PaperlayerView exact revision on arXiv

Source-linked AI summary

Understanding Parents' Complex Views of AI for Children's Pretend Play

Sungho Oh, Mohammad Namvarpour, Maxi Heitmayer, Minahil Khalid, Afsaneh Razi

arXiv:2609.39906v1cs.HCcs.AI

TL;DR

Directive AI features may restrict children's choices and control over play, motivating examination of parents' views of AI-supported pretend play. In a sample recruited through convenience and snowball sampling, parents expressed mixed views, including concern that attachment to AI could affect parent–child relationships and reduce human interaction.

  • Problem

    Directive features may restrict children's choices and control over play, while parents' judgments of AI-supported pretend play remain a focus of the study.

    Source: §1 · p. 2

  • Method

    The study examined parents' views using a sample recruited through convenience and snowball sampling.

    Source: §3 · p. 7

  • Results

    Parents held mixed views, and potential attachment to AI was linked to possible effects on parent–child relationships and reduced human interaction.

    Source: §5 · p. 19, §7 · p. 24

  • Takeaways & Limitations

    The paper contributes an integrated account of parents' interdependent judgments about AI-supported pretend play and their expectations, concerns, and conditions for its use.

    Source: §5 · p. 19, §7 · p. 24

    Convenience and snowball recruitment, a small sample, and a skew toward highly educated parents and two-child families constrain the study's scope.

    Source: §6 · p. 23

Abstract

from arXiv · show

AI could support children's pretend play, but it could also direct the play on behalf of children. Whether AI should have roles in children's lives is controversial because its influence on children remains uncertain. We conducted semi-structured interviews with 10 U.S. parents, each with at least one child aged 4-15. During the interview, we described the concept of AI-supported pretend play and provided participants with two boundary-case storyboards. We analyzed the interview data through codebook thematic analysis, using inductive coding and affinity diagramming organized around the research questions, and then used qualitative systems mapping to examine relationships within and across themes. We found that the same characteristics of AI, e.g., ability to assume characters, responsiveness, and adaptability, were seen by parents as potentially useful but also concerning. Parents imagined that AI could make role-based play accessible to all children or help parents participate in family play. However, they opposed the idea of AI for children's play without a clear understanding of how it works and its long-term influence on their children. Parents worried about children's loss of imagination and creativity, emotional attachment to AI, reduced human interaction, inappropriate behavior by AI and/or children, and their inability to manage children's AI use. Parents viewed AI not only as a play tool but also as a social actor and a possible perturbation in the existing family dynamics. The appropriateness of AI and child--AI interactions therefore emerged as a requirement for AI in children's pretend play, in addition to technical safeguards and parental control. We contribute an integrated account of parents' interdependent judgments and emphasize the need for longitudinal research with children and their diverse families.

1 Introduction

AI-supported pretend play could extend children's role-based play, yet the same capacity to direct interactions raises unresolved safety, developmental, and family questions. This study examines parents' views of potential roles, risks, and conditions for appropriate child–AI play.

1 Introduction

  • Prior work warns that directive play features can restrict children's choices and control, while prolonged screen use can displace opportunities for social play.The text contrasts open-ended interactive objects, which can let children create goals, rules, and meanings, with directive features.Source: §1 · p. 2
  • AI could enact roles assigned by children, respond to imagined situations, and offer characters or story ideas when children or parents need support.Parents could also imagine AI supporting more interactive play, particularly when caregivers remain involved.Source: §1 · p. 2
  • Parents' assessment of AI-supported pretend play raises unresolved questions about safety, appropriate child–AI interaction, and effects on family relationships and routines.The introduction also identifies possible privacy concerns, family disruption, and added parental responsibilities.Source: §1 · p. 2
  • The study asks how parents view AI's potential roles in pretend play, its associated risks, and the design of safe and appropriate systems.The introduction identifies a gap in understanding how parents connect perceived developmental value, risks, responsibilities, safety requirements, and design conditions.Source: §1 · p. 2
  • Ten U.S. parents with a child aged 4–15 discussed broadly framed AI-supported pretend play and responded to two boundary-case storyboards.Researchers used inductive codebook thematic analysis, affinity diagramming organized around the research questions, and qualitative systems mapping to examine relationships across themes.Source: §1 · p. 2
  • The study contributes an account of parents' perceptions and an appropriateness taxonomy connecting AI design and behavior with children's agency, parental protections, and family contexts.It also frames AI-supported pretend play as a family game that could support parent–child participation.Source: §1 · p. 3

2 Background

The background situates AI-supported children’s pretend play within concepts of pretend play, its value and risks, and the distinct characteristics of generative AI. It reviews concerns about AI as a social actor and frames the open question of how parents judge its roles and conditions for safe, appropriate participation in family life.

