Source-linked AI summary
StoryBuddy: A Human-AI Collaborative Chatbot for Parent-Child Interactive Storytelling with Flexible Parental Involvement
Zheng Zhang, Ying Xu, Yanhao Wang, Bingsheng Yao, Daniel Ritchie, Tongshuang Wu, Mo Yu, Dakuo Wang, Toby Jia-Jun Li
TL;DR
Parents face barriers to interactive storytelling, while fully automated question generation can exclude parents and overlook educational appropriateness and engagement. StoryBuddy combines AI question generation with configurable parent involvement across storytelling modes, and a user study found that parents could successfully collaborate with it while children remained engaged and entertained.
Problem
Parents may lack time or suitable questions for interactive storytelling, while fully automated systems can exclude parent involvement and produce questions poorly aligned with educational goals or children’s needs.
Method
StoryBuddy uses formative and participatory design to support AI-assisted co-reading, automated bot-reading, question configuration, follow-up generation, and child-progress tracking.
Results
Parents successfully collaborated with StoryBuddy to create interactive storytelling experiences, and the system performed well in keeping children engaged and entertained.
Takeaways & Limitations
Flexible interaction design can accommodate different levels of parental control while changing how children interact with parents and the agent during storytelling.
Takeaways & Limitations
The user study sample underrepresented parents from underrepresented racial groups, lower socioeconomic backgrounds, and non-traditional family structures.
Abstract
from arXiv · showhide
Despite its benefits for children's skill development and parent-child bonding, many parents do not often engage in interactive storytelling by having story-related dialogues with their child due to limited availability or challenges in coming up with appropriate questions. While recent advances made AI generation of questions from stories possible, the fully-automated approach excludes parent involvement, disregards educational goals, and underoptimizes for child engagement. Informed by need-finding interviews and participatory design (PD) results, we developed StoryBuddy, an AI-enabled system for parents to create interactive storytelling experiences. StoryBuddy's design highlighted the need for accommodating dynamic user needs between the desire for parent involvement and parent-child bonding and the goal of minimizing parent intervention when busy. The PD revealed varied assessment and educational goals of parents, which StoryBuddy addressed by supporting configuring question types and tracking child progress. A user study validated StoryBuddy's usability and suggested design insights for future parent-AI collaboration systems.
1 INTRODUCTION
StoryBuddy addresses barriers to parent-child interactive storytelling by combining AI-generated questions with flexible parental involvement. Its design emerged from formative and participatory studies and supports both parent collaboration and more autonomous storytelling.
- Interactive storytelling can improve children’s language, communication, cultural and emotional awareness, and cognitive development.
- Parents face difficulty creating appropriate questions, multitasking-related cognitive load, and challenges tracking children’s progress.
- Existing systems often use manually crafted questions, limiting interactive storytelling to a restricted set of prepared stories.
- Automatically generated questions can be factually correct yet trivial, plot-irrelevant, linguistically difficult, or disconnected from educational goals.
- StoryBuddy was designed through formative investigation and participatory design to support varied parental goals and flexible involvement.
- The system offers assisted parent-AI co-reading and asynchronous automated bot-reading, while configuring questions and tracking child progress.
2 RELATED WORK
Prior work demonstrates educational uses of interactive storytelling and conversational agents, but many systems rely on fixed content or overlook parent involvement. StoryBuddy extends this work through question generation for arbitrary stories and collaborative control over storytelling.
- Dialogic reading research favors open-ended Wh- questions, including recall prompts and Theory of Mind prompts about characters’ thoughts, feelings, and intentions.
- Digital storytelling systems have supported computational thinking, literacy, bilingual acquisition, science learning, vocabulary, and other developmental goals.
- StoryBuddy supports question-answering on any stories and flexible parent involvement through two modes for parent-present and parent-absent situations.
- Conversational agents can support verbal communication and parenting practices, but children’s limited communication skills make interaction breakdowns difficult to prevent or repair.
- Unlike model-focused QAG research, StoryBuddy emphasizes parents’ pre-configuration, question selection, and follow-up question generation.
3 FORMATIVE STUDY
The formative study examined families’ storytelling practices and expectations for digital systems. Parents valued educational, interactive, customizable technologies that could support both children’s learning and family interaction.
