Layer 1 / 7Concepts & question
Frame the question
Set the purpose, the question worth reasoning about, and the concepts the whole house rests on.
Purpose
Decide whether K-12 schools should adopt AI tools, and under what guardrails, so the reasoning holds up to teachers, parents, and administrators.
Overarching question
Should AI be used in schools?
Key concepts
Academic integrityEquitySupervisionData privacy
Layer 2 / 7Stakeholders
Build the perspectives
Reason from each stakeholder in turn. Assign perspectives to co-builders so the work divides cleanly.
6 perspectives
StudentsMRMaya R.78
Students benefit most when AI acts as a tutor that explains its steps, not an answer key that finishes the work for them.
Sub-questions · 3
Does AI help students learn, or help them avoid learning?
Splits on supervision. Guided, worked-example use raises outcomes; unsupervised use rewards shortcutting.
Who gets left behind when tools cost money?
Access gaps widen unless districts fund equal access, so equity has to be designed in from the start.
How do students with disabilities and ELLs fare?
Strongest-gain group. Scaffolding, read-aloud, and translation lower long-standing barriers.
Supporting evidence
Intelligent tutoring systems raised test scores by a median of 0.66 standard deviations across 50 controlled evaluations.
Counterarguments
Over-reliance may erode independent writing and problem-solving over a 1-2 year horizon.
Self-reported engagement can mask shallow, test-brittle learning.
TeachersMRMaya R.64
Teachers gain real prep and feedback leverage, but inherit new assessment-design and oversight responsibilities.
Sub-questions · 3
Does AI save teacher time or add oversight burden?
Net time saved on prep and first-draft feedback; time added on verifying authorship and integrity.
Can teachers reliably detect AI-assisted work?
Detection alone is unreliable. Redesigning tasks for process and defense matters more than catching output.
What training does this actually require?
Ongoing professional development, not a one-time rollout, or the tool adds load instead of removing it.
Supporting evidence
25 percent of U.S. K-12 teachers used AI tools for instructional planning or teaching in the 2023-24 school year.
Counterarguments
Without sustained training, tools add workload rather than removing it.
Assessment validity is threatened unless tasks are redesigned around reasoning.
ParentsDKDevan K.55
Parents are cautiously supportive when given transparency about data use and a low-friction opt-out.
Sub-questions · 2
What happens to my child's data?
Concern centers on third-party sharing and retention windows, not the tutoring itself.
Can I opt my child out without penalty?
Opt-outs are essential to trust and must not create a second-class classroom experience.
Supporting evidence
The FTC's 2025 COPPA Rule amendments require separate parental consent before children's data can be used for targeted advertising or shared with third parties.
Counterarguments
Opt-outs can quietly create a two-tier classroom.
Transparency promises are hard for a parent to verify in practice.
SocietyDKDevan K.60
Societal benefits are plausible but, today, distributed unevenly by district funding and geography.
Sub-questions · 2
Does AI in schools widen or close gaps?
Either outcome is possible depending on funding equity and procurement terms.
What is the civic risk of AI-shaped learning?
Homogenization of reasoning if every classroom leans on a few converging models.
Supporting evidence
Fewer than 10 percent of schools and universities surveyed worldwide have formal guidance on using generative AI.
Counterarguments
A policy vacuum tends to hit low-resource districts hardest.
EmployersYOYou72
Employers strongly value AI-fluent graduates and are already re-bundling entry-level roles around oversight.
Sub-questions · 2
Which skills do employers actually want?
Judgment, verification, and framing over raw generation.
Are entry-level roles disappearing?
Shifting rather than vanishing. Tasks are re-bundled toward review and quality control.
Supporting evidence
Entry-level task composition is shifting toward oversight and review of AI output.
Counterarguments
Employer demand may outpace how fast curricula can change.
AdministratorsYOYou58
Administrators face a hard balance of cost, liability, and measurable outcomes on a district budget.
Sub-questions · 2
What is the true total cost of ownership?
Licenses plus training plus compliance and support, not the sticker license price.
Who is liable when the AI is wrong?
Unsettled. Procurement contract terms carry most of the real risk allocation.
Supporting evidence
Districts must meet the updated COPPA Rule's consent, retention, and security obligations by April 2026, adding real compliance lift to procurement.
Counterarguments
Measurable outcomes tend to lag adoption by several years.
Layer 3 / 7Sourced facts
Ground it in evidence
Add facts with citations. Research Mode finds sources for you, and every claim links back to something checkable.
3 sourced facts
YO
Intelligent tutoring systems raised test scores by a median of 0.66 standard deviations across 50 controlled evaluations.
YO
25 percent of U.S. K-12 teachers used AI tools for instructional planning or teaching in the 2023-24 school year.
YO
Fewer than 10 percent of schools and universities surveyed worldwide have formal guidance on using generative AI.
Layer 4 / 7Foundations
Surface the assumptions
Name what has to be true for the reasoning to hold. Weak footings show up here first.
4 foundational assumptions
01AI systems are accurate enough for educational use under teacher supervision.YO
02Technology adoption in schools will continue to expand over the next decade.DK
03Districts can fund ongoing teacher training, not just one-time rollouts.MR
04Student-data privacy frameworks will mature alongside the technology.YO
Layer 5 / 7Where it lands
Draw the conclusion
State the central conclusion and the reasoning that carries the perspectives into it.
Central conclusion
AI is a net benefit in schools when strong guardrails exist: teacher supervision, ongoing training, and FERPA-compliant procurement.
Equity safeguards and opt-outs are essential to prevent widening gaps between high- and low-funding districts.
Reasoning summary
Across all six perspectives, evidence converges on net benefit when guardrails are present. Where guardrails are absent, risks of dependency, inequity, and privacy breaches grow quickly enough to erase the gains. The conclusion holds most strongly for supervised tutoring use, and weakens for unsupervised assessment.
Layer 6 / 7Consequences
Trace the implications
Map what follows if the conclusion holds, sorted by how positive, negative, or uncertain each consequence is.
8 implications mappedSorted by register and tagged with time horizon and who it lands on.
Positive · 3
Personalized tutoring at scale, especially for students with disabilities and ELLs
Near-termStudents
Meaningful reduction in teacher prep and feedback time
Near-termTeachers
Stronger workforce readiness for AI-fluent roles
Long-termSociety
Negative · 3
Erosion of independent writing and reasoning if unsupervised
Long-termStudents
Widening divide between high- and low-funding districts
Long-termSociety
Compliance and liability exposure under student-privacy laws
Near-termDistricts
Uncertain · 2
Long-term impact on critical-thinking habits beyond a 1-2 year horizon
Long-termStudents
Whether free AI tools will close or widen the access gap
Long-termSociety
Signals to watch · would change the conclusion
→Unsupervised assessment use rising faster than teacher training can keep pace.
→Vendor pricing climbing sharply once districts are locked in.
→Early evidence that heavy AI reliance dampens independent problem-solving.
Layer 7 / 7Score & publish
Review house strength
See how the house scores across evidence, logic, and coverage, what is driving each number, and what would raise it.
68/ 100
Solid, with gaps
A solid house with clear soft spots. The steps below are the fastest way to firm it up.
The three scores
EvidenceDeveloping68
How well each claim is backed by a cited, checkable source.
Driving this score3 sourced facts
LogicDeveloping66
Whether assumptions are surfaced and the conclusion follows from them.
Driving this score4 assumptions, conclusion set, 8 implications
CoverageDeveloping70
The range of stakeholder perspectives the house accounts for.
Driving this score6 perspectives
How the overall is weightedEvidence 40% · Logic 35% · Coverage 25%