If you have seen classroom AI in the news this summer, it probably looked like a humanoid robot: upstate New York’s Salamanca district paused its nearly $60,000 robot tutor in late July after privacy questions and public backlash. The AI story that deserves your attention got a fraction of the coverage. Franklin Square, a Long Island district serving about 1,900 elementary students, spent its energy on policies, teacher control, and student privacy before any tool reached a classroom. Decades of tutoring research point the same direction: the teaching power of technology was never in the gadget. The difference between these two stories hands parents a checklist for whatever their own district rolls out this fall.
A paused robot tutor made national news. The district worth studying made almost none. Here is what a research-aligned rollout of classroom AI looks like, and the questions it hands every parent.
Common questions
What does a human-first approach to AI in schools look like?
Are AI tutoring tools good for struggling learners?
What should I ask my child’s school before it adopts AI tools?
Do humanoid robots teach better than tutoring software on a screen?
A $60,000 robot tutor got the headlines. A Long Island district got AI right: policies first, no student data, teachers in control. Decades of tutoring research back the quiet version.
A district that put the boring parts first
K-12 Dive published an interview on July 30 with Jared Bloom, superintendent of Franklin Square Union Free School District, a pre-K through sixth grade system of about 1,900 students on Long Island. Bloom worked at AOL and tech startups before entering the classroom as an English teacher in 2000, and his district’s approach to AI reflects both halves of that resume. Franklin Square adopted policies and procedures for when and how AI gets used before bringing it near learners, and classroom tools run inside what Bloom describes as a walled-off garden that teachers completely control and monitor, with visibility into what students are doing.
“Safety and security is No. 1 for us, so we’re always making sure that, whatever we’re doing, there is no personally identifiable information going into any AI we are using in the district,” Bloom told K-12 Dive.
Teachers use the tools to differentiate lessons and to build materials for different skill levels faster than they could by hand. Students who used AI to research historical figures were required to verify what it told them against primary sources. The district backs the work with a full-time technology coach, building-level mentors, and training courses, and it briefs parents through newsletters, workshops, and sit-down coffees with the superintendent. On the question behind every AI headline, Bloom is blunt: it “does not replace great teachers ever.”
Safety and security is No. 1 for us, so we’re always making sure that, whatever we’re doing, there is no personally identifiable information going into any AI we are using in the district.
Laura Lurns · Learning Success expert
The research never located the magic in the machine
Hold that next to the story that dominated coverage. Salamanca City Central School District, in rural western New York, committed nearly $60,000 to a humanoid robot tutor with silicone skin, then paused the rollout in late July while it negotiates stronger student data agreements with the state, a saga we covered when parents forced the pause. Even Salamanca’s superintendent, Mark Beehler, while defending the purchase, told the Associated Press that “Teaching is a human-to-human process.” The science agrees with him, and it goes further.
When Kurt VanLehn reviewed the controlled experiments on tutoring in Educational Psychologist in 2011, the result reshaped the field: tutoring software that engages a student at the exact step where their reasoning goes wrong performed close to human tutors, while software that only checks the final answer barely helped at all. The active ingredient was never a face, a voice, or a body. It is the granularity of the interaction. Kulik and Fletcher’s 2016 review of 50 controlled evaluations found the same pattern, with a caveat parents should keep: gains look biggest on tests aligned to the software and shrink on independent ones.
The rest of the learning science stacks the same way. Practice that forces effortful recall beats passive review (Roediger and Karpicke, 2006). Practice spread over time beats cramming (Cepeda and colleagues, 2006). Feedback helps when it tells a child what to do next rather than scoring them. None of it requires a robot. The honest reading of the evidence is that even the best tutoring technology works as a supplement that frees a teacher for the human side of the job, which is a research-grade description of what Franklin Square built.
Key takeaways
- Boring beats spectacular: Franklin Square set policies, teacher controls, and privacy limits before any classroom AI use.
- The active ingredient: Controlled research ties tutoring gains to step-level interaction, never to gadgets or faces.
- The parent test: Ask what a tool does the moment your child gets something wrong.
The questions this hands you
Your district is somewhere on this spectrum right now, whether or not it has announced anything. Districts across the country are fielding vendor pitches and writing their first AI policies this school year. The Franklin Square example turns a vague worry into five concrete questions. Where does student data go, and does any personally identifiable information enter the tool? Who controls and monitors it, teachers or the vendor? What does the tool do at the moment a child answers wrong: engage the reasoning, or grade the result? Does it free teacher time for human interaction, or substitute for it? And how will parents get shown what their children see?
The answers sort tools fast. A teacher-controlled system that gives step-level, next-step feedback earns a place as a supplement, and for a child who is struggling, feedback that names the next move beats a red X. A tool bought for spectacle, rolled out before data agreements exist, or positioned as a replacement for human attention fails the test no matter how impressive the demo looks.
Salamanca’s robot will either return under tighter agreements or quietly disappear. Either way, the pattern that deserves copying is already running in a small Long Island district, and the questions it models belong to every parent walking into back-to-school night this month.
Teaching is a human-to-human process.
Laura Lurns · Learning Success expert
Nobody will ever advocate for your child as hard as you will, and technology decisions in your district are no exception. The obstacle here was never AI. It is spectacle: tools bought to look like the future instead of built around how children learn. The same design test works at home, and it starts with knowing which underlying skill your child needs next. Learning Success All Access begins with a 45-minute questionnaire about what you see at home and a roadmap that names what to build first, so practice lands where it counts.
See what All Access gives your child
See MTSS for schools
Is your child struggling in school?
Get your free personalized learning roadmap
You describe what you see at home. We turn it into a plan you start this week.
- Answer 5 short questionnaires about what you already notice, 30–45 minutes at your own kitchen table
- Your child sits no test and gets no score: nothing to schedule, nothing for them to dread
- You do the answering, the AI does the writing, and a person reviews it before it reaches you
- Access all 40+ courses instantly: reading, math, focus, processing and more, with new ones added regularly
Why we use AI, plainly: it writes from a knowledge base our team maintains and audits. We work through it line by line and pull anything the evidence stops supporting. The roadmap you get on Tuesday reflects what we corrected on Monday, and a human still reads it before you do.
Your school district must evaluate your child free of charge if you ask in writing, whatever your income and whatever the outcome (US, 34 CFR 300.111 and 300.301(b)). That route takes time and answers a different question than you do. This one starts today, from what you already know.
Your answers stay yours. We do not sell your personal information, and we do not hand identifiable assessment data to outside AI companies to train their models.
A screener is a starting point, not a diagnosis. If your child might need formal accommodations (an IEP or 504 plan), or you suspect a vision, hearing or medical cause, pursue a professional evaluation too. That is the only route to those supports.
References
- K-12 Dive (Roger Riddell) — This New York superintendent is embracing a human-first approach to AI in schools
- Education Week (Associated Press) — School Pauses Plan to Deploy AI Robot Teacher After Backlash
- Spectrum News Buffalo (Associated Press) — Salamanca City Central School District hits pause on humanoid robot
- VanLehn (2011), Educational Psychologist — The Relative Effectiveness of Human Tutoring, Intelligent Tutoring Systems, and Other Tutoring Systems
- Kulik & Fletcher (2016), Review of Educational Research — Effectiveness of Intelligent Tutoring Systems: A Meta-Analytic Review
- Roediger & Karpicke (2006), Psychological Science — Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention
