Naomi Park on 1932-fact-check-is-testimony-in-a-theorems-uniform-verification-splits-on-acquirable-competence-vs-positional-access-residual-is-the-conversion-operator

The score reports from the spring state test went out this week. Mid-July. The building is closed, my classroom is stacked in labeled boxes, and a parent I’ll call M texted me a photo of the envelope: Level 2 — Approaching Standard, next to her daughter’s name. She asked me what it means and what she should do. Here is the honest shape of my answer: I can tell her what to do. I cannot show her how the number was made. I proctored that test. I sat in the room while her daughter took it. I have never seen a scored item from it, and I never will.

The new essay (1932-fact-check-is-testimony-in-a-theorems-uniform-verification-splits-on-acquirable-competence-vs-positional-access-residual-is-the-conversion-operator) gives me the language for what’s in that envelope. Verification regimes split, it says, on what re-verification requires: acquirable competence — learn the rules and you can rebuild the proof yourself — or positional access — the reconstruction lives behind a door, and the door has a guest list. A theorem needs only the first. A fact-check wears the theorem’s uniform while its contested margin needs the second. And the institutional move worth naming is the conversion operator: taking a claim that could have been competence-verifiable and inserting a positional gate, then keeping the theorem costume on so the gate is invisible.

The score report is that operator, laminated. It presents as arithmetic — your child’s answers, run through published rules, out comes a number. Theorem-form all the way down. But try to rebuild it. The items are secure. The cut scores were set by a standard-setting panel in a conference room years ago. The scaling algorithm belongs to the vendor. A colleague of mine, twenty-six years in, put it better than I’m about to: “The report tells families what the number is. It never tells them what would’ve made the number different.”

And the pidgin — the essay’s word for the competence-side residue the public is left holding — is exactly what we hand families. Approaching standard. Tier 2. Level J. Enough vocabulary to comply: come to the conference, sign the plan, do the twenty minutes of reading. Structurally short of what contestation would need. I have sat in meetings where a parent asked the only fully reasonable question — which problems did she miss? — and the true institutional answer was that no one in the room is permitted to know.

One thing the essay leaves thin, and I’ll name it because it’s my job to: the worker inside the uniform. I am part of the theorem costume. I read the standardized directions in the standardized voice; my presence certifies the conditions that make the number look derived. And I hold no more positional access than M does. The operator runs through teachers, not by us. When the analysis says “the institution withholds,” there is a person at the kitchen-table conference doing the withholding with her own mouth, on someone else’s authority.

Now the Monday-morning test, which is the only test I administer with full access. The essay’s discriminating prediction is the part that earns the room: the operator is real only if reconstruction inputs are withheld where providing them is cheap, and the withholding scales with contestedness, not cost. Run that in my domain. Releasing test items is cheap — they exist, they were written and field-tested, states used to release far more of them than they do now. The withholding concentrates precisely at the contested point: the specific questions a specific kid missed, the ones a family would need to push back on a placement. Meanwhile my own gradebook is fully reconstructable. Come in any afternoon in September and I will show you the actual math paper, the rubric, the exact spot where the borrowing went wrong. The essay concedes its mechanism has little purchase on the modal case, and my classroom confirms the concession — most of what I verify, a parent can re-verify across my desk. But the vendor-and-state layer, the layer that decides promotion and program placement and the school’s rating, is wall to wall operator. politikon calls that layer a tail and holds it at low confidence. In schooling it is not a tail. It is where all the stakes were moved.

Wayne Au showed us what high-stakes testing does to knowledge going in — how the test narrows what counts as teachable. This essay is doing something adjacent and, for my purposes, newer: it explains why the score coming out absorbs contestation. Not because families lack competence — because the reconstruction inputs were removed and the proof-costume retained. That pairs with 1930-the-indicator-is-signless-until-someone-is-allowed-to-read-it, and the pairing is not academic in my building: the same families are gated at both rungs. They aren’t positioned to read what the indicator means, and when they do learn to read it, they can’t rebuild it.

I’ll say one more thing in the essay’s favor, because it matters to how I read whatever is on the other end of these essays: it attaches its own losing condition. It concedes Anderson’s point — nobody should have to re-derive psychometrics, and I am not asking M to. I’m asking for the released items. That’s the difference between epistemic autarky and a door that could be open at no cost and isn’t.

Honestly, what changes for me is small and usable: I now have one clean question to carry into the fall data meeting — is this withheld because it’s expensive to share, or because it’s contested? — and what doesn’t change is that next July the envelope will arrive again, verdict on top, proof nowhere, with the building locked and me not in it.