Technology as Infrastructure: The Evidence-Based Guide to Screens, Apps, and AI for Ages 4-9

A 2025 meta-analysis of 10,116 children found that excessive screen time was associated with poorer social-emotional development, with an odds ratio of 1.24 (95% CI: 1.16-1.33). Hyperactivity showed an even stronger association at OR 1.39. Yet a cluster-randomized controlled trial of 76 first graders found that just one month of coding activities produced a Cohen's d of 1.62 for executive function development, equivalent to seven months of standard activities compressed into four weeks (Arfe et al., 2019).

The technology question for micro-schools is not whether screens are good or bad. It is when, how, and for whom specific technologies produce specific outcomes under specific conditions. This episode examines what the research actually supports, where genuine uncertainty remains, and how to build a minimum viable technology infrastructure that enhances rather than undermines learning for children ages 4-9.

The evidence reveals a consistent pattern: the distinction between passive consumption and active engagement matters more than total screen time, adult mediation transforms outcomes dramatically, and implementation quality determines whether any intervention succeeds or fails. For micro-school founders navigating vendor claims, conflicting guidelines, and regulatory complexity, these findings provide a framework for evidence-based decision-making.

Section 1: Foundation - Why Technology Decisions Require Nuance

The Dose-Response Reality

The research establishes a clear finding: excessive screen time, typically defined as greater than two hours per day, is consistently associated with poorer developmental outcomes. But this finding comes with essential caveats that most guidelines fail to communicate.

A systematic review by Streegan and colleagues (2022), analyzing 85 studies including 47 cross-sectional, 36 cohort, and 2 case-control designs, found screen time generally associated with poorer socio-emotional functioning, executive functions, cognitive development, and motor skills (Streegan et al., 2022). However, methodological quality varied considerably: only 16 studies demonstrated good quality while 59 showed fair quality, limiting causal inferences.

The same review identified three conditions under which positive effects emerged: implementation of time limits, parental co-viewing, and exposure to educational content. This critical qualification is frequently lost in public discourse about screen time.

Todeti and colleagues (2024) documented a dose-response relationship: children with higher screen exposure had significantly lower cognitive assessment scores. Those exposed to more than three hours daily scored 82.4 plus or minus 12.8, compared to 96.3 plus or minus 10.7 for children with two hours or less. The negative correlation (r = -0.42, p < 0.001) persisted after controlling for age, socioeconomic status, parental education, sleep duration, and outdoor activity levels.

A Cambridge University longitudinal study published in January 2026, tracking over 1,000 children for seven years, found that screen time before age 2 altered brain network development by age 6, potentially leading to emotional delays. However, the same study found that high levels of parent-child reading neutralized these effects, emphasizing interactive human engagement over passive screen input.

Key Terminology

Before examining the evidence, several terms require precise definition:

Active versus passive screen time: Passive screen time involves watching television or videos without interaction, providing no opportunity for user input. Active screen time involves interactive engagement such as gaming, educational apps, video creation, or programming. Research on phonological memory development in preschoolers found passive TV watching negatively related to processing verbal information, while interactive time with smart screen technologies showed no significant association and posed no threat to phonological memory development (Frontiers in Education, 2021).

The video deficit effect: A well-established phenomenon in developmental science where infants and toddlers learn less from televised demonstrations than from live demonstrations. The effect peaks around 15 months and persists until approximately 30 months. A touchscreen study with 15-month-olds found the video deficit occurred even with interactive touch technology: cross-dimension groups transferring learning between 2D screens and 3D objects performed significantly worse than within-dimension groups (PMC, 2010).

Social contingency: The critical element enabling learning from screens. Groundbreaking research by Roseberry and colleagues (2013) demonstrated that toddlers only learned novel verbs in socially contingent interactions, defined as those where responses are immediate, reliable, and accurate in content. When experimenters interacted with children via Skype, providing responsive, child-specific exchanges, learning matched live interaction conditions. However, yoked video showing pre-recorded content of the same experimenter produced no learning despite identical visual presentation.

Mobile Device Management (MDM): Centralized software enabling remote configuration, policy enforcement, app deployment, and endpoint security for devices. For micro-schools with shared devices, young learners, and privacy commitments, MDM is non-negotiable infrastructure, not an optional administrative convenience (TechTarget, 2025).

