K-12Issue 363 records
K–12
Fresh K–12 signal centers on policy and procurement gaps more than new learning RCTs. Connecticut’s October 2026 CCERC evaluation of a six-district AI pilot finds skill-building success concentrated in teacher preparation and teacher-designed classroom supports, with accuracy/math-error concerns and student ambivalence. Same-day Compass Vermont reporting shows state grade-banded chatbot discouragement without tracking which schools enable Gemini, and thin local AI policy uptake. Prince George’s County’s required Artificial Intelligence Declaration Agreement illustrates a concrete vendor control that forces AI disclosure, no-retain student-data terms, disable switches, and audit rights.
- Evidence records
- 3
- Cross-source patterns
- 1
- Evidence classes
- 1 government evaluation1 independent reporting1 standards or public-body guidance
- Outcomes
- 1 mixed1 cautionary1 emerging
- Source freshness
- 2 undated1 new this fortnight
- Research completed
- 2026-10-13
Choose a role to see its takeaway beside every record in the ledger.
Patterns across the evidence
Synthesis · Lighthouse Advisory interpretation
Classroom AI skill gains and developmental guidance still outrun clear local parameters and enforceable vendor AI declarations
Connecticut’s pilot evaluation ties engagement to clear, consistent AI-use parameters and teacher-designed supports, while Vermont reporting shows non-mandate guidance without statewide Gemini enablement tracking and historically thin local AI policies/privacy agreements. PGCPS’s AIDA shows one large district forcing vendor AI disclosure, no-retain student PII terms, disable rights, and audit/remediation into purchasing—controls many LEAs still lack.
Operating questionBefore expanding student-facing generative AI, do districts publish grade-banded use rules and require signed AI declarations that ban retaining student prompts?
Supporting evidenceCenter for Connecticut Education Research Collaboration (CCERC)Compass VermontPrince George’s County Public Schools
Evidence ledger
Full record · every source keeps its link and limitations
CCERC October 2026: Connecticut AI pilot evaluation finds preparation/differentiation gains, math-error wariness, and demand for clear AI-use policies
October 2026 CCERC executive summary evaluates Connecticut’s state-funded AI-in-education pilot across six LEAs (East Hartford, Lebanon, Rocky Hill, Seymour, Westport, Odyssey Community School) in grades 5–12 during 2024–25. Mixed methods: student survey n=735, teacher survey n=58, student focus groups n=32 (3 districts), teacher focus groups n=24 (4 districts). Pilot generally helped students and teachers develop AI skills. Teachers found tools most useful outside class for material preparation and differentiation; in class, teacher-created content-specific resources (e.g., rubric chatbots) were most useful. Both groups raised ethics, accuracy, and reliability concerns; math-related AI mistakes were frequent; most students were ambivalent about expanded use. Recommendations emphasize PD/peer collaboration, AI literacy, content-specific teacher-leveraged use, and clear district AI-use policies.
Why it matters, evidence and limitations
- Why it matters
- High for K–12: a current state-commissioned evaluation of multi-district AI pilots with actionable recommendations on PD, content-specific use, and policy clarity.
- Evidence and measured results
- Executive-summary mixed-methods evaluation funded via CCERC at CSDE request; surveys and focus groups; not an achievement RCT.
- Limitations and uncertainty
- Executive summary only; self-report methods; six pilot districts; no learning-outcome effect sizes.
Compass Vermont: state discourages ChatGPT-style chatbots through grade 8 but does not track which schools enable Gemini
Published Oct 12, 2026. Reports Vermont Agency of Education advises keeping ChatGPT-style chatbots away from students through eighth grade via January 2026 non-mandate guidance with grade bands (no chatbots PreK–2; no open-ended chatbots grades 3–5; education-specific only under teacher direction grades 6–8; piloted use grades 9–12). Agency told Compass it does not collect which systems enable Gemini or at what grades. Latest published tech survey (2023–24) found about 9% (5 of 54) of school systems had any AI policy/notice; 16 of 54 did not require vendors to sign student data privacy agreements. Notes Google Classroom Gemini features and monitoring gaps (teacher visibility; Docs/app Gemini not logged). Contrasts Boston teach-AI vs NYC/LA restrictions; flags pending/failed VT bills H.650/H.830 and Senate interest in a model AI policy.
Why it matters, evidence and limitations
- Why it matters
- High for K–12 governance: documents the gap between developmental guidance and operational visibility into student Gemini enablement and local AI policy coverage.
- Evidence and measured results
- Independent reporting citing Agency of Education guidance, emailed agency statements, and the state’s 2024 Annual Technology Survey.
- Limitations and uncertainty
- Survey wave predates January 2026 guidance; 2025 survey unpublished; article does not independently verify each LEA’s current Gemini settings.
PGCPS Artificial Intelligence Declaration Agreement (AIDA): mandatory vendor AI disclosure, no-retain student PII, disable rights, and audit/remediation
Prince George’s County Public Schools Purchasing Appendix P requires every digital-tool vendor to complete an Artificial Intelligence Declaration Agreement (AIDA). If AI is present, vendors must disclose models/training-data sources/local-vs-cloud storage/improvement methods; primary user audience; FERPA/COPPA/state-privacy confidentiality verification; Student Data points used for personalization; verification that generative models will not retain/store Student Data or PII entered during use and that outputs are not linked to individuals; ban on secondary use without PGCPS and parental approval (termination risk); bias/privacy/harm monitoring with results shareable on request; enable/disable directions; explainability for advancement-impacting recommendations; and third-party processor disclosure including geography. Vendors must notify within 60 days of new AI-containing versions; PGCPS may enable/disable AI features, terminate for unilateral T&C/privacy changes, audit AI systems/data practices, and require remediation within 60 days at vendor cost. AIDA supersedes vendor AI policies. Why now: newly relevant alongside Vermont reporting of uneven privacy-agreement practice—shows a concrete district-level AI declaration control not previously archived.
Why it matters, evidence and limitations
- Why it matters
- High for K–12 procurement: a reusable contract addendum that forces AI feature disclosure and student-data retention limits before tools ship.
- Evidence and measured results
- Official district purchasing PDF (Appendix P); contractual requirements without published compliance statistics.
- Limitations and uncertainty
- Undated PDF; no published rates of vendor completion or audit findings; filename misspells ‘declaration’.
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.
- Government evaluation
- A public body’s measured evaluation or documented pilot.
- Independent reporting
- Independent reporting with attributable sources but without a formal evaluation design.
- Standards or public-body guidance
- Normative or advisory guidance from a standards body or public institution.