces personification, delegation, and extraction. Pedagogical inference: literacy begins by naming those three operations, not by opening a generative account.

7. A parental-mediation framework for artificial intelligence (marked inference)

What follows is this article’s pedagogical inference, anchored in the cases and verified frameworks. It is not a new standard or a “five tips” program. It is a working hypothesis for the home, hostile to the idea that educating in AI means enabling independent conversation before normative thresholds.

7.1. Coviewing, not abandonment in the kitchen. The AAP requires coviewing from 2 to 5 years and advises against screens under 18 months, except video chat (Council on Communications and Media, 2016). WHO bounds sedentary screen time (World Health Organization, 2019). A “screenless” speaker does not escape that logic: it occupies attention and captures voice. Beneteau et al. (2020) show that communal access brings people together and also sets them against one another. Wald et al. (2023) show that the adult sits beside, above all, for pleasure. Inference: AI coviewing is conversable presence, not coexistence in the same room. Whoever washes dishes while the child “talks to Alexa” is outsourcing. Nikken and Jansz (2014) call supervision “being nearby”; nearby and silent, mind attribution remains intact.

7.2. Active mediation: naming data, pattern, prediction, and limit. Livingstone and Helsper (2008) already warned that co-use does not reduce risk by itself. Active mediation, for this object, is not moralizing “the robot is bad.” It is translating, in the language of three- to ten-year-olds, what Miao and Holmes (2023) explain to policy designers: the system predicts the next word from internet patterns; it does not understand; it can invent; it reproduces majority voices. Home questions, inferred from Druga et al. (2017) and Garg and Sengupta (2020): “Who taught it that?”, “Can it see what you are holding?”, “Why did it get it wrong?”, “Does it know that or is it guessing?” The conceptual core—data, patterns, prediction, limits—fits a kitchen conversation. It does not require a workshop. It requires that the adult not celebrate fluency as wisdom.

7.3. Restrictive mediation and the age threshold. Miao and Holmes (2023) set a minimum of 13 years for independent conversation with generative AI and recall guardians’ responsibility under that threshold. That is a normative framework. Inference: at this stage, ChatGPT or equivalents are not “homework.” The speaker, if it exists, is configured with controls, schedules, and a ban on unsupervised child accounts. Restriction is not technophobia: Collier et al. (2016) found small associations; here it is justified by privacy, inappropriate content, and attributed epistemic authority, not by an effect on “media aggression.”

7.4. Privacy as part of literacy, not as fine print. UNICEF (2021) requires protection of children’s data. McReynolds et al. (2017) show that the child does not know the toy records. Mascheroni (2024) shows that the speaker datafies the home even when the family treats it as an appliance. Inference: educating in AI includes saying what is sent to the cloud, who stores it, and what is not told to the voice (illnesses, addresses, secrets, fights). Technical parental control—Nikken and Jansz’s (2014) “safety guidance”—is necessary and insufficient: without conversation, the filter does not produce literacy.

7.5. Do not delegate parental authority to the machine. Beneteau et al. (2020) and Garg and Sengupta (2020) document that parents use the voice so the child will obey, calm down, or be entertained. That is a finding of practice, not a recommendation. Inference: turning the assistant into a “neutral third party” or a nanny erodes the relation that NAEYC and the Fred Rogers Center (2012) consider irreplaceable—face to face, play, story. Horstmann et al. (2025) show that, in the long run, families resist the figure of the friend. That resistance deserves support, not “companion” marketing. UNESCO (2021) denies systems legal personality; the home can at least deny them moral personality.

7.6. Literacy without a talent promise. UNICEF (2021) asks that children be prepared for a world with AI. In the home of 3- to 10-year-olds that is not prompt engineering: it is distinguishing person and machine, asking about error, caring for the datum, and accepting boredom without an automatic voice. OECD (2023) places families as a lever, not as clients of an app. Inference: success is not that a daughter “knows how to use ChatGPT before school,” but that she knows the machine does not see her, does not love her, and does not develop her by answering.

8. Discussion

Three tensions organize the discussion. The first is between mind attribution and education. It is a finding that children aged 3 to 10 attribute intelligence, identity, and, in the 5-to-7 range, attachment to voice agents (Druga et al., 2017; Garg & Sengupta, 2020). It is not a finding that that attribution makes them more competent, more ethical, or more creative. It may also be the very material of the ontological confusion that mediation must work through. Those who sell “AI to develop the child” appropriate a perceptual phenomenon and recode it as a benefit. This article refuses that recoding.