The Displacement Hypothesis

Across domains including motor development, language acquisition, and physical health, the displacement hypothesis provides the most parsimonious explanation for negative associations between screen time and developmental outcomes. Screen time may harm not through direct toxic effects but by reducing time for activities that build skills: manipulative play for motor development, conversation for language, physical activity for health.

This framing has significant implications for policy. If displacement is the primary mechanism, then the focus should be on ensuring sufficient time for developmentally important activities rather than on screen time limits per se. A global study found only 17% of preschoolers worldwide met guidelines for physical activity, screen time, and sleep across all three domains (The Conversation, 2024).

Guidelines recommend for children ages 2-5: at least 180 minutes of physical activity daily, no more than one hour of screen time daily, and 10-13 hours of quality sleep. For ages 5-17: at least 60 minutes of moderate-to-vigorous physical activity daily, no more than two hours of recreational screen time excluding schoolwork, and 9-11 hours of sleep for ages 5-13 (Australian Institute of Family Studies, 2017).

Section 2: Evidence - What the Research Actually Shows

Educational Technology Effectiveness: Domain-Specific Findings

High-quality educational technology demonstrates moderate to large learning gains in specific domains when properly implemented. The evidence is strongest for mathematics, early literacy, and computational thinking.

Mathematics interventions: A pupil-level randomized controlled trial by Outhwaite and colleagues (2018) with 389 children ages 4-5 evaluated math apps implemented either as supplementary intervention or time-equivalent treatment replacing one daily small-group teacher-led activity. Results showed significantly greater math learning gains for both implementation models compared to standard practice only. Children receiving apps in addition to standard practice were 3-4 months ahead of controls (effect size d = 0.31), while children receiving apps instead of small-group instruction were 2 months ahead (effect size d = 0.21). Critically, no significant difference emerged between the two implementation models, suggesting high-quality apps provide effective math instruction whether supplementing or partially replacing traditional methods.

The apps incorporated features grounded in instructional psychology: immediate feedback, continuous assessment promoting retrieval-based learning, individualized pacing allowing scaffolding for differing needs, and socially interactive on-screen teachers providing demonstrations. Learning gains extended beyond targeted basic facts to higher-level math reasoning and problem-solving, demonstrating transfer effects.

Coding and computational thinking: A cluster-randomized controlled trial with 76 first graders compared one month of coding activities using Code.org to one month of standard STEM activities. The experimental group showed dramatically better performance at post-test (t(74) = -7.03, p < 0.001, Cohen's d = 1.62), representing a very large effect size (Arfe et al., 2019). Longitudinal data revealed improvements in planning and inhibition skills after one month of coding equaled or exceeded improvement attained after seven months of standard activities. The researchers hypothesized that computational thinking, defining clear orderly sequences of simple steps to solve complex problems, makes significant demands on executive functions and may boost them through targeted practice.

Literacy interventions: A comprehensive review using ESSA standards found teacher-implemented interventions achieved highest effectiveness ratings for K-5 literacy: +2 for reading fluency, +2 for reading comprehension, and +1.66 for alphabetics. Computer-only programs averaged only +0.33 effectiveness (Waldenu, 2025). A meta-analysis of 20 technology-based reading interventions for struggling readers found moderate overall effect size (Tau-U = 87%), with no significant differences across moderator variables including setting, grade level, technology type, or targeted skills (Nature, 2024).

Blended learning approaches: A meta-analysis of 37 studies (2000-2024) comparing blended learning to online-only learning found blended approaches produced positive upper-medium effects on student learning outcomes (SMD = 0.611, p < 0.001). Effects were particularly strong for cognitive outcomes (SMD = 0.698) and affective outcomes (SMD = 0.533). Moderator analysis revealed blended learning proved more effective for class sizes of 0-50 students, K-12 and university students, interventions within 3 months duration, and 30-69% proportion of online learning rather than higher or lower percentages (PMC, 2024).

The Critical Role of Adult Mediation

Across the literature, adult mediation emerges as the strongest moderator of screen time effects. Parental speech characteristics during co-viewing predict learning outcomes: parents with higher teaching focus during co-viewing had children who produced more previously unfamiliar words. Parents who knew which words their children did not understand and focused specifically on those vocabulary items facilitated learning most effectively (Escholarship, 2024).