- 3 FORMATIVE STUDY: The study investigated families’ storytelling practices, digital-device use, and needs for digital storytelling systems.
- 3.1 Method: Ten families with children aged three to eight were interviewed across a university community and a nearby working-class, Spanish-speaking community.
- 3.1 Method: Researchers used inductive memoing, thematic coding, and periodic cross-checking by two coders to analyze interview transcripts.
- 3.2 Key Insights: Nine of ten parents recognized digital technologies and AI as supporting children’s language, literacy, and story-comprehension learning.
- 3.2 Key Insights: Eight parents preferred interactive storytelling because they associated it with more engaging and active learning.
- 3.2 Key Insights: Six parents believed technology could positively affect family interaction, while parents also described difficulty generating questions during reading.
- 3.2 Key Insights: Parents carefully selected and wanted to customize children’s digital content, including language appropriate to their child’s understanding.
4.1 Process
The participatory design process used parent interviews and low-fidelity storyboards to elicit design goals across different storytelling contexts. Storyboards varied parental presence and device type while foregrounding user needs and system capabilities.
- 4.1 Process: Participatory design sessions with four parents were conducted to uncover concrete design goals and strategies for StoryBuddy.
- 4.1 Process: Participants were primary caregivers, mostly mothers, with children ranging from four years old to older than five and frequent storytelling routines.
- 4.1 Process: Each remote session lasted about an hour and used four low-fidelity storyboards representing alternative StoryBuddy scenarios.
- 4.1 Process: Interviews before each session examined current storytelling practices, question-answering, and encountered challenges.
- 4.1 Process: Storyboard variations differed by whether the parent was present and whether the system ran on a tablet or smart speaker.
- 4.1 Process: Personas represented parents who value educational technology, parent-child interaction, content curation, and occasional unavailability.
- 4.1 Process: The storyboards emphasized parent motivations, goals, constraints, and proposed capabilities rather than detailed interface designs.
4.2 Findings
Parents viewed interactive storytelling as a way to sustain engagement, support learning assessment, and strengthen parent-child interaction, but their desired involvement and assessment goals varied. StoryBuddy therefore needed flexible interaction patterns, AI assistance, and configurable parental control.
- Engagement: Parents considered high child engagement important whether they were present or absent during storytelling.Children could become bored or seek unavailable parents’ attention, creating frustration for both parties.
- Engagement and assessment: Joint reading and questioning-and-answering were used to motivate children and assess or develop their comprehension.Parents reported taking turns reading and asking questions about story facts, actions, pictures, colors, and simple math.
- Assessment goals: Parents identified varied needs for question specificity, difficulty, timing, and follow-up based on children’s ages and preferences.Some children preferred questions before or after reading, while others might dislike interruptions or need more challenging questions.
- AI assistance: Parents found real-time question generation difficult because cognitive load could cause them to finish stories without asking questions.This occurred despite recognizing the benefits of question-answering.
- AI assistance: Parents liked AI support for identifying question opportunities and generating questions, especially when similar questions were grouped by themes, entities, or types.Grouping was intended to make question planning easier during joint reading.
- Assessment goals: Parents disagreed about assessment-focused features because embedded assessment goals might reduce children’s interest in the storytelling agent.One parent welcomed capability-based assessment, while another feared children would dislike being assessed.
- Parental involvement: Parents wanted active control over content selection and configuration, but differed in how finely they wanted to edit individual questions.Most accepted configuring question types without editing every recommendation, whereas one parent wanted granular control.
- Parental involvement: Parents faced a conflict between being present to strengthen relationships and being unavailable for live storytelling.They valued system variants that could support both involvement and periods of absence.
4.3 Design Strategies
The participatory design process produced six strategies for sustaining engagement, supporting parental questioning and assessment goals, accommodating control preferences, and handling parent availability.
- Engagement: DS1 recommends varying interaction means and approaches among the parent, child, and storytelling agent to maintain attention and engagement.
- Question facilitation: DS2 recommends assisting joint-reading question-answering by identifying opportunities and suggesting questions for parents to ask.
- Assessment: DS3 supports different parental goals regarding whether interactive storytelling should emphasize assessment.