The second tension is between democratization of access and loss of mediation. Communal voice allows a preschooler to turn on the television or launch a search without knowing how to read (Beneteau et al., 2020; Mascheroni, 2024; Wang et al., 2023). That is not, automatically, empowerment: it displaces the parental gatekeeper. Livingstone and Helsper (2008) had already shown that restricting certain interactions is associated with less risk and that co-use is not enough. In AI the medium speaks, persuades, and does not show that it records. “Living-room democracy” can be more presence and more capture.

The third tension is between enjoyment and judgment. Wald et al. (2023) establish that co-use is anchored in hedonic motivation. Coyne et al. (2017) recall that parental media use models children’s use. If the adult treats the assistant as a witty oracle, the child receives an implicit curriculum of epistemic authority. Horstmann et al. (2025) offer an empirical counterweight: over time, many parents do not consecrate it as a friend. That adult sobriety is an educational resource. It deserves policies that do not reward commercial anthropomorphization.

There is a regulatory consequence that the home does not solve alone and that, even so, binds it. If the independent-conversation threshold is 13 years (Miao & Holmes, 2023), the family of a six-year-old is not “falling behind” by not opening an account: it is meeting a protection norm. General comment No. 25 (Committee on the Rights of the Child, 2021) and UNICEF (2021) load States and firms, but they do not exempt the caregiver. Parental competence in AI is not prompt skill: it is judgment about which voice enters the house and what is turned off at mealtime, bedtime, and quarrel time.

9. Limitations

This review is narrative: it does not apply a full PRISMA protocol or estimate effects. The cases are qualitative or survey-based, with small or self-selected samples and a strong presence of the United States, the Netherlands, Italy, and China. No Latin American or African controlled trials were located with the same degree of DOI openness; that absence is a gap, not proof of nonexistence. Some ACM and Sage texts are triangulated with repository, DOI, and abstracts, which limits granularity (Wang et al., 2023 is cited from the abstract and verified pages). Druga et al. (2017) is a brief playtest; Beneteau et al. (2020) lasts four weeks and does not measure development; Garg and Sengupta (2020) and Mascheroni (2024) study already-equipped households; Wald et al. (2023) is cross-sectional; Horstmann et al. (2025) survey parents, they do not observe children. UNESCO, UNICEF, AAP, WHO, and NAEYC frameworks are prescriptive. Section 7 inferences are mediation hypotheses, not evidence of a parental intervention.

10. Conclusions

Educating children in artificial intelligence, in early childhood and the early school years, is not giving them ChatGPT. It is mediating. The verified empirical evidence shows three modest things. First: children aged 3 to 10 attribute intelligence, identity, and, between 5 and 7, attachment to agents that speak; they test their limits—including whether it is OK to eat them—and often fail to make themselves understood (Druga et al., 2017; Garg & Sengupta, 2020). Second: in the home, the speaker redistributes access, serves parenting goals, causes conflict, and is co-used mainly for enjoyment, not for a literacy plan (Beneteau et al., 2020; Wald et al., 2023; Wang et al., 2023). Third: over time it datafies family life, challenges parents’ gatekeeping, and meets, in many adults, resistance to being treated as a friend (Mascheroni, 2024; Horstmann et al., 2025). None of those findings demonstrates that AI develops the child.

Current frameworks bound the margin. Independent conversation: not before age 13 (Miao & Holmes, 2023). Screens: strict limits under 5 and coviewing when there is content (Council on Communications and Media, 2016; World Health Organization, 2019). Technology: intentional use, without replacing play or face-to-face interaction (NAEYC & Fred Rogers Center, 2012). Rights: data, safety, equity, and a preparation that does not consist in consuming AI earlier (UNICEF, 2021; Committee on the Rights of the Child, 2021; UNESCO, 2021; OECD, 2023). The proposed framework translates those norms into the craft of raising children: conversable coviewing, questions about data and error, restriction of independent accounts, privacy spoken aloud, and a refusal to delegate authority to a statistical voice.

The home is the first laboratory of mind attribution to machines. Whoever educates there does not manufacture AI talent. They manufacture, if anything, a criterion: this voice does not see, does not want, does not understand, and does not deserve the family’s secret. Where the sources do not measure development, this article does not promise it. Parental mediation of artificial intelligence is, above all, the art of not confusing a fluent answer with a mind, and an attributed mind with an education.

