K-12Issue 164 records
K–12
Fresh K-12 signal pairs a binding megadistrict operating floor with supervised-tutoring promise and early-adopter realism. NYCPS’s official 2026–27 AI and screen-time guidance bans student-facing generative AI in grades 2K–8, caps 1:1 screen time by band, requires two HS AI-literacy modules, and confines high-school GenAI to five named, dosage-capped pilots (Quill, Edia, Brisk Teaching, Playlab, Intel AI-Ready Schools) while barring staff AI for grading, behavior monitoring, or placement decisions. Against that constrained-pilot design, a Google/Eedi exploratory RCT of LearnLM in five UK secondary schools (N=165) finds human-supervised, pedagogically fine-tuned tutoring matched or beat human-only tutors—including +5.5 pp on novel subsequent problems—when experts could edit every AI draft. CRPE’s May 2026 Early Adopter brief (45 districts/20 states) warns sophistication is not transformation: 58% remain System Improvers, with procurement lag and usage-score evaluation dominating. Idaho’s Title 33 Chapter 70 statute hardens the human-teacher floor statewide. Together: treat supervised, curriculum-tied pilots as the evidence path; do not equate tool access or Early Adopter branding with instructional redesign.
- Evidence records
- 4
- Cross-source patterns
- 2
- Evidence classes
- 2 standards or public-body guidance1 academic research1 independent research
- Outcomes
- 2 cautionary1 effective1 emerging
- Source freshness
- 2 recent2 older, newly relevant
- Research completed
- 2026-09-22
Choose a role to see its takeaway beside every record in the ledger.
Patterns across the evidence
Synthesis · Lighthouse Advisory interpretation
Human-supervised pedagogically tuned tutoring shows exploratory gains while megadistrict floors ban open student GenAI
LearnLM’s supervised UK exploratory RCT reports novel-problem gains when experts edit AI drafts, while NYCPS bans student-facing GenAI in 2K–8 and confines HS use to five vetted, dosage-capped pilots—implying districts should evaluate coach-mode, human-gated tutors, not open chatbots.
Operating questionAre student GenAI pilots limited to human-supervised, curriculum-tied tools with edit gates and dosage caps rather than open chatbot access?
Supporting evidenceLearnLM Team (Google) & EediNew York City Public Schools (NYCPS)
Early Adopter sophistication mostly reinforces existing models while statutes and megadistrict floors reassert human-teacher control
CRPE finds 58% of Early Adopter districts are System Improvers stuck in procurement lag and usage-score evaluation, even as Idaho’s Chapter 70 requires that GenAI not replace human teachers and NYCPS bars AI from grading and student trajectory decisions—so governance work must target instructional redesign capacity, not tool count.
Operating questionDo we measure AI programs by instructional-model change and human-decision integrity, or only by adoption and usage dashboards?
Supporting evidenceCenter on Reinventing Public Education (CRPE)Idaho State Legislature / Idaho Code
Evidence ledger
Full record · every source keeps its link and limitations
NYCPS publishes 2026–27 AI and screen-time floor: GenAI moratorium in 2K–8, five dosage-capped HS pilots, staff AI barred from grading and placement
NYCPS’s official guidance for 2026–27 imposes a moratorium on student-facing generative AI in grades 2K–8, recommends grade-banded 1:1 screen-time limits, requires two 45-minute high-school AI literacy modules, and confines HS GenAI to five centrally approved pilots (Quill, Edia, Brisk Teaching, Playlab, Intel AI-Ready Schools) with weekly dosage caps and one-pilot-per-student exposure. Staff may use AI for planning/operations under ERMA privacy review but must not use AI for grading, behavior monitoring, or placement/promotion/graduation decisions. IEP/504 assistive technology remains excepted.
Why it matters, evidence and limitations
- Why it matters
- Largest-U.S.-district operating model other LEAs will be asked to match or explain against: grade-banded GenAI floors, named pilot allowlists, staff high-stakes AI prohibitions, and privacy review gates.
- Evidence and measured results
- Opened the NYCPS guidance page end-to-end via WebFetch: Chancellor note, grade/AI/screen table, 2K–8 moratorium language, HS literacy and five-pilot tables with dosage examples, staff-use limits, ERMA explanation, accessibility exceptions, and afterschool/CBO alignment commitments.
- Limitations and uncertainty
- Guidance without outcome evaluation yet; enforcement and the full list of disabled AI features in the ~38 tools are incomplete on-page; publicationDate aligned to Sept 2026 announcement/rollout rather than an on-page timestamp.
LearnLM exploratory RCT (UK): human-supervised pedagogically fine-tuned tutoring matches human tutors and lifts novel-problem success +5.5 pp
In an exploratory RCT with 165 students across five UK secondary schools, expert tutors supervised LearnLM drafts on the Eedi math platform and could edit before sending. Tutors approved 76.4% of drafts with zero or minimal edits. Students supported by LearnLM performed at least as well as human-tutor-only peers on measured outcomes and were 5.5 percentage points more likely to solve novel problems on subsequent topics (66.2% vs 60.7%). Authors frame results as evidence for pedagogically fine-tuned, human-gated tutoring—not unsupervised chatbots—and call for longer longitudinal and standardized-assessment follow-up.
Why it matters, evidence and limitations
- Why it matters
- Transferable design constraint for U.S. districts running narrow HS AI tutoring pilots: require coach-mode models, human edit gates, and unaided novel-problem checks rather than access-only deployments.
- Evidence and measured results
- Opened arXiv PDF via curl+pdftotext: abstract sample and design, supervision protocol, 76.4% approval statistic, +5.5 pp novel-problem contrast, tutor interview themes on Socratic questioning, and limitations calling for longitudinal standardized outcomes.
- Limitations and uncertainty
- Exploratory, N=165, UK secondary math only; Google & Eedi authorship; human tutors always present—does not justify unsupervised student GenAI; Dec 2025 preprint used now because it maps directly onto 2026–27 supervised-pilot operating models (e.g., NYCPS).
CRPE Early Adopter brief: 58% of leading districts are System Improvers—AI sophistication without instructional transformation
Drawing on surveys and interviews with leaders from 45 Early Adopter districts in 20 states, CRPE finds coordinated AI adoption is rising but transformation is rare: 58% are System Improvers strengthening existing models, 16% System Changers amplifying mastery/competency reforms, and ~7% Reimaginers. AI fluency does not substitute for educational vision; districts hit teachers-first/students-later sequencing, delayed family engagement, procurement lag, and evaluation that leans on usage or standardized scores. Support networks help responsible adoption more than instructional redesign. Why now: explains why 2026 ban/pilot divergence may not yield learning redesign without new capacity.
Why it matters, evidence and limitations
- Why it matters
- Diagnostic for district AI roadmaps—separates responsible-adoption checklists from true instructional-model change and warns against mistaking Early Adopter branding for outcomes.
- Evidence and measured results
- Opened CRPE brief page via WebFetch: key findings on pathways (58%/16%/~7%), system-strain mechanisms (procurement lag, family engagement delay, usage-score evaluation), and recommendations for districts, TA providers, states, and funders.
- Limitations and uncertainty
- Early Adopter self-selected sample; May 2026 date; qualitative/survey synthesis without student-outcome causal estimates; pathway percentages are leader classifications, not audited instructional audits.
Idaho Title 33 Chapter 70: statewide GenAI framework required; local policies mandatory; GenAI may not replace human teachers
Idaho Code Title 33 Chapter 70 requires the State Department of Education to develop a State Board-approved statewide GenAI-in-education framework that prioritizes human-centered oversight, transparency, safety, and data security; explicitly ensures GenAI does not replace or eliminate a human teacher; and underpins local policy, PD, procurement, and standards. Each local district and public charter must adopt a GenAI policy for students and employees covering school settings and school-issued devices, defining appropriate and prohibited uses. Framework updates convene legislators, practitioners, industry, and workforce representatives.
Why it matters, evidence and limitations
- Why it matters
- Binding state floor other Western/SEA clients will cite: human-teacher non-replacement, mandatory local policies, and board-approved statewide framework as procurement and operating constraints.
- Evidence and measured results
- Opened the legislature PDF via curl+pdftotext: legislative intent, definitions of generative AI, framework duties including human-teacher non-replacement, local policy adoption requirements, and review committee composition.
- Limitations and uncertainty
- Statutory requirements without published compliance rates here; quality of local policies will vary; SDE framework document already archived separately—this resource is the binding chapter text.
How to read this edition
Source findings, measured results and limitations come from the cited publications. Patterns, operating questions, role takeaways and implementation considerations are Lighthouse Advisory interpretation, stated as questions to validate locally rather than guaranteed outcomes. Vendor and operator claims are labeled as claims. Full research method.
- Standards or public-body guidance
- Normative or advisory guidance from a standards body or public institution.
- Academic research
- Research produced through an academic institution or peer-reviewed venue.
- Independent research
- Research conducted outside the implementing organization.