K-12Issue 435 records
K–12
Tonight's K-12 signal is that AI access is being decided as device configuration while the evidence for paying more for 'AI' stays thin. A large pre-registered trial in rural China, newly out in a peer-reviewed journal, found an AI-driven adaptive math program did no better than simpler software or none for 8,647 primary students, though it tested 2021 adaptive software rather than chatbot tutors. Issaquah, Washington now gates student AI by grade band through a monitoring vendor and a Copilot study mode, while a Saint Paul coalition asks its district to switch Gemini off entirely. Nova Scotia shows what full local discretion produces: uneven rules and teachers returning to paper. Peninsula, Washington built its own teacher agent and claims savings without a stated cost basis.
- Evidence records
- 5
- Cross-source patterns
- 2
- Evidence classes
- 3 independent reporting1 academic research1 standards or public-body guidance
- Outcomes
- 3 cautionary2 emerging
- Source freshness
- 4 new this fortnight1 recent
- Research completed
- 2026-10-20
Choose a role to see its takeaway beside every record in the ledger.
Patterns across the evidence
Synthesis · Lighthouse Advisory interpretation
Student AI access is being set by grade band at the device layer, and those settings are now contested
Issaquah enforces no core AI below grade 6 and monitored AI chat above it; a Saint Paul coalition, building on a national letter, asks for Gemini to be disabled on devices entirely; Nova Scotia's regions vary AI use by grade and centre with no provincial rule.
Operating questionCan we show, per grade band, which AI features are enabled on student devices, who decided, and what is logged?
Supporting evidenceIssaquah School District 411Intentional Technology Saint Paul (coalition of parents, teachers and community members), published by Midway Como Frogtown MonitorCBC News (Nova Scotia); quotes from Department of Education and Early Childhood Development, Nova Scotia Teachers Union and Mount Saint Vincent University
AI value claims are outrunning outcome evidence
The week's strongest causal study finds no gain from AI personalization; Peninsula reports savings and broad teacher use without a cost basis or measurement; a Nova Scotia parent group argues schools assume benefit without data and seeks a pause.
Operating questionBefore scaling or renewing an AI product, what comparison against a simpler option and what cost baseline will we require?
Supporting evidenceYue Ma, Tianli Feng, Robert Fairlie, Chengfang Liu, Prashant Loyalka, Scott Rozelle, Xinwu Zhang (Stanford, UESTC, UCLA/NBER, Peking University, NUIST); CESifo Working Paper 12837Peninsula School District, reported by District Administration (DA Changemakers podcast)CBC News (Nova Scotia); quotes from Department of Education and Early Childhood Development, Nova Scotia Teachers Union and Mount Saint Vincent University
Evidence ledger
Full record · every source keeps its link and limitations
Large rural China RCT: an AI-driven adaptive math program did no better than non-adaptive software or no software for 8,647 primary students
Researchers randomly assigned 315 grade 4–6 classes (8,647 students) in 105 rural Sichuan schools to an AI-driven adaptive math program (AI-CAL), a non-adaptive program matched on content, time, hardware, gamification and feedback (R-CAL), or no CAL. Students used CAL 80 minutes a week for 11 weeks (March–June 2021). In the preferred specification AI-CAL vs R-CAL was −0.040 SD (not significant; effects above 0.01 SD ruled out) and AI-CAL vs pure control was −0.036 SD (SE 0.027). Attrition was under 1% per arm. Teachers cited older hardware and low student motivation. The authors conclude paying a premium for adaptive features is hard to justify in these settings. Why now: the peer-reviewed journal version was indexed as published October 17 and surfaced after last night's research, and it is one of the few large causal tests of AI personalization.
Why it matters, evidence and limitations
- Why it matters
- Directly relevant to district purchasing of 'AI-personalized' practice software: the strongest new causal result this week is a precisely estimated null.
- Evidence and measured results
- Pre-registered cluster RCT with active and pure controls, cluster-robust standard errors, minimal attrition and a replication package. Outcome is a curriculum-aligned math test at the end of 11 weeks; no long-term retention measured.
- Limitations and uncertainty
- Intervention ran in 2021 and tested pre-generative adaptive learning, not chatbot or LLM tutors. Rural, under-resourced Chinese primary schools; results may not transfer to U.S. districts or higher-engagement settings. Heterogeneity analyses were exploratory. The two programs were different products, not one program with adaptivity switched off. Journal version not opened (ScienceDirect returned HTTP 403); findings come from the CESifo working paper.