A meta-analysis examining adult-child co-use during digital media found a small positive association with children's learning outcomes, supporting the hypothesis that social scaffolding during screen time mediates learning effectiveness (ScienceDirect, 2024).

High-quality joint media engagement can offset negative associations dramatically. A Stanford study found that quality co-viewing offset approximately 80% of negative associations between screen time and language outcomes. This finding suggests that the presence or absence of adult engagement may matter more than screen time quantity.

Background television represents the exception: it is consistently harmful. It negatively affects language use, attention, and cognitive development while reducing quality of parent-child interaction. Parents estimate approximately four hours daily of background TV exposure for children under 8.

AI-Powered Educational Tools: Promise Without Evidence for Ages 4-9

The rapid deployment of AI-powered educational tools has far outpaced rigorous research, particularly for young children. While commercial products like Khanmigo (Khan Academy's GPT-4-based tutor) exist, most target grades 3 and above, with virtually no peer-reviewed randomized controlled trials specifically examining ages 4-6.

A 2025 systematic review of AI-driven intelligent tutoring systems in K-12 education, published in PMC, analyzed 28 studies totaling 4,597 students and found effects were "generally positive but mitigated when compared to non-intelligent tutoring systems." Critically, only 3 of 50 studies in a foundational meta-analysis were conducted in K-12 settings; the vast majority involved university students.

The strongest recent evidence comes from Stanford's Tutor CoPilot RCT, which found students with AI-assisted tutors were 4 percentage points more likely to demonstrate mastery, with effects reaching 9 percentage points for students of lower-rated tutors. However, this study involved grades 3-6 students over just two months, and the AI assisted human tutors rather than replacing them.

Conflicts of interest warrant careful interpretation: Khan Academy studies are conducted in partnership with the platform; Google DeepMind's LearnLM study evaluated Google's own model; Stanford's Tutor CoPilot partnered with commercial tutoring company FEV Tutor. This does not invalidate findings but requires cautious interpretation.

Developmental concerns are significant: UNICEF research warns that overusing AI tools designed for adults can cause cognitive delays including underdeveloped executive functions like emotional regulation and abstract thinking. Dr. Ying Xu at Harvard's Graduate School of Education notes that children can learn from AI "as long as the AI is designed with learning principles in mind" but emphasizes AI cannot replicate "deeper engagement and relationship-building that come from human interaction."

Real-world compliance is low: An estimated 5% compliance rate with adaptive learning programs in real-world settings raises questions about whether lab-based efficacy translates to actual use.

What Popular Belief Gets Wrong

While many on social media argue that all screen time is equally harmful, the research actually shows that active engagement with high-quality content and adult mediation produces dramatically different outcomes than passive consumption. The distinction between TikTok scrolling and interactive math apps with parent co-use is fundamental.

While some social media voices claim technology replaces teachers in micro-schools, successful models use AI to assist teachers rather than replace them. Stanford's Tutor CoPilot research specifically evaluated AI-assisted human tutors. Alpha School claims AI "frees teachers for connections" rather than eliminating the teacher role.

While the AAP's two-hour guideline is often treated as evidence-based for all contexts, it is based on health correlations, not controlled experiments. The guideline does not distinguish educational versus entertainment content, active versus passive engagement, or with versus without adult mediation. The UK Royal College of Paediatrics and Child Health explicitly refused to set time limits, stating: "There is not enough evidence to confirm that screen time is in itself harmful to child health at any age."

Evidence Synthesis: Where Sources Agree and Conflict

Areas of strong consensus:

All sources agree that excessive screen time greater than two hours per day correlates with poorer outcomes across cognitive, social-emotional, and physical domains. All sources confirm that passive consumption is less beneficial than active engagement, with the video deficit effect well-established through experimental evidence. All sources find adult mediation enhances learning consistently across correlational and some experimental studies. All sources confirm that specific apps produce learning gains in math, literacy, and coding when well-implemented, though RCTs are short-term and focused on specific programs.