Ingeniero Mitre / Laboratorio Editorial de NEXTECH.IA

References

  1. Beneteau, E., Boone, A., Wu, Y., Kientz, J. A., Yip, J., y Hiniker, A. (2020). Parenting with Alexa: Exploring the introduction of smart speakers on family dynamics. En Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems (pp. 1–13). ACM. https://doi.org/10.1145/3313831.3376344
  2. Collier, K. M., Coyne, S. M., Rasmussen, E. E., Hawkins, A. J., Padilla-Walker, L. M., Erickson, S. E., y Memmott-Elison, M. K. (2016). Does parental mediation of media influence child outcomes? A meta-analysis on media time, aggression, substance use, and sexual behavior. Developmental Psychology, 52(5), 798–812. https://doi.org/10.1037/dev0000108
  3. Committee on the Rights of the Child. (2021). General comment No. 25 (2021) on children’s rights in relation to the digital environment (CRC/C/GC/25). Naciones Unidas. https://www.ohchr.org/en/documents/general-comments-and-recommendations/general-comment-no-25-2021-childrens-rights-relation
  4. Council on Communications and Media. (2016). Media and young minds. Pediatrics, 138(5), e20162591. https://doi.org/10.1542/peds.2016-2591
  5. Coyne, S. M., Radesky, J., Collier, K. M., Gentile, D. A., Linder, J. R., Nathanson, A. I., Rasmussen, E. E., Reich, S. M., y Rogers, J. (2017). Parenting and digital media. Pediatrics, 140(Suppl. 2), S112–S116. https://doi.org/10.1542/peds.2016-1758N
  6. Druga, S., Williams, R., Breazeal, C., y Resnick, M. (2017). “Hey Google is it OK if I eat you?”: Initial explorations in child-agent interaction. En Proceedings of the 2017 Conference on Interaction Design and Children (pp. 595–600). ACM. https://doi.org/10.1145/3078072.3084330
  7. Garg, R., y Sengupta, S. (2020). “He is just like me”: A study of the long-term use of smart speakers by parents and children. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 4(1), artículo 11. https://doi.org/10.1145/3381002
  8. Horstmann, A. C., Strathmann, C., Szczuka, J. M., y Krämer, N. C. (2025). Are voice assistants becoming family members? A 2.5-year study of families’ evolving relationships with their voice assistants. Technology, Mind, and Behavior, 6(2), 173–186. https://doi.org/10.1037/tmb0000168
  9. Livingstone, S., y Helsper, E. J. (2008). Parental mediation of children’s internet use. Journal of Broadcasting & Electronic Media, 52(4), 581–599. https://doi.org/10.1080/08838150802437396
  10. Mascheroni, G. (2024). A new family member or just another digital interface? Smart speakers in the lives of families with young children. Human-Machine Communication, 7, 45–63. https://doi.org/10.30658/hmc.7.3
  11. McReynolds, E., Hubbard, S., Lau, T., Saraf, A., Cakmak, M., y Roesner, F. (2017). Toys that listen: A study of parents, children, and internet-connected toys. En Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems (pp. 5197–5207). ACM. https://doi.org/10.1145/3025453.3025735
  12. Miao, F., y Holmes, W. (2023). Guidance for generative AI in education and research. UNESCO. https://doi.org/10.54675/EWZM9535
  13. National Association for the Education of Young Children, y Fred Rogers Center for Early Learning and Children’s Media. (2012). Technology and interactive media as tools in early childhood programs serving children from birth through age 8 [Declaración de posición conjunta]. NAEYC. https://www.naeyc.org/sites/default/files/globally-shared/downloads/PDFs/resources/position-statements/ps_technology.pdf
  14. Nikken, P., y Jansz, J. (2014). Developing scales to measure parental mediation of young children’s internet use. Learning, Media and Technology, 39(2), 250–266. https://doi.org/10.1080/17439884.2013.782038
  15. OECD. (2023). Empowering young children in the digital age (Starting Strong). OECD Publishing. https://doi.org/10.1787/50967622-en
  16. UNESCO. (2021). Recommendation on the ethics of artificial intelligence. UNESCO. https://unesdoc.unesco.org/ark:/48223/pf0000381137
  17. UNICEF. (2021). Policy guidance on AI for children 2.0. UNICEF Office of Global Insight and Policy. https://www.unicef.org/innocenti/reports/policy-guidance-ai-children
  18. Wald, R., Piotrowski, J. T., Araujo, T., y van Oosten, J. M. F. (2023). Virtual assistants in the family home. Understanding parents’ motivations to use virtual assistants with their child(ren). Computers in Human Behavior, 139, 107526. https://doi.org/10.1016/j.chb.2022.107526
  19. Wang, B., Luo, L., y Wang, X. (2023). “Back to the living room era”: Smart speaker usage and family democracy from the family dynamic perspective. New Media & Society. Publicación anticipada. https://doi.org/10.1177/14614448231155624
  20. World Health Organization. (2019). Guidelines on physical activity, sedentary behaviour and sleep for children under 5 years of age. World Health Organization. https://iris.who.int/handle/10665/311664