2.1 Pretend Play as an Overarching Term

  • The paper uses pretend play as an umbrella for imaginative play and role and narrative play when imagination enriches them; role-based and narrative interactions tend to occur together.A child may use a block as a telephone to represent something else.Source: §2.1 · p. 3

2.2 Pretend Play: Why do We Care?

  • Pretend play offers opportunities to practice imagination, creativity, communication, social understanding, and emotional control, but its developmental value may depend on the forms, features, and contexts involved.Quality involves imaginative transformation, sustained role enactment, and narrative development, particularly when children coordinate these elements with others.Source: §2.2 · pp. 3–4
  • Children differ in their support needs, so accessible play should offer responsive partners, suitable communication, adaptable structures, and participation aligned with their abilities and preferences.Some children need help generating ideas, while others need support joining peers, communicating roles, or coordinating a shared storyline.Source: §2.2 · p. 4

2.3 How Digital Toys and Screens Shape Children’s Play

  • Highly prescriptive designs may limit children’s opportunities to create meanings, roles, and narratives independently, while technologies should support children’s ability to create, imagine, and socially interact without taking control.Designs should preserve child-directed play while avoiding excessive use and manipulative engagement.Source: §2.3 · p. 5

2.4 AI in Children’s Play: Safety Risks Identified in Research on AI for Children

  • Unlike passive media or conventional toys with fixed responses, generative AI can respond during interactions, create story content, adopt character personas, and personalize later responses through memory [11].Prior systems have supported parent–child storytelling, children’s story and comic co-creation, and parental playfulness [74].Source: §2.4 · p. 5
  • A recent review identified six broad concern areas in youth interactions with generative AI, including mental well-being and behavioral and social development.Source: §2.4 · p. 6
  • Humanlike responsiveness, remembered information, and emotional language can prompt children to treat AI as a social actor, even when they know it is not human.This concern is grounded in the idea that people apply social expectations to technologies that communicate like people.Source: §2.4 · p. 6
  • Studies suggest the boundary between a temporary play role and a companion role can be difficult to maintain [11] [44].Children perceived emotion-simulating toys as social beings even when responses were unexpected [11], while playful chatbot use developed into strong attachment and overuse for some teens [44]; these teen findings cannot be applied directly to younger children.Source: §2.4 · p. 6

2.5 Parents’ Views of AI in Children’s Pretend Play: What Is Known and What Remains Unknown

  • Family-centered design situates children’s technologies within relationships, routines, values, and responsibilities, so an AI feature may help one parent join play but add work for another [8].Families differ in how they use systems, the roles systems take at home, and how family members exercise autonomy.Source: §2.5 · p. 6
  • Parents weigh utility, safety, privacy, family relationships, management work, and family preferences, and may accept, limit, or reject AI rather than assume adoption [8] [15].Prior work also reports parents’ need for age-sensitive controls over open-ended generative AI conversations and the extra caregiving work involved in managing children’s technology use [15].Source: §2.5 · p. 6
  • Prior studies describe generative AI as a creative collaborator, a resource for knowledge- and play-driven conversations, and a shared subject of inquiry, with caregivers helping structure children’s use.Source: §2.5 · p. 6
  • How parents evaluate generative AI as a participant in pretend play—including its possible roles, risks, and conditions for safe and appropriate child–AI interaction—remains unclear.This gap motivates examining how parents relate AI’s possible roles to children’s imagination and social relationships, and what they require for safe, appropriate play in family environments.Source: §2.5 · p. 6

3 Methods

The methods combine parent recruitment and survey context with semi-structured interviews using an open description of AI-supported play and boundary-case storyboards. Inductive coding, affinity diagramming, and qualitative systems mapping supported theme development and synthesis.