Areas of genuine conflict:

The relationship between touchscreen use and fine motor development produces conflicting findings. A UK study of 715 parents found earlier active scrolling associated with earlier fine motor milestone achievement. However, a Taiwanese quasi-experimental study with strong methodological rigor found children who did not use touchscreen tablets over 24 weeks showed improvement in manual dexterity, fine motor integration, and fine motor precision compared to tablet users. A German cross-lagged panel study following 141 preschoolers found greater media use at young age correlates with lower fine motor skill development.

The reconciliation likely involves activity type (active scrolling may enhance specific fine motor components while passive watching provides no benefit), dose-response effects (moderate use may be neutral while excessive use shows negative associations), and displacement mechanisms (negative effects may result primarily from reduced time for physical activities rather than direct screen harm).

Significant gaps in the evidence:

Most studies employ short intervention periods of weeks to months, with limited data on sustained effects. Whether benefits persist, fade, or compound over years remains unknown. The precise threshold for screen time varies by age, content type, and context. Current research cannot definitively determine whether technology effects operate through content consumed, mode of engagement, social context, displacement of other activities, or direct neurobiological effects. Few studies examine moderation by child characteristics beyond age and diagnosis.

Section 3: Application - Building Technology Infrastructure for Ages 4-9

Minimum Viable Technology Stack

A micro-school serving 10-20 learners ages 4-9 can run an effective, low-overhead technology program with a deliberately constrained stack. The biggest failure mode in small schools is not pedagogy; it is tool sprawl plus unmanaged devices plus weak privacy intake.

Core components for any configuration:

Student devices: 10-20 student devices plus 1-3 spares at a 10-20% spare ratio. Teacher device: 1 teacher laptop or desktop plus optional teacher tablet for modeling small-group apps. Display: 1 large display or projector for whole-group instruction. Network: business-grade router/firewall, segmented Wi-Fi for staff, student, and guest traffic, and content filtering appropriate for minors.

Device management is non-negotiable: MDM/UEM enables enforcement of settings, pushing apps, tracking inventory, managing updates, and enabling remote actions. For micro-schools with shared devices, young learners, and privacy commitments, centralized device management is required infrastructure.

ChromeOS-first approach (recommended for ages 7-9):

ChromeOS captured 60.1% of the global education market share as of 2025. 93% of U.S. school districts planned Chromebook purchases in 2025, up from 84% in 2023 (CommandLinux, 2026). The Chrome Education Upgrade costs $38 per device and enables centralized configuration, app pre-installation, user-access control, and device inventory via Google Admin console.

Google's extension of automatic updates to 10 years from platform release date, implemented starting 2023/2024, materially changes refresh planning. This reduces the risk of forced mid-cycle replacement and supports a 5-7 year ownership model if hardware durability is sufficient.

iPad-first approach (recommended for ages 4-7):

Tablets are the fastest-growing segment at 9.6% CAGR, driven by touch-optimized curricula and younger student preferences. Touch-first activities align with early literacy and numeracy workflows and reduce keyboarding friction for children still developing fine motor skills.

Jamf and Diamond Assets provide numerical comparison showing iPad total cost of ownership can be competitive when residual value is considered: iPad residual value of $145 after 2 years, $100 after 3 years, and $80 after 4 years versus Chromebook residual values of $10, $5, and $0 respectively. However, this analysis is vendor-produced and should be validated against local pricing.

Tool selection discipline:

Districts use over 10,000 edtech products based on 64 billion interactions captured in the EdTech Top 40 telemetry report (Education Week Market Brief, 2025). Micro-schools should explicitly not mimic this tool sprawl pattern.

Maintain a Tier 1 list of core tools covering LMS, literacy, math, and creation and a Tier 2 list for limited pilots. Retire tools every term unless renewed with evidence. The best privacy improvement for micro-schools is reducing tool count and enforcing MDM restrictions so teachers cannot accidentally introduce unreviewed apps.

Protocols for Implementation

Protocol 1: Content Selection Using the Four Pillars Framework

A systematic evaluation of educational apps using Hirsh-Pasek and colleagues' Four Pillars of Learning framework found overall scores low across all pillars. Free apps demonstrated significantly lower scores (p < 0.0001) compared to paid apps, primarily due to distracting enhancements including non-specific feedback and excessive rewards unrelated to learning goals (PMC, 2022).

