K-12Issue 134 records
K–12
Fresh K-12 signal pairs state staff-AI governance supply with learning-integrity caution. OSSE’s September 2026 LEA AI Model Policy for Staff Use gives DC a customizable stoplight that Red-bars AI for surveillance, student discipline, teacher evaluation, and IEP/504 eligibility; Yellow allows oversighted grading/IEP drafting; Green covers planning and operations—enterprise tools only for PII, with pre-use and annual training. That policy answers OSSE’s own February 2026 LEA survey (39 LEAs): only 45% had staff AI policies and 34% student policies, with model instructional guidance ranked the top SEA need. Same-day operating tension appears in Montgomery County, where Bethesda Magazine (Sept 18) reports teacher Wild West norms, board-mandated emerging-tech vetting, PTA calls to block student GenAI until learning value is shown, and only a Canva image generator approved for students pending review. A CEPR/NBER Chinese secondary panel (n=26,811) newly frames those fights: GenAI can lift homework scores ~18% while cutting closed-book exams ~20%, with losses concentrated in homework outsourcing. Together: ship staff stoplight controls and training gates; evaluate student homework AI on unaided learning, not packet speed.
- Evidence records
- 4
- Cross-source patterns
- 2
- Evidence classes
- 1 standards or public-body guidance1 government evaluation1 academic research1 independent reporting
- Outcomes
- 2 emerging1 mixed1 cautionary
- Source freshness
- 2 older, newly relevant1 recent1 new this fortnight
- Research completed
- 2026-09-19
Choose a role to see its takeaway beside every record in the ledger.
Patterns across the evidence
Synthesis · Lighthouse Advisory interpretation
SEA staff AI model policies are landing to fill documented LEA readiness gaps
DC’s LEA survey showed only 45% staff and 34% student AI policies, with model instructional guidance ranked the top SEA need; OSSE’s September staff stoplight policy is the direct operational response. Together they show governance supply following measured readiness demand—not waiting for student-outcome RCTs.
Operating questionBefore enabling staff AI features, do we have written local policy coverage, enterprise allowlisting, and training gates matching SEA model expectations?
Supporting evidenceOSSE with Regional Education Laboratory at MathematicaOffice of the State Superintendent of Education (OSSE), District of Columbia
Assisted homework productivity is a weak proxy for retained learning when outsourcing dominates
Large-scale Chinese secondary evidence finds GenAI can raise homework scores while cutting closed-book exam performance, concentrated among outsourcing users; Montgomery County parent leaders simultaneously argue AI’s purpose is to remove the mental struggle that produces learning. U.S. operators should evaluate classroom AI on unaided outcomes and effort preservation, not homework throughput.
Operating questionWhat unaided or delayed assessments and anti-outsourcing controls prove our homework AI is producing retained learning rather than completed packets?
Supporting evidenceCEPR Discussion Paper / NBER conference paper (Strömberg, Lei, Wu)Bethesda Magazine
Evidence ledger
Full record · every source keeps its link and limitations
OSSE releases DC LEA AI Model Policy for Staff Use with stoplight prohibitions on discipline, teacher eval, and IEP eligibility
OSSE’s September 2026 Model AI Policy for Staff Use gives DC LEAs a customizable stoplight: Red bars AI for physical/biometric surveillance, student discipline decisions, teacher performance evaluations, and IEP/Section 504 eligibility determinations; Yellow allows limited use with enhanced oversight (e.g., drafting IEP language, reviewing/grading student work, device monitoring); Green covers lesson planning, tutoring plans, communications, and logistics with human review. Staff must use only approved enterprise tools for PII, complete training before use and renew annually, and keep humans accountable for outputs. The policy explicitly does not set student-use or procurement rules.
Why it matters, evidence and limitations
- Why it matters
- A transferable SEA staff-use template for districts that need clear prohibited vs. conditional AI use cases, enterprise-tool mandates, and FERPA/COPPA/CIPA/IDEA-aligned vendor expectations without waiting for binding statewide student-AI rules.
- Evidence and measured results
- Opened the September 2026 OSSE PDF booklet end-to-end (executive summary, guiding principles, stoplight tables, Red/Yellow/Green narratives, training and procurement recommendations). Cross-checked with OSSE’s Sept 1, 2026 press release describing the same Red/Yellow/Green categories and 45% staff-policy baseline from the LEA survey.
- Limitations and uncertainty
- Voluntary model policy, not legal advice and not mandatory adoption. Staff-only scope—student GenAI and tool procurement remain local LEA work. Evidence of instructional effectiveness is acknowledged as limited (cites Stanford SCALE 2026 review).