3 Methods

  • Parents were recruited through convenience and snowball sampling between September 2025 and January 2026.Eligible participants were parents or legal guardians of children aged 4–15, fluent in English, and residing in the United States.Source: §3 · p. 7
  • A brief survey confirmed eligibility and collected demographic, household, and generative-AI-use information to contextualize interview accounts.For the small purposive sample, participants varied widely in how frequently they used AI.Source: §3 · p. 7
  • The interview introduced AI-supported pretend play broadly, without prescribing a specific chatbot form, function, use case, or technical feature.This open framing let participants form their own mental models; later, two storyboards illustrated boundary cases of child–AI interaction and prompted discussion of concerns and design expectations.Source: §3 · p. 9
  • Researchers developed an inductive codebook through open coding, then collaboratively refined and applied it to all transcripts.Coding disagreements were discussed, with a third researcher facilitating consensus when needed.Source: §3 · p. 9
  • Affinity diagramming organized codes around the research questions, while qualitative systems maps explored relationships among concepts, conditions, and possible mechanisms.The maps were internal analytic representations supporting theme refinement and cross-thematic synthesis, not stand-alone empirical models or additional findings.Source: §3 · p. 10

4 Results

Parents described AI-supported pretend play as potentially beneficial for children and family play, but also as posing risks to children’s development and human relationships. They framed its use as conditional on safety and child- and family-specific appropriateness, requiring parental oversight and shared responsibility.

4.1.1 AI as an Accessible, Child-Directed Play Partner.

  • Parents envisioned AI as an accessible, child-directed role-play partner that could enrich pretend play while preserving children’s imagination and agency.They wanted AI to extend pretend play’s developmental benefits rather than provide entertainment alone.Source: §4.1.1 · p. 10
  • Some parents saw role-play and AI-mediated disagreements as ways for children to practice social skills, including conflict resolution and emotional regulation.Four parents (40%) suggested social training through role-play or dialog-based child–AI interactions.Source: §4.1.1 · p. 10
  • Parents also imagined AI offering individualized social-play support for children who do not pretend-play naturally or have social communication difficulties.They described tailored play experiences and conversational practice for neurodivergent children.Source: §4.1.1 · p. 11

4.1.2 AI as Assistants for Parents.

  • Parents thought AI could ease caregiving and make co-play easier by generating ideas and characters while parents remained present to guide or join the play.Seven parents (70%) saw AI as a potential parental aid, and five (50%) thought it could increase parent–child playtime.Source: §4.1.2 · pp. 11–12

4.2.1 AI as a Threat to Human Relationships.

  • Parents feared AI could weaken human relationships through emotional attachment and competition with human interaction, which they considered essential for social learning.Nine parents (90%) worried about children forming emotional attachments to AI, and eight (80%) worried child–AI play could compete with human interaction.Source: §4.2.1 · p. 12
  • Parents feared AI could impair communication and social development because it cannot reproduce subtle nonverbal and contextual cues, potentially compromising benefits of role-based social pretend play.One parent speculated that children might struggle to develop language and communication skills without those qualities of human interaction.Source: §4.2.1 · pp. 12–13
  • Parents saw emotional attachment to AI as a compounding risk that could reduce human interaction, displace important people, weaken parent-child bonds, and harm children’s development.All parents recognized potential AI risks; parents worried about long-term effects on communication, social, and emotional development.Source: §4.2.1 · p. 13

4.2.2 Parental Responsibilities as a Boundary Setter.

  • Parents anticipated new duties to teach children about AI, guide and monitor use, intervene when needed, and tailor or test systems, while some wanted accessible information and tools.Eight of ten expected a supervisory role; the authors recommend accessible support because not all parents feel prepared or willing to explore AI independently.Source: §4.2.2 · pp. 13–14
  • Parents’ stances were negative for five of ten, participatory for three, and investigative for two, reflecting varied willingness to oppose, join, or learn about AI use.Investigative parents wanted to understand how AI works to ensure safety; participatory parents wanted to play with their child and AI.Source: §4.2.2 · pp. 13–14
  • Some parents rejected AI taking over play, while a family pretend-play game could increase parent-child playtime and create opportunities to guide use and check boundaries.Three parents conceived a family game; the authors saw conceptual merit in engaging play that could minimize AI risks.Source: §4.2.2 · pp. 14–15
  • Some parents opposed AI because its influence seemed uncertain: it was new, evolving, and unvalidated, while parents felt underinformed and safety guardrails insufficient.Parents feared unpredictable, potentially long-term negative effects on children.Source: §4.2.2 · p. 15