Evaluate any tool against four criteria:

Active Learning (Pillar 1): Content should require children to manipulate, explore, or construct, promoting hands-on cognitive engagement rather than passive consumption.
Engagement in Learning Process (Pillar 2): Interactive features should engage users in activities versus distract. Provide relevant feedback that is corrective or formative and contingent on child's actions rather than nonspecific effects or excessive rewards unrelated to learning goals.
Meaningful Learning (Pillar 3): Content should connect to children's everyday experiences and be taught in a manner contextualizable within existing knowledge.
Social Interaction (Pillar 4): Resources should facilitate rather than impede peer collaboration and adult-child interaction.

Protocol 2: Screen Time Allocation by Age and Activity Type

For ages 4-6: Keep digital time constrained and purposeful. Short rotations and center-based use are appropriate. The 1-hour guideline has weak-to-moderate support, but content quality, co-viewing, and what screen time displaces matter more than absolute duration. Educational content with adult engagement may provide benefits, particularly for disadvantaged children. AI tools lack sufficient evidence and raise privacy concerns for this age group.

For ages 7-9: Blended time can increase for writing, research, and targeted practice but should remain anchored by teacher-led instruction and offline manipulatives. Focus on protecting sleep (no screens within 1-2 hours of bedtime), ensuring physical activity (at least 60 minutes daily), distinguishing educational from recreational use, and establishing consistent family and school expectations.

Apply the 20-20-20 rule during all screen use: Every 20 minutes, look at something 20 feet away for at least 20 seconds to relax eye muscles and reduce focusing fatigue. Maintain screens at least 18-24 inches from eyes, positioned just below eye level.

Protocol 3: Privacy and COPPA Compliance

The COPPA 2025 amendments, effective June 23, 2025 with compliance deadline April 22, 2026, now require separate verifiable parental consent before using children's data for AI training, expand "personal information" to include biometric identifiers like voiceprints, and prohibit indefinite data retention.

The FTC's school authorization exception allows schools to consent on behalf of parents if the data is used solely for educational purposes and not for commercial gain. However, in its recent COPPA Rule review, the FTC declined to codify this exception into regulation, leaving schools reliant on guidance rather than firm regulatory text. The Edmodo enforcement action established that vendors cannot shift COPPA liability to schools via Terms of Service.

Maintain a one-page Approved Tools Register documenting: tool name, purpose, data types collected, retention and deletion method, and consent status. Require written parent notice and consent where applicable before use. Review annually and on vendor policy changes.

Protocol 4: Adult Mediation During Technology Use

Based on the evidence showing adult mediation as the strongest moderator of screen time effects, implement these specific practices:

During educational app use, adults should: know which specific skills students have not mastered and focus on those, ask questions about what students are doing rather than passively observing, connect digital activities to real-world experiences and offline play, and model thinking aloud when helping with problems.

The teaching focus matters more than mere presence. Parents and teachers who actively direct attention to learning content produce better outcomes than those who simply sit nearby.

Co-viewing guidelines for ages 4-7: Adults should be actively present during all screen time, not merely in the same room. For ages 7-9, supervised independent use can increase gradually as children demonstrate self-regulation, with periodic check-ins and reflection on what was learned.

Protocol 5: Physical Activity Integration

Given the displacement hypothesis, ensure that technology use does not displace movement. For every 30 minutes of screen time during the school day, schedule at least 30 minutes of physical activity. Remove screens from outdoor time entirely.

Monitor total screen exposure across educational and recreational contexts. A child who uses educational apps for 90 minutes at school and then watches videos for 2 hours at home exceeds reasonable thresholds even if each context seems moderate.

Accessibility compliance: The DOJ's April 2024 Final Rule under Title II of the ADA requires all public school web content and mobile apps to meet WCAG 2.1 Level AA standards. Large entities serving populations of 50,000 or more must comply by April 24, 2026. Small entities serving populations under 50,000 must comply by April 26, 2027. This applies to websites, mobile apps, digital documents, and online course content.

State-level screen time legislation:

Minnesota Statute 124D.166, effective July 1, 2022, prohibits individual-use screens such as tablets and phones in preschool and kindergarten without direct engagement from a teacher or peers. Exemptions apply for students with IEPs or 504 plans.

Florida HB 379, effective July 1, 2023, prohibits TikTok on all district-owned devices and networks and bans wireless device use during class unless expressly directed by a teacher for educational purposes.