OSSE/REL Mathematica LEA survey: only 45% of DC LEAs had staff AI policies; model guidance ranked top need
In February 2026, OSSE and REL Mathematica surveyed DC’s 69 LEAs (59 responses collapsed to 39 unique LEAs). Only 45% reported written staff AI guidance and 34% reported student-use guidance. Procurement maturity was uneven (formal review 41%, informal/ad hoc 26%, not procuring 28%, unsure 5%). LEAs ranked model instructional AI policy as the most useful OSSE support, ahead of privacy guidance and educator training. Top adoption challenges included data privacy/security, tool quality, lack of federal/state guidance, bias, staff readiness, and uncertain instructional value. Why now: OSSE’s Sept 1, 2026 staff model policy was built explicitly in response to this gap and was not yet in the SLED k12 archive.
Why it matters, evidence and limitations
- Why it matters
- Quantifies the policy-readiness gap that SEAs and large districts must close before scaling staff AI—and explains demand for model policies over procurement templates alone.
- Evidence and measured results
- Opened the May 2026 PDF brief: policy prevalence (45%/34%), procurement pie chart categories, ranked desired supports, use-case interest bars, and challenge severity distributions. Cross-checked with OSSE press release citing the same 45% staff-policy figure as the impetus for the September model policy.
- Limitations and uncertainty
- Self-report LEA-leader survey (39 LEAs), not classroom outcome measurement; published May 2026 (pre-policy) so pair with the September model policy for current operational meaning. Interest bars are not evidence of effective use.
CEPR/NBER evidence: GenAI raises homework scores but cuts closed-book exam learning, concentrated in homework outsourcing
Using 30 months of panel data on 26,811 Chinese students in grades 7–12, staggered GenAI adoption raises homework scores ~18% and cuts completion time ~30%, but lowers monthly closed-book exam scores ~20% within six months; high-stakes entrance exams fall ~18–24% with the full penalty emerging after about two years. Losses concentrate among roughly 80% of AI users whose behavior matches homework outsourcing (very short completion time with high homework scores); users who keep similar time-on-task see small losses. Why now: U.S. districts debating bans on AI-completed assignments (e.g., APS draft; Montgomery PTA concerns about removing mental struggle) need measured evidence that assisted homework productivity is not retained learning.
Why it matters, evidence and limitations
- Why it matters
- Gives SLED buyers a quantitative caution against treating homework speed/score gains as proof of learning, and supports instructional designs that preserve effortful practice and off-tool assessment.
- Evidence and measured results
- Opened the NBER conference PDF (full paper) and the CEPR DP21577 abstract page. Verified abstract statistics (n=26,811; 18%/30%/20%; entrance-exam 18 and 24%; ~80% outsourcing concentration) and ethics approval note. CEPR hosted PDF URL returned an HTML paywall/login page, so the conference PDF is the verification object.
- Limitations and uncertainty
- China secondary context; DiD/staggered adoption not a U.S. RCT; college entrance exams are out of pure K-12 scope except as long-run learning signal; transferability to U.S. curricula and tool ecosystems is limited and must be stated.
Bethesda Magazine: Montgomery County AI policy/regulations advance amid teacher Wild West and PTA push to block student GenAI
Fresh Sept 18 reporting describes Montgomery County classrooms improvising AI practice in a guidance vacuum while county and state policy catch up: MCPS Board Policy IGS (May 2026) directs formal vetting of emerging technologies for accuracy, bias, privacy, security, and educational value, with implementing regulations expected mid-to-late September; Maryland’s AI Ready Schools Act and MSDE guidance add statewide literacy and collaborative structures. MCCPTA and regional parent coalitions urge blocking student-facing GenAI until learning value is shown, arguing AI removes the mental struggle that produces learning. Teachers and students report mixed classroom norms; disability advocates seek IEP-accessible AI. Officially, MCPS says only a Canva image generator is approved for student use pending extensive AI tool review.
Why it matters, evidence and limitations
- Why it matters
- Shows the live U.S. operating tension among board vetting mandates, parent precautionary blocks, teacher improvisation, and accessibility claims—exactly the change-management problem SLED operators face after headline AI rules.
- Evidence and measured results
- Opened the full Bethesda Magazine article (David Montgomery, Sept 18, 2026). Verified Policy IGS language on AI supporting rather than replacing teacher-led instruction and requiring formal vetting; Lopez quotes on regulations timing and Canva-only student approval; MCCPTA/Betsy Tao struggle quote; Holton-Arms private-school literacy practices as contrast.
- Limitations and uncertainty
- Single-county long-form journalism with interviews; not a representative survey or causal evaluation. Private-school examples are not MCPS policy. Regulations’ final text not retrieved in this run.
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.
- Government evaluation
- A public body’s measured evaluation or documented pilot.
- Academic research
- Research produced through an academic institution or peer-reviewed venue.
- Independent reporting
- Independent reporting with attributable sources but without a formal evaluation design.