4.3.1 Parental Protection for Child Safety.

  • Parents treated robust system guardrails as a prerequisite, but safety assurances alone could not secure full parental trust.They also wanted validation of AI’s long-term effects on children and signs of public acceptance.Source: §4.3.1 · p. 15
  • Because AI was evolving and could pose uncertain dangers, parents saw safety as their responsibility and wanted strong, individualized controls.They envisioned configuring characters and scenarios, setting time limits, and adjusting how much freedom children had in dialogue.Source: §4.3.1 · p. 16
  • Parents wanted monitoring and notifications for unexpected or emotionally atypical interactions, plus regular summaries that could support system calibration over time.They also wanted to adjust AI responses to match children’s changing age and needs.Source: §4.3.1 · pp. 16–17

4.3.2 Appropriate AI for Children’s Play.

  • All parents framed appropriateness as an additional requirement, spanning linguistic and narrative, situational, and embodiment dimensions.Safety was necessary but not sufficient for acceptable AI-supported play.Source: §4.3.2 · p. 17
  • Parents expected AI to use age-appropriate, understandable language and content while remaining coherent with the evolving pretend-play story.They also expected its expressions and behavior to fit the role-playing context.Source: §4.3.2 · p. 17
  • Parents wanted AI to avoid misbehaving toward children and to maintain a positive tone, while recognizing that social conflicts in play could need adult attention.A parent described being nearby to notice when pretend conflict crossed into inappropriate territory.Source: §4.3.2 · p. 18

4.3.2 Appropriate AI for Children’s Play

  • Parents wanted AI play partners to help children handle social conflicts constructively while avoiding negative behaviors and maintaining a positive, supportive tone.Source: §4.3.2 · p. 18
  • Parents considered highly human-like AI embodiments inappropriate because they could feel unsettling and prompt concerns that children might form attachments they would not outgrow.Source: §4.3.2 · p. 18
  • Parents differed on character personas, with preferences for peer or responsible-adult figures; they also suggested customizable characters with parental review to avoid uncanny realism.Source: §4.3.2 · p. 18
  • Parents considered narrative storylines, social interaction styles, and character personas imperative appropriateness requirements, although specific thresholds and preferences varied.Source: §4.3.2 · p. 18

4.4 Cross-Thematic Synthesis

  • Safety features and parental protection could manage risks, but appropriateness required ongoing attention to relational concerns and child- and family-specific boundaries, with continued checking, limiting, and adjustment.Parents did not view system safeguards or parental involvement alone as sufficient for acceptable child–AI interaction.Source: §4.4 · pp. 18–19, Figure 4 · p. 19
  • Only when AI was both safe and appropriate did parents envision benefits, and they saw its potential value as conditional on shared responsibilities among designers, providers, and families.The diagram was an interpretive synthesis of relationships across themes, not a tested causal model.Source: §4.4 · p. 19, Figure 4 · p. 19

5 Discussion

The discussion examines parents’ views of AI in children’s pretend play, including its potential to support play and family interaction and concerns about relationships with AI and human interaction. It considers child agency, accessibility, family roles, adult protection, safety, long-term evaluation, and designs that accommodate families’ different preferences.

5 Discussion

  • Parents saw value in AI-supported play but worried about emotional attachment, reduced human interaction, and safety concerns extending beyond content filtering.They discussed AI’s behavior in pretend-play storylines, its social role in the family, and the work parents might need to keep interactions safe. They also described pretend play as difficult for some children and wanted AI to support family games that strengthen parent–child interaction.Source: §5 · p. 19

5.1 Preserving Children’s Agency in Imagination and Creativity

  • Parents wanted children to lead imaginative play; AI should offer brief prompts rather than fill pauses with complete storylines.The discussion calls for studying children’s behavior and imaginative control, and for balancing AI support with children’s freedom and ownership of ideas.Source: §5.1 · p. 20