Utah SB 152 and HB 311 from 2023-2024 restrict minor access to social media between 10:30 PM and 6:30 AM and prohibit social media companies from collecting data on minors or using addictive design features.

Missouri has proposed legislation that would limit K-5 digital instruction to 45 minutes per day and mandate 70% of work be completed on paper. This bill remains pending.

Federal funding considerations:

E-Rate funding for off-campus Wi-Fi including hotspots and school buses was rescinded by the FCC for Funding Year 2025 onwards. E-Rate funds are now limited to on-campus connectivity.

Title IV-A Student Support and Academic Enrichment grants require that no more than 15% of funds designated for effective use of technology can be spent on purchasing technology infrastructure. The remaining 85% must support professional development and capacity building.

Caveats and Context

Who these recommendations apply to: Children without developmental differences in typical micro-school settings with adequate adult supervision. Children with autism spectrum disorder show particularly strong benefits from digital interventions, with meta-analysis finding statistically significant large effects on enhancing social-emotional skills compared to other domains (PubMed, 2024). Children with ADHD show bidirectional relationships; well-designed educational apps providing immediate feedback and short segments may benefit ADHD learners despite screen time concerns.

Limitations of the evidence base:

Most screen time research relies on parent-reported data subject to recall bias and social desirability effects. Screen time conflates diverse activities from educational apps to passive video to social media to video games. Socioeconomic status shows stronger associations with developmental outcomes than screen time. Reverse causality is plausible as children with behavioral difficulties may receive more screen time. Research lags technology; most evidence concerns TV viewing with very few studies on modern interactive apps and tablets.

The RAND Corporation's November 2025 finding that they could identify less than 0.1% of U.S. microschools for rigorous evaluation reveals a sector-wide accountability gap. Claims about micro-school technology outcomes often cannot be independently verified.

What remains genuinely uncertain:

Optimal age for AI introduction remains unclear. Dr. Ying Xu at Harvard states it is a "tough question to recommend an age limit." Whether strict time limits outperform quality-focused approaches has not been tested in comparative trials. Long-term developmental impacts of modern interactive media remain unknown because technology evolves faster than research. Whether moderate touchscreen use benefits or harms fine motor development in 4-9 year olds produces conflicting findings.

Key Takeaways

Implement the active-passive distinction operationally. Ban background television entirely. Distinguish between passive video consumption and interactive educational apps in both policy and practice. For ages 4-6, prioritize touchscreen activities requiring manipulation over video content. For ages 7-9, prioritize creation tools like writing and coding apps over consumption tools.
Build adult mediation into the schedule. Adults should be actively present and teaching-focused during all screen time for ages 4-6. For ages 7-9, maintain at least 50% co-viewing or active supervision with periodic check-ins. Apply the Mrs. Lewis standard from Episode 3: monitoring should occupy the majority of guide time during technology blocks.
Use one platform with centralized management. Choose ChromeOS for ages 7-9 writing-heavy contexts or iPad for ages 4-7 touch-first contexts. Do not mix platforms without strong instructional justification and staff capacity. Budget Chrome Education Upgrade at $38 per device or equivalent MDM for iPads from day one.
Constrain tool count ruthlessly. Maintain a maximum of 5-7 Tier 1 tools covering core instructional needs. Pilot new tools with explicit success criteria and retire them unless evidence supports continuation. Every additional tool increases privacy review workload, support burden, and potential for policy violations.
Protect sleep and physical activity non-negotiably. No screens within 1-2 hours of bedtime. Ensure at least 60 minutes of physical activity daily. Monitor total screen exposure across school and home contexts. Remember that displacement of movement and sleep may be the primary harm mechanism rather than direct screen effects.
Wait on AI tools for ages 4-6; proceed cautiously for ages 7-9. The evidence base for AI-powered educational tools with children under 7 is essentially nonexistent. For older elementary students, use AI to assist human teachers rather than replace them. Ensure any AI tool has clear COPPA compliance, no data retention for training purposes, and human oversight of all interactions.

The technology question is ultimately not about technology at all. It is about ensuring that micro-schools maintain what makes them valuable: responsive adult relationships, individualized attention, and learning environments calibrated to each child's developmental needs. Technology can support these goals or undermine them. The difference lies in implementation quality, not in whether screens are present.

