Young Children’s Perception and Interaction with Intelligent Agents: A Scoping Review

• In Event: Computational Playgrounds: Foundations and Futures of AI, Play, and Pedagogy in Early Childhood

Abstract

Intelligent agents (IAs), socially responsive agents powered by computer systems, are ubiquitously shaping young children’s lives through diverse interactive formats and progressively evolving concepts of intelligence (Druga et al., 2019; Rapti & Sapounidis, 2024). Despite growing interest (e.g., Li et al., 2022), the field lacks a comprehensive understanding of how children perceive and engage with IAs, critical for designing child-friendly IAs, advancing developmentally appropriate educational practices, and informing research.
To this end, we conducted a scoping review of 80 empirical studies published from 2020 to 2024, systematically identified across education, psychology, and computer science databases, to examine how children aged 3 to 8 interact with and perceive IAs (see Tables 1-1 and 1-2 for search terms and inclusion/exclusion criteria). Based on Arksey and O’Malley’s (2005) approach (see Figure 1-1), our review addressed the following questions: (1) What is the current state and classification of IA research involving young children? (2) How do children interact with IAs and what shapes these interactions? (3) How do children perceive IAs and what influences these perceptions? and (4) How do children’s perceptions and interactions interrelate?
Our review revealed four major findings. First, related research has surged in recent years, yet is hindered by inconsistent IA definitions and typologies, making cross-study comparisons challenging. To address this, we propose a two-tiered classification system, organizing IAs by their morphology and level of anthropomorphism (Low, middle, and high). Additionally, current studies predominantly rely on Western, homogenous samples and overlook the impact of gender, socioeconomic status, and cultural diversity. Second, children’s interactions with IAs are context-dependent, ranging from direct physical engagement to indirect virtual presentation or no active interaction at all (e.g., drawing robots). Most studies emphasize individual child-agent interactions, with less attention given to collaborative contexts involving pairs, peers, or parents. The interaction quality notably depends on IA’s expressiveness and reliability, the child’s age, prior experience, and broader cultural context. We caution against the prevalent assumption under interactions that positive, trust-based relationships with IAs are universally beneficial and advocate for critical examination. Third, children frequently attribute mental, cognitive, and socioemotional qualities to IAs—sometimes at levels comparable to humans or animals. Age and prior experience significantly affect children’s attribution of agency, trustworthiness, and moral standing. Moreover, IA design, particularly anthropomorphic features, profoundly influences children’s initial perceptions and ongoing interactions. Fourth, our analysis revealed a dynamic relationship between children’s perceptions and interactions: initial perceptions shape how children engage, while interaction experiences, in turn, reflect, refine, and reinforce their views of IAs’ intelligence, trustworthiness, and social presence.
In sum, this review underscores the importance of moving beyond anthropomorphic novelty in IA design, advocating for developmentally informed, culturally inclusive, and flexible roles for IAs in children’s lives. Educators might thoughtfully integrate IAs to support human connection, foster AI literacy, encourage critical reflection, and empower children’s agency. Future research should broaden participant diversity, adopt more child-centered and qualitative methodologies, and explore the long-term impacts of IA engagement to ensure ethical, equitable, and meaningful integration of intelligent agents in early childhood education.

Authors

• Grace Yaxin Xing, University at Buffalo - SUNY

• Zhuoyun Cai, University at Buffalo - SUNY

• Yuya Yamamoto, University at Buffalo - SUNY

• X. Christine Wang, University at Buffalo - SUNY