NJDOE Gave Us a 749-Page Adaptive Testing Proposal. It Still Doesn’t Prove the System Works.

So we are approaching our year-long quest to get data from the NJ DOE on evidence that the new adaptive testing model rolled out last year is reliable, measures what it is supposed to, and allows the public to compare district performance to prior tests.

What we finally received was a massive, 749-page technical proposal from Cambium Assessment, the testing vendor for ELA and Math which provided an outline of the testing platform, scoring specifications, quality-control measures, psychometric work, adaptive testing structure and technical reporting.

But it’s not what we needed.

A proposal isn’t proof – it’s a promise.

In its June 23, 2026 response, NJDOE issued broad denials.

The Department indicated some of the requests were over-broad and required research. For others, it said it does not make or maintain responsive records. And for still others, the DOE indicated that the records were exempt as drafts or advisory, consultative and deliberative materials.

Most remarkably, when we asked for standard errors or score-accuracy data, NJDOE wrote:

“The Department of Education does not make or maintain records responsive to your request. Accordingly, the request is denied. Please note, this data is not yet available because it has not yet been calculated.”

The 749-page proposal was provided to us as a ‘courtesy’.

The proposal itself is evidence that we’re asking the right questions as it indicates CAI committed to producing the exact records we’ve been looking for.

-Detailed scoring, analysis, reporting specs, rules and processes.

-Simulations, scoring validation and quality-control processes which include work product intended to ensure point estimates and standard errors are on target.

-There’s an entity (HumRRO) that’s responsible for psychometric analyses, external psychometric and scoring checks and validity studies.

-Annual technical reports that deal with reliability, validity, comparability and the quality of the assessments.

To be fair, there is an important technical distinction here.

In computer-adaptive testing, item banks are often pre-equated meaning, individual test questions may carry statistical values before the test is administered. Those values help the system pick questions and estimate student performance during the test.

So the lack of a final statewide report doesn’t necessarily mean the test had no scoring mechanism at all.

The recently published performance reports provide some evidence that a scoring/reporting mechanism exists, but it isn’t the same thing as evidence that the system was validated.

Did the adaptive algorithm match the blueprint?

Were item exposure rates too high?

Were standard errors calculated and reviewed?

Were performance classifications stable?

Were scores comparable to prior fixed-form NJSLA results?

Were the results validated for accountability use?

Did the scoring checks actually work?

These tests feed school performance reports, state and federal accountability systems, district comparisons and budget debates so if this one significant tool the public uses as a basis for evaluation changes it should be able to see the machinery – we shouldn’t have to send multiple OPRA requests or play the ‘use the right words to unlock the treasure’.

By shielding Technical Advisory Committee agendas, materials, recommendations and meeting records behind a deliberative-process exemption, NJDOE is turning a standard technical review process into a black box.

While some of the material may be legally exempt the entire technical review process shouldn’t disappear from public view.

At a minimum, we should know what records exist, when they were created, who received them, what topics were reviewed and why NJDOE believes specific records are exempt.

But we aren’t giving up and we aren’t taking this to court just yet – we filed a more targeted follow-up OPRA request.

We’ aren’t asking the Department to conduct new research – just the actual deliverables, technical memoranda, simulation logs, quality-control records, specifications, reports – basically the work product connected to the proposal.

If NJDOE claims the reports themselves are exempt as drafts or deliberative materials, the state should still identify what exists including document titles, dates, authors, recipients, transmittal records, project indexes, portal logs, deliverable logs and records showing what was provided to NJDOE and when.

Until NJDOE releases the actual validation data, technical reports, comparability evidence, TAC records, deliverable logs, or document indexes showing what exists, the central question remains:

Why should New Jersey ask the public to trust a statewide testing system when the Department still has not released the technical evidence showing how that system was validated?

Source Documents:
Proposal

DOE Response to OPRA