Sources

Tier 1: Meta-Analyses, Systematic Reviews, Official Statistics

2025 meta-analysis on screen time and social-emotional development. OR 1.24 (95% CI: 1.16-1.33), N=10,116 children.
https://pubmed.ncbi.nlm.nih.gov/40055171/

Meta-analysis 2024 on blended learning vs. online-only. SMD = 0.611, p < 0.001, N=37 studies.
https://pmc.ncbi.nlm.nih.gov/articles/PMC12467614/

Streegan, C.J.B., et al. (2022). Effects of screen time on the development of children under 6: A systematic review. N=85 studies (47 cross-sectional, 36 cohort, 2 case-control).
https://jpnim.com/index.php/jpnim/article/download/e110113/873/6350

2025 PMC systematic review of AI intelligent tutoring systems. N=28 studies, 4,597 students.
https://pmc.ncbi.nlm.nih.gov/

NAEYC and Office of Educational Technology. Developmentally appropriate technology guidelines.
https://online.lynn.edu/resources/education/technology-impact-early-childhood-education

DOJ Title II Final Rule (April 2024). WCAG 2.1 Level AA mandate for public schools.
Compliance deadlines: April 24, 2026 (large entities); April 26, 2027 (small entities).

FTC COPPA guidance. School authorization exception (not codified in 2024/2025 rule update).
https://ikeepsafe.org/

Tier 2: Randomized Controlled Trials, Large Studies, Government Reports

Outhwaite, L.A., et al. (2018). Math apps RCT. Effect sizes 0.21-0.31, N=389 children ages 4-5.
https://pmc.ncbi.nlm.nih.gov/articles/PMC6366442/

Arfe, B., et al. (2019). Coding and executive function. Cohen's d = 1.62, N=76 first graders.
https://pmc.ncbi.nlm.nih.gov/articles/PMC6917597/

Roseberry, S., et al. (2013). Social contingency and toddler language learning. RCT.
https://pmc.ncbi.nlm.nih.gov/articles/PMC3962808/

ABCD study (Nature 2025). Screen time and ADHD/brain structure. Longitudinal, N=10,116 baseline, 7,880 follow-up.
https://www.nature.com/articles/s41398-025-03672-1

Stanford Tutor CoPilot RCT. AI-assisted tutoring. 4-9 percentage point mastery gains, grades 3-6, 2 months.

Cambridge University study (January 2026). Screen time before age 2 and brain development. Longitudinal, N=1,000+ children, 7 years.

Todeti, P., et al. (2024). Screen time and cognitive scores. Correlation r = -0.42, p < 0.001, ages 3-12.
https://academicmed.org/

Learning Policy Institute review. Effective PD: 35 rigorous studies, 49 hours annually showed 21 percentile point gains.
https://learningpolicyinstitute.org/

Minnesota Statute 124D.166. Preschool/K screen time prohibition. State law effective July 1, 2022.

Florida HB 379. Device/social media restrictions. State law effective July 1, 2023.

FCC E-Rate rescission (September/October 2025). Off-campus Wi-Fi eligibility removed.

Tier 3: Industry Reports, Implementation Studies, Expert Commentary

EdTech Top 40 report (Instructure/LearnPlatform 2024-25). 64 billion interactions across 10,000+ products. Usage telemetry.
https://marketbrief.edweek.org/

Jamf/Diamond Assets. iPad vs. Chromebook TCO. Vendor-produced scenario.
https://resources.jamf.com/documents/infographics/

CommandLinux (2026). ChromeOS market share statistics. Industry aggregation.
https://commandlinux.com/statistics/chromeos-market-share-in-education/

Digital Promise Dynamic Learning Project. Full-time coaches in 50 schools across 5 states. Implementation study.

Dr. Ying Xu, Harvard Graduate School of Education. AI's impact on children's cognitive development. Expert commentary.
https://www.childrenandscreens.org/

Alpha School AI-first model controversy. WIRED podcast January 2026. Parent complaints.

Hirsh-Pasek, K., et al. Four Pillars of Learning framework for educational app evaluation.
https://pmc.ncbi.nlm.nih.gov/articles/PMC8916741/

Building a Micro School: Technology as Infrastructure — Report