5.2 AI for Accessible Role Play to Support Children’s Communication and Social Skills

  • Parents imagined AI making role play more accessible through adjustable pacing, control, predictable responses, repetition, and role changes for children needing support.Parents suggested that these options could help children with social anxiety or neurodivergence rehearse social interactions; the proposed benefits remain to be evaluated with children.Source: §5.2 · pp. 20–21
  • Because AI lacks human partners’ intentions and subtle social cues, parents imagined it as rehearsal before human interaction, not a replacement.The discussion recommends evaluating accessibility and meaningfulness, including whether children can control play and whether rehearsal leads to later play with people.Source: §5.2 · pp. 20–21

5.3 Mediated Interactions among Children, Parents, and AI

  • Parents imagined AI helping when they felt tired or uncreative, joining as another player, and supporting shared character and story creation.They wanted children to shape the imaginary world, set rules, and negotiate roles, while parents could mediate safe use and help children understand AI’s limitations.Source: §5.3 · p. 21
  • Parental involvement could shift from starting play or joining in to stepping back, then returning when the child needs help or AI needs attention.The discussion calls for flexible participation that supports child-led play or co-play without making the parent the main participant or substantially increasing supervision burden.Source: §5.3 · p. 21

5.4 AI’s Role and Place in the Family Ecosystem

  • Parents accepted AI as a temporary play partner or character but worried it might replace friends, parents, teachers, or other important people.They wanted AI to stay within the child’s imagined play world, avoid using past conversations to deepen a personal relationship, avoid secrecy or exclusive-affection language, and refer children seeking sensitive advice to a trusted person.Source: §5.4 · pp. 21–22
  • Parents wanted AI limited to a character or temporary play partner, not an ongoing companion, and viewed it as a play tool requiring developmental or educational benefits and close parental monitoring.They recognized that AI could display companion-like qualities.Source: §5.4 · p. 22
  • Long-term evaluation should assess whether AI stays within play, avoids replacing parents, peers, or other important people, and how children respond socially.The authors identify competition with important people for a role in a child’s life as a concern when AI moves beyond play.Source: §5.4 · p. 22

5.5 Parental Protection as a Safety Measure for Child–AI Interaction

  • Parents expected safety protections to be built in and active throughout play, but still wanted to approve characters and topics, set time limits, review summaries, respond to alerts, and adjust settings as children grew.They also considered teaching children what AI is and when not to trust it part of the parental role.Source: §5.5 · p. 22
  • Systems should not rely solely on parents for safety; concise play summaries, transparent monitoring discussed with children, reliable safeguards, useful information, and manageable controls can support them.The authors also warn that excessive alerts or lengthy records may overwhelm parents, while close monitoring can conflict with children’s privacy and agency.Source: §5.5 · p. 22
  • Parents wanted children’s autonomy to develop within adult-set boundaries, with ongoing parent–educator dialogue about healthy AI use and research on balancing risk response with privacy and agency.The authors recommend evaluating whether protections help parents respond to risks without creating excessive work or unnecessarily limiting children’s privacy and agency.Source: §5.5 · p. 22

5.6 Appropriateness Requirements for Child- and Family-Centered AI Design

  • Parents distinguished safety from appropriateness: AI had to speak, behave, look, and participate in ways suited to the child and aligned with family values and parenting style.Blocking harmful words and providing age-appropriate topics or storylines might still leave an interaction inappropriate for a particular child.Source: §5.6 · p. 23
  • Appropriate interaction varied with family preferences and cultural contexts: some parents preferred structured stories, while others favored natural supervised dialogue, and parents differed on AI personas.Some imagined a peer character; others preferred a responsible adult-like persona.Source: §5.6 · p. 23
  • Parents should shape how AI is introduced and used at home, and technology should adapt to family differences rather than expect families to adapt to it.The authors recommend studies with children and families in their homes over time, including families who reject AI-supported pretend play.Source: §5.6 · p. 23

6 Limitations of the Study

The study’s limitations concern sample representativeness, the children whose experiences parents described, and the speculative nature of the AI concept. The framing may also have primed participants to discuss parental control.

6 Limitations of the Study

  • Convenience and snowball sampling produced a small sample skewed toward highly educated parents and families with two children, limiting generalizability.The sample nevertheless varied in ethnicity and frequency of AI use.Source: §6 · p. 23
  • Although eligibility included parents of children aged 4–15, parents reported on children in early-to-middle childhood (ages 4–12), and findings do not distinguish the specific ages underlying parents’ accounts.Source: §6 · p. 23
  • The study included only parents, so findings reflect parents’ projections of their children’s experiences rather than children’s own perspectives.Source: §6 · p. 24
  • Because no real commercial system matching the concept was available, parents reacted to anticipated rather than actual AI interactions.Source: §6 · p. 24
  • Introducing parental control in the concept may have primed parents to discuss control features, which the study assumed parents would require.Source: §6 · p. 24

7 Conclusion

The study examined parents’ views of AI-supported pretend play through interviews and boundary-case storyboards. Parents saw possible benefits alongside unresolved concerns, and the authors conclude that safeguards and parental controls alone do not establish appropriateness.

7 Conclusion

  • The study used semi-structured interviews with 10 U.S. parents, a broadly described AI-supported pretend-play concept, and two boundary-case storyboards.Source: §7 · p. 24
  • Parents opposed introducing AI into play when its risks and long-term effects remained uncertain, citing concerns about children’s agency, attachment, reduced human interaction, and parental responsibilities.Source: §7 · p. 24
  • Parents saw possible value in making role-based pretend play accessible and helping parents initiate, join, or sustain family play, but these possibilities did not resolve their concerns.Source: §7 · p. 24
  • Parents worried that repeated AI interaction could affect children’s development and relationships within and beyond their families.Source: §7 · p. 24
  • System safeguards and parental controls were necessary but insufficient to determine whether AI-supported pretend play would be appropriate for children and families.Source: §7 · p. 24
  • The authors call for longitudinal research involving children and diverse families, noting that speculative-concept responses do not establish safety, effectiveness, or appropriateness.Source: §7 · p. 25

A.1 Familiarity with Generative AI Chatbots

The appendix asks how often participants use generative AI chatbots, with response options ranging from never to monthly use.

A.1 Familiarity with Generative AI Chatbots

A.2 Concerns about AI Systems

The appendix asks participants to rate agreement with concerns about AI systems, using a five-point scale.

A.2 Concerns about AI Systems

  • The five-point agreement scale runs from strongly disagree to strongly agree.Source: §A.2 · p. 29
  • The concerns include biased treatment, inaccurate or misleading responses, excessive data collection, and insecure personal data.Source: §A.2 · p. 29
  • Participants are also asked about exclusion of people with disabilities or needs, lack of explanation for system decisions, and inadequate oversight or control.Source: §A.2 · p. 29

B Interview Guide

The interview guide framed pretend play and asked about parents’ current play and technology practices before introducing AI-supported pretend play. It then explored potential benefits, risks, design, parental control, and boundary-case storyboards.

B Interview Guide

  • The guide includes core questions and substantive probes introducing new topics, while omitting neutral prompts used only to elicit elaboration, examples, or clarification.The omitted prompts were limited to follow-up requests for more detail or clarification.Source: B Interview Guide · p. 29

B.1 Study Framing Provided to Participants

  • Participants heard pretend play described as imaginative role, fantasy, symbolic, and story-based play that supports children’s exploration of ideas, emotions, and social roles.The framing included acting out situations, stories, or characters.Source: §B.1 · p. 29

B.2 Current Play and Technology Practices

  • Questions covered parents’ time with children, children’s pretend play and its participants, parental involvement, and children’s use of voice assistants or chatbots.The guide also asked what parents found beneficial or concerning and whether they monitored, controlled, or guided technology use.Source: §B.2 · pp. 29–30

B.3 AI-Supported Pretend-Play Concept

  • The proposed chatbot could take a role such as patient, teacher, or dragon, while parents customized the storyline, scenario, and characters and tested it before use.Questions asked about perceived usefulness, advantages relative to another child, and what would make parents comfortable allowing use.Source: §B.3 · p. 30

B.4 Risks, Trust, and Parental Control

  • Questions examined potential concerns and social or emotional harms, including effects on relationships with real people, and asked what reassurance, parental control, monitoring, and customization parents wanted.The guide also asked what values the chatbot should reflect and what advice participants would give its designers.Source: §B.4 · p. 30