Issaquah School District introduces grade-banded student AI: no core AI for K–5, Securly AI Chat for 6–12, Copilot set to Study and Learn for high school
Beginning this fall, Issaquah is introducing district-supported AI tools for grades 6–12, selected for privacy, safety, age appropriateness and educational value. K–5 students have no access to core AI platforms and restricted access to others on district devices and network. Middle and high school students use Securly AI Chat; Securly also filters internet access and supports classroom screen monitoring. High school students also get Microsoft Copilot configured to default to Study and Learn. Teachers may use Colleague AI's Classroom feature (University of Washington National R&D Center of AI in Classroom Instruction) and decide when and how AI is used. AI literacy is taught in grades 6 and 9; a Tech Advisory committee will gather stakeholder input.
Why it matters, evidence and limitations
- Why it matters
- A concrete district configuration that ties student AI access to grade bands, a filtering and monitoring vendor, and a default study mode.
- Evidence and measured results
- District announcement; no outcome data, contract terms, cost or data-flow detail.
- Limitations and uncertainty
- Self-description by the district; no measured results. Specific product behavior (e.g., Study and Learn defaults) not independently verified.
CBC: Nova Scotia leaves classroom AI rules to regional centres, producing uneven guidelines; teachers' union reports a shift back to paper assessments
Nova Scotia has no provincewide classroom AI policy; the province leaves rules to regional centres. CBC found Halifax and Chignecto-Central have written guidelines, Cape Breton-Victoria varies use by grade with teacher judgement, Strait is finalizing guidelines, and the francophone board has a committee but no policy. A senior department official called provincewide guidelines an 'ongoing conversation.' The teachers' union president said many teachers have reverted to in-class paper assessments and that, apart from a voluntary summer camp, there is no provincewide AI-focused professional development. A parent group is petitioning for a New York-style pause.
Why it matters, evidence and limitations
- Why it matters
- Mirrors the U.S. pattern of local discretion without state rules, and adds a union account of assessment changes and missing training.
- Evidence and measured results
- News report based on responses from each regional centre and on-record interviews; no survey or outcome data.
- Limitations and uncertainty
- Canadian jurisdiction with different governance from U.S. states; claims about chatbot-written work are the union president's account, not measured. Opened via a fetch tool because direct retrieval from the agent's host timed out.
Saint Paul parent-teacher coalition asks SPPS to fully disable YouTube and Gemini AI on school devices, extending a national open letter to the 150 largest districts
The coalition's letter to Superintendent Stacie Stanley and the SPPS board acknowledges steps SPPS has already taken to limit YouTube and Gemini AI on school-issued devices and urges the district to fully disable both and to include a systematic review of student internet use in its iPad evaluation and AI policy work. It forwards the October 5 national open letter from 13 organizations and experts to the 150 largest districts (about 12.5 million students), which argues litigation and reporting show unacceptable risk, says Google enabled Gemini for students without asking schools or parents first, and says Google shifts legal risk to districts through indemnification. A filtering vendor's October 11 response (Lightspeed) argues blocking is a configuration choice, not the only answer, and lists equity and visibility costs of switching tools off.
Why it matters, evidence and limitations
- Why it matters
- Shows advocacy pressure moving from the largest districts to mid-size ones, framing AI access on devices as a liability and contract question.
- Evidence and measured results
- Advocacy letter; the national letter's legal and product claims are the signatories' assertions, not adjudicated findings. The vendor response is visibly attributed vendor material with its own usage data.
- Limitations and uncertainty
- Opinion and advocacy; no independent verification of litigation claims here. SPPS's response and current settings are not reported. The vendor counterpoint comes from a company that sells filtering tools.
Peninsula School District (Wash.) builds its own agentic AI assistant for teachers, claims nearly $300,000 saved versus buying
Peninsula built an in-house agentic AI assistant for educators so it could embed district student and teacher data and its own educational frameworks, and rolled it out this school year. According to Kris Hagel, most teachers use it to offload administrative work (one built a QR-code classroom checkout), and the district saved nearly $300,000 by building rather than buying. He credits years of work on foundational data systems and advises leaders to get their 'data house' in order first.
Why it matters, evidence and limitations
- Why it matters
- A rare district build-versus-buy example for agentic AI, with the data-readiness prerequisite stated plainly.
- Evidence and measured results
- Operator claims relayed by trade press from a podcast; no usage metrics, cost basis for the savings figure, security review or evaluation.
- Limitations and uncertainty
- Single-source operator claims; savings figure has no stated baseline; details deferred to a January conference talk.
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.
- Independent reporting
- Independent reporting with attributable sources but without a formal evaluation design.
- Academic research
- Research produced through an academic institution or peer-reviewed venue.
- Standards or public-body guidance
- Normative or advisory guidance from a standards body or public institution.