- Follow-up questions: DS4 recommends relevant follow-up questions connected to semantic entities or story themes to improve engagement and assessment effectiveness.
- Configuration: DS5 accommodates varied preferences for configuring interaction content through multiple flexible levels of control.
- Parent availability: DS6 supports both synchronous parent-child joint reading with AI assistance and asynchronous AI-facilitated storytelling configured by parents.
5 STORYBUDDY
StoryBuddy is an AI-enabled tool for configuring, augmenting, and automating interactive storytelling while allowing parents flexible involvement in both content configuration and delivery. It provides joint-reading and automated modes to accommodate parent presence or absence.
- System overview: StoryBuddy supports interactive storytelling with flexible parental involvement in configuration and delivery, alongside children’s skill development, progress assessment, and engagement.
- System architecture: The system combines parent configuration, parent-child co-reading assistance, an automated conversational agent, progress tracking, and machine-learning question generation.
- Interaction modes: StoryBuddy provides distinct modes to reconcile parents’ desire for active participation with periods when they are unavailable.
- Parent presence: In joint-reading mode, AI identifies question opportunities, recommends questions, and adds conversational-bot interactions without displacing the parent’s role.
- Parent presence: Parents can select stories from the story library and view story text, illustrations, and AI-recommended questions linked to corresponding story content.
- Parent control: Parents retain control over storytelling while optionally delegating question-asking, answer-checking, progress tracking, or reading to the agent.
- Engagement: Dynamic interaction lets parents switch among reading, questioning, and conversational-agent assistance to reduce boredom.
- Parent absence: In automated mode, the agent engages children without a parent nearby while parents configure interactions and track or assess progress asynchronously.
5.3 Preference Configuration of Question Types
StoryBuddy lets parents configure the question types generated for stories, aligning interactive questioning with their children’s learning and development goals.
- Preference configuration: Parents can optionally configure preferred question types through a dedicated preference panel.
- Question types: The system supports questions about characters, settings, feelings, actions, causal relationships, outcomes, and future predictions.
5.4 Dashboard for Child Progress Tracking and Performance Assessment
StoryBuddy’s dashboard supports parent assessment of children’s storytelling performance across individual sessions and weekly progress. It provides question-level answers and aggregated accuracy statistics, alongside the system’s configurable question-generation pipeline.
- Dashboard functions: The dashboard shows a child’s performance for an individual storytelling session or aggregated progress over a week.Parents can review prior sessions and weekly progress.
- Dashboard functions: For each question in a selected session, parents can inspect the child’s attempts and the correct answer.
- Performance assessment: The dashboard reports overall accuracy, accuracy by question type, and the proportion of question types used in a session.
- Performance assessment: Weekly views include overall question-answering accuracy, question-type statistics, and the proportion of each question type across the week.Parents can tailor the dashboard to a particular question type.
- Question-generation support: StoryBuddy’s question pipeline generates question-answer pairs, supports specified question types, and groups questions to identify follow-up candidates.The QAG model combines rule-based answer generation, BART-based question generation, and ranking.
6 USER STUDY
A remote study with 12 parent-child pairs examined StoryBuddy’s two reading modes, user behaviors, perceived usefulness, and usability. Participants generally completed and valued the system, while observations and interviews revealed varied involvement preferences, age-related strategies, and interface limitations.
- Study procedure and completion: All 12 parent-child pairs completed both parent-AI co-reading and automated bot-reading sessions, averaging around 18 and 17 minutes respectively.
- Observed behaviors: Parents adopted different reading strategies, with younger children more often receiving parent-led or emotional reading and older children more often leading reading themselves.Five parents used auto-reading throughout; three parents of children aged 6–7 had children lead reading.
- Observed behaviors: Parents varied in question use: two asked every displayed question, seven selected questions based on perceived comprehensibility, and five did not use generated follow-ups.
- Observed behaviors: Six parents kept default question-type settings, while six modified them to adapt generated questions to their goals and needs.
- Observed behaviors: Eleven children interacted with the chatbot, although engagement varied from answering only the first question to making additional attempts after incorrect answers.One child skipped all chatbot interactions and used StoryBuddy only to read the book.