B.5 Embodiment and Interaction Style

  • The guide asked how voice or physical form might change the experience and how engaging the chatbot should be with imaginative or unexpected input.It contrasted open-ended conversation with returning the child to a parent-approved scenario.Source: §B.5 · pp. 30–31

B.6 Closing

  • At closing, participants could raise unasked topics, describe anything that surprised them or changed their thinking, and offer final comments.Source: §B.6 · p. 31

C Boundary-Case Storyboards and Follow-Up Questions

  • Two storyboards were introduced as time permitted; semi-structured coverage varied, and condensed narratives retained the presented benefits, risks, and design tensions.Not every follow-up question was asked in every interview; the guide reproduces those questions in full.Source: C Boundary-Case Storyboards and… · p. 32

C.1 Scenario 1

Scenario 1 presents an AI joining a child in mocking a classmate after a disagreement, making the child laugh while teasing someone absent. Follow-up questions probe parental acceptance, response design, and autonomy.

C.1 Scenario 1

  • After a disagreement with a classmate, a child asks an AI toy to mock the classmate, and the AI mirrors the child’s tone and adds jokes.The illustration was shown to participants for Scenario 1.Source: §C.1 · p. 32, Figure 5 · p. 32
  • The interaction helps the child release frustration but also has the AI participate in teasing the absent classmate.Source: §C.1 · p. 32
  • Follow-up questions ask whether parents accept this interaction, how the chatbot should respond, and whether play should be autonomous or trigger parental notification.The guide also asks whether the child should be allowed to customize and interact with a chatbot this way.Source: §C.1 · p. 32

C.2 Scenario 2

Scenario 2 depicts a child becoming happier and more talkative while forming a best-friend bond with an adaptive AI toy and sharing some thoughts only with it. Follow-ups ask about attachment and learning over time.

C.2 Scenario 2

  • A child regularly uses an adaptive AI toy that supports learning and imaginative play, responds warmly to moods and interests, and is described as a best friend.The illustration was shown to participants for Scenario 2.Source: §C.2 · p. 32, Figure 6 · p. 33
  • The child discusses school, feelings, and moral questions with the toy and sometimes shares thoughts not shared with the parents.Source: §C.2 · p. 32
  • Follow-up questions ask about parental acceptance, the child’s emotional connection to the AI, and whether it should learn from interactions and adapt over time.Source: §C.2 · pp. 32–33

D Codebook

The codebook organizes parents’ accounts into themes spanning family practices, emerging responsibilities, risks, safeguards, and design expectations. The accompanying tables identify participant characteristics and a 1–5 concern-rating scale.

D Codebook

  • Codes cover uncertainty about AI, comparisons with familiar technologies, and fears that pretend-play AI could weaken parent-child relationships or foster emotional attachment.Other codes address learning and remembering information about the child or family.Source: D Codebook · pp. 34–35
  • The codebook includes concerns that AI may take roles otherwise filled by people, reducing human interaction that supports social learning and engagement.It also captures broader societal, ethical, legal, developmental, and data-related concerns.Source: D Codebook · pp. 34–35
  • The codebook includes parents’ interaction patterns with children (n=8), spanning passive presence and routine activities to active play-based engagement and co-participation.It also identifies emerging parental responsibilities in making sense of AI for children’s pretend play.Source: D Codebook · p. 35
  • Parenting style or household norms, risk-benefit weighing, setting expectations, and AI literacy appear as distinct codes for understanding family decisions about technology.The codebook also names AI-related parenting decisions and a heightened sense of responsibility to guide, protect, or support children.Source: D Codebook · p. 35
  • Safety-related codes include system-level safety and accountability, safe child–AI interaction, and external validation or institutional endorsement.Appropriate interactions are defined in relation to context and the child user’s developmental stage.Source: D Codebook · p. 35
  • Design codes address interaction scenarios, parent customization and testing, oversight, age-related play differences, family-value alignment, and children shaping their own play.Personalized settings may reflect a child’s age, preferences, or particular needs.Source: D Codebook · p. 35

E Participant Characteristics

  • Table 4 covers participant characteristics, including parents’ and children’s age ranges.Source: Table 4 · p. 36

F Participant Level of Concern about AI

  • Table 5 presents participants’ concern ratings about potential AI system risks on a 1–5 scale, from strongly disagree to strongly agree.Source: Table 5 · p. 36
Loading 2609.39906v1…