EducationMixed Methods

AI Literacy Begins at Age Five: Computational Thinking, Gender, and the Preschool Frontier

The STEM gender gap begins before children can read. New research from Spain, South Korea, and Turkey examines whether introducing AI literacy and computational thinking in preschool can build foundations for more equitable STEM participationโ€”and whether current approaches risk replicating the biases they aim to prevent.

By Sean K.S. Shin
This blog summarizes research trends based on published paper abstracts. Specific numbers or findings may contain inaccuracies. For scholarly rigor, always consult the original papers cited in each post.

If the gender gap in STEM fields begins in childhoodโ€”and decades of research suggest it doesโ€”then the logical place to intervene is before the gap opens. The emerging field of early childhood AI literacy takes this logic to its limit: introducing computational thinking and AI concepts to children as young as four or five, before gender stereotypes about "who does science" have fully crystallized.

The idea is both promising and fraught. Promising because early intervention has the potential to establish cognitive foundations and positive associations with technology before social pressures intervene. Fraught because the preschool classroom is a complex environment where well-intentioned technology integration can go wrong in subtle waysโ€”reinforcing the very patterns it aims to disrupt.

Gender in the Preschool Classroom

Cotino-Arbelo, Molina-Gil, and Gonzรกlez-Gonzรกlez (2025) examine what they describe as "an overlooked frontier": preschool classrooms. Despite decades of global initiatives aimed at increasing female participation in technology fields, gender-interest stereotypes continue to discourage girls from pursuing STEM. Their study analyzes computational thinking and AI literacy through a gender-based lens among early learners.

The significance of preschool-age intervention lies in developmental timing. Research in social cognition suggests that children begin to internalize gender stereotypes about academic domains between ages four and seven. By the time gender-differentiated STEM attitudes are typically measured (ages 10-12), the patterns are already entrenched. Intervening at the preschool level targets the formation period rather than the reinforcement period.

The study's gender-based analysis among early learners provides evidence on whether preschool AI literacy activities produce differential engagement and learning outcomes by genderโ€”a question that is critical for designing interventions that do not inadvertently advantage boys (who may receive more technology exposure at home) while claiming to serve all children equally.

Robot-Assisted AI Literacy in South Korea

Lee, Ku, and Ko (2025) explore the role of AI humanoid robots in developing preschoolers' AI literacy and computational thinking in South Korea. Using a mixed-methods approach, the study examines both learning outcomes and the experiences and perceptions of teachers and parents.

The study's context is significant: South Korea has invested heavily in AI education as a national priority, and the Korean Ministry of Education has introduced AI-related content into the national curriculum. The deployment of AI robots in preschool classrooms represents an early implementation of this national strategy.

The mixed-methods design allows the study to capture not just whether children learn from robots but how teachers and parents experience the integration. This stakeholder perspective is important because preschool AI education depends on adult mediatorsโ€”teachers who facilitate activities and parents who support or resist technology exposure at home. A technically effective AI education tool that teachers find burdensome or parents find concerning will not be adopted sustainably.

Teacher Perspectives: Between Enthusiasm and Anxiety

Kรถlemen and Yฤฑldฤฑrฤฑm (2025) evaluate the role of AI in preschool from the perspective of preschool teachers. Their study, involving 101 preschool teachers selected through purposeful sampling, provides a rare window into how educatorsโ€”the people who must actually implement AI literacy curriculaโ€”view the technology.

The teacher perspective reveals tensions that policy documents and curriculum guidelines tend to smooth over. Teachers recognize AI's potential to improve children's skills such as AI literacy and computational thinking, but express concerns about privacy and security (including personal data protection), infrastructure limitations, and their own insufficient AI literacy and competencies. Many teachers report feeling expected to integrate AI into their practice without adequate professional developmentโ€”a pattern that mirrors the broader challenge of technology integration in education.

The study's finding that teachers did not feel sufficient regarding AI literacy and competencies, and that concerns centered on content knowledge gaps, infrastructure, and classroom readiness, has practical implications. Investment in teacher professional development may yield greater returns than investment in technology hardware.

The Evidence from Meta-Analysis

Park, Min, and Chae (2025) contribute a meta-analysis on the effects of AI and digital play activities in early childhood. Their analysis synthesizes research on how AI technologies are being incorporated into educational content to foster key competencies such as computational thinking.

Meta-analytic findings provide a useful corrective to the enthusiasm of individual case studies. By aggregating across multiple studies with different designs, populations, and outcome measures, the analysis can identify whether effects are consistent, conditional, or artifactual. For a field as young as early childhood AI educationโ€”where publication bias toward positive results is likelyโ€”meta-analytic evidence is particularly valuable.