- Interview findings: Parents described StoryBuddy as useful for language learning, reading-skill development, reducing storytelling burdens, and personalizing reading experiences.Ten parents reported reduced burden, nine discussed personalization, and seven discussed reading-skill development.
- Design opportunities: Parents identified a need for age-based customization because questions and bot guidance could be too difficult for younger children or too slow and unhelpful for older children.
- Usability issues: Participants reported inaccurate speech recognition, widgets that were too small for children, and a desire for more colorful or cartoon-oriented interface elements.
7 DISCUSSION AND DESIGN IMPLICATIONS
StoryBuddy frames parent-AI storytelling as partial automation that preserves parental expertise while adapting involvement to parents’ goals and availability. Its user study found successful collaboration and sustained child engagement, while exposing opportunities to refine AI timing and roles.
- User study outcomes: Parents can collaborate with StoryBuddy to create interactive storytelling experiences, and the system performed well in keeping children engaged and entertained.Its interaction design supported changing how children interacted with the parent and agent throughout storytelling.
- Partial automation in parent-AI collaboration: StoryBuddy’s AI generates questions, answers, and follow-ups, while parents customize them using knowledge of their child, goals, and current constraints.This division of capabilities motivates partial automation rather than fully automated storytelling.
- Supporting flexibility in parent involvement: The system supports varied parental goals through question-preference configuration, child-progress tracking, and optional manual editing of questions and answers.Parents focused mainly on engagement or entertainment can skip these controls.
- Supporting flexibility in parent involvement: StoryBuddy converts some synchronous parental involvement into asynchronous control before sessions and progress monitoring afterward when parents are unavailable.The agent facilitates story delivery and question-answering, while parents configure interaction content and review child performance.
- Role of AI in parent-child interaction: The AI assists with question opportunities and follow-ups without displacing the parent’s central role, while also acting as a child-facing companion or mediator.These roles can reduce parents’ cognitive load and support relationship-building, skill development, and engagement.
- Design implications and limitations: Parents reported that proposing questions after every story section could disrupt their normal reading flow, indicating that AI timing and role boundaries need refinement.The authors recommend formative studies and iterative participatory design to define appropriate AI roles around multifaceted family goals.
8 FUTURE WORK
Future work extends StoryBuddy toward child-led interaction, adaptive question generation, broader devices and users, and age-sensitive designs. The authors also propose studying classroom use and children’s ability to configure the system themselves.
- Child-led interaction: Future versions may support child-led question-answering, allowing children to ask about story plots while the agent answers and proposes follow-ups.This direction is intended to engage children more directly and support learning with less reliance on adults.
- Device support: Smart-speaker deployment could address some parents’ screen-time concerns, but would require solutions for conversational context, turn-taking, story grounding, and progress display.StoryBuddy currently supports smartphones, tablets, laptops, and desktops.
- Adaptive question generation: A learned model could infer parent preferences from question selections and edits, reducing configuration effort and adapting question difficulty and types to child performance and engagement.The current question-asking plan remains static after configuration.
- Age-sensitive design: Future studies will examine how children of different ages interact with StoryBuddy and adapt vocabulary, complexity, cognitive demands, guidance, and visual-content interactions accordingly.The proposed design changes target differences in developmental progress across age groups.
- Classroom applications: Because the user study focused on home settings, future work will explore automatic question generation for teachers and classroom instruction.The authors identify personalized interactive reading instruction as a possible classroom application.
- Evaluation and deployment: The authors plan larger-scale deployment and further study of how parents configure stories, choose modes, and use interaction paradigms.The current usability validation does not resolve these broader usage questions.
9 CONCLUSION
StoryBuddy is an AI system for parent collaboration in interactive storytelling, informed by formative and participatory design studies. It provides distinct modes and dynamic interaction paradigms for flexible parental involvement before and during storytelling.
- Conclusion: StoryBuddy enables parents to collaborate with AI in creating interactive storytelling experiences with question-answering.Its design was informed by formative and participatory design studies.
- Conclusion: The system supports flexible parental involvement through two distinct modes and dynamic interaction paradigms across storytelling configuration and delivery.These features address different degrees of parent participation in the storytelling experience.