Computational Thinking Models for Early Science Literacy

Salleh and Omar (2025) provide a structured review of computational thinking models specifically designed for early childhood science literacy. The integration of CT models in early childhood science literacy has gained increasing attention due to their potential to enhance cognitive skills and problem-solving abilities in preschool learners.

The review identifies several CT components that can be meaningfully introduced at the preschool level: pattern recognition (identifying regularities in the environment), decomposition (breaking complex problems into simpler parts), algorithmic thinking (sequencing steps to accomplish a task), and abstraction (identifying essential features while ignoring irrelevant details). These components can be embedded in age-appropriate activitiesโ€”sorting games, building challenges, storytelling sequencesโ€”without requiring screens or digital devices.

This finding is important because it decouples computational thinking from technology use. CT can be developed through unplugged activities that require no hardware investment and no screen timeโ€”addressing two of the primary concerns that teachers and parents raise about early childhood technology education.

Claims and Evidence

<
ClaimEvidenceVerdict
Gender differences in STEM attitudes are already present at preschool ageCotino-Arbelo et al. (2025): gender-interest stereotypes discourage girls before formal schoolingโœ… Supported
AI robots effectively develop preschoolers' computational thinkingLee et al. (2025): mixed-methods evidence from Korean preschools with positive learning outcomesโœ… Supported
Preschool teachers are prepared to implement AI literacyKรถlemen & Yฤฑldฤฑrฤฑm (2025): teachers recognize potential but report inadequate training and supportโŒ Refuted
Computational thinking requires digital technologySalleh & Omar (2025): CT components can be developed through unplugged, screen-free activitiesโŒ Refuted
Early AI education consistently improves outcomesPark et al. (2025): meta-analysis provides aggregated evidence; effect consistency depends on implementationโš ๏ธ Uncertain

Open Questions

  • What is the appropriate developmental threshold for AI literacy? Can four-year-olds meaningfully understand "how AI works," or should early childhood focus on pre-computational skills (pattern recognition, sequencing) without the AI framing?
  • How do we assess computational thinking in pre-literate children? Standard assessment instruments assume reading and writing ability. Performance-based, observational, and play-based assessment methods exist but lack standardization.
  • Does early AI exposure reduce or reinforce the gender gap? If boys receive more technology exposure at home, school-based AI education may level the playing fieldโ€”or it may provide boys with a head start that amplifies existing advantages.
  • What are the long-term effects? Current evidence is short-term. Does preschool CT education translate into sustained STEM interest, engagement, and achievement through elementary school and beyond?
  • How do cultural contexts shape early AI education? The studies reviewed span Spain, South Korea, Turkey, and Malaysia. Cultural attitudes toward technology, childhood, gender, and education vary substantially. How should early AI curricula be culturally adapted?
  • Implications

    The research suggests that early childhood is a viable and potentially valuable context for AI literacy educationโ€”but only if implementation respects developmental constraints, invests in teacher preparation, and maintains equity as a design principle rather than an afterthought.

    The gender dimension deserves particular attention. If early AI education is introduced without deliberate attention to gender equity, it risks becoming another context in which boys receive more encouragement, more complex tasks, and more sustained engagement with technology than girlsโ€”replicating at age five the patterns that decades of intervention have failed to change at age fifteen.

    The path forward requires collaboration between early childhood educators (who understand development), computer scientists (who understand AI), and equity researchers (who understand how well-intentioned interventions can reproduce inequality). No single discipline has the expertise to get this right alone.

    References (5)

    [1] Cotino-Arbelo, A.E., Molina-Gil, J., & Gonzรกlez-Gonzรกlez, C. (2025). Computational Thinking and AI Literacy: A Gender-Based Analysis Among Early Learners. Proc. IEEE EDUCON 2025.
    [2] Lee, B., Ku, S., & Ko, K. (2025). AI Robots Promote South Korean Preschoolers' AI Literacy and Computational Thinking. Family Relations, 74(2), 13189.
    [3] Kรถlemen, E.B. & Yฤฑldฤฑrฤฑm, B. (2025). A New Era in Early Childhood Education (ECE): Teachers' Opinions on the Application of Artificial Intelligence. Education and Information Technologies, 30, 13478.
    [4] Park, Y., Min, H., & Chae, Y. (2025). Meta-Analysis on the Effects of AI and Digital Play Activities in Early Childhood. Korean Society for Young Children's Brain & Cognition, 15(2), 1.
    [5] Salleh, S.M. & Omar, R. (2025). Elements of Computational Thinking Models in Early Childhood Science Literacy: A Comprehensive Structured Review. International Journal of Research and Innovation in Social Science, 9(2), 121.

    Explore this topic deeper

    Search 290M+ papers, detect research gaps, and find what hasn't been studied yet.

    Click to remove unwanted keywords

    Search 8 keywords โ†’