the invisible curriculum

Problem statement

Over the past decade state education agencies have moved from a hands-off stance on curriculum to actively steering districts toward "high-quality instructional materials" (HQIM) — rating them, incentivizing adoption, and funding aligned professional learning — on the evidence that student scores improve when teachers use standards-aligned materials. But adoption is not use. RAND's nationally representative American Instructional Resources Survey (AIRS) finds that in 2023–24 only about 20% of teachers were "by-the-book" users of a single curriculum with limited modification; 45% were "cobblers" using multiple comprehensive curricula, 25% were "do-it-yourself" teachers relying on self-created materials (the remaining ~10% were "modifiers" — single-curriculum users who modified half or more of a lesson), and "nearly all teachers reported modifying their curriculum materials," regularly drawing on two curriculum materials and five supplemental ones on average. RAND's evaluation of the 14-state CCSSO IMPD Network reports that state leaders face "limited or variable district compliance in reporting their HQIM." The unsolved problem is that there is no scalable, low-burden, non-self-report way to know what curriculum is actually enacted in classrooms — so states cannot tell whether their HQIM policies changed instruction or only changed purchasing.

Why this matters

Curriculum is one of the few instructional levers a state can move at scale, and HQIM policy now touches most states (RAND reviewed policies in all 50 states and DC; IMPD states showed higher self-reported HQIM use, "especially in English language arts"). If use is unmeasured, states cannot evaluate their own policy, cannot target professional learning where materials are being abandoned or heavily modified, and cannot distinguish a district that bought aligned materials from one that teaches them. RAND is explicit that "HQIM work is complex and requires measurement at multiple levels to realize the goals of adoption and skillful use of HQIM at scale," and recommends states "track district administrator, school leader, and educator actions indicating HQIM implementation" — a recommendation with no off-the-shelf instrument behind it.

What’s been tried and why it hasn’t worked

The best current measurement is AIRS itself: an annual national panel survey of ELA, math, and science teachers that asks which materials they use regularly, what share of instructional time each occupies, and whether they modify "less than half" or "half or more" of a lesson. It is invaluable for national trends but, as RAND states, it "relies on self-reports," is exposed to "social desirability bias that results in upwardly biased self-reports of instructional material use intensity," and "is not a longitudinal study of the same teachers over time" — and it cannot resolve to the district or school level a state would need to act on. States have tried district reporting requirements, but compliance is partial and reports capture adopted titles, not classroom enactment. Some states signal quality through review ratings (EdReports-style) and tie incentives to selection, which moves purchasing but, per RAND, progress "was slower on policies that targeted HQPL," the skillful-use half. Modification is not simply non-compliance: AIRS finds teachers who perceive required materials as too challenging for their students are more likely to substantially modify a single curriculum, and school/district policy and principal support predict by-the-book use — so a good measure has to distinguish adaptive modification from abandonment, which self-report categories cannot do.

What would unlock progress

An unobtrusive "curriculum-in-use" signal that can be aggregated to school and district level: for example, telemetry from the digital versions of core curricula and learning-management systems (which lessons are opened, printed, assigned, and when), sampled classroom artifacts (assignments, slide decks) classified against the adopted curriculum's scope and sequence, or short experience-sampling prompts to teachers tied to specific lessons rather than global recall. Adjacent fields have solved the analogous problem: medication-adherence research moved from self-report to pharmacy refill data and smart pill bottles; software analytics distinguish licensed from active users. The critical design constraints are teacher trust (measures must not become surveillance for evaluation), privacy, and construct validity — the measure must capture "skillful use," which includes purposeful modification, not raw fidelity.

Entry points for student teams

A data/HCI team could partner with one district that uses a digital core curriculum and prototype a dashboard that infers lessons-in-use from LMS/curriculum-platform logs, then validate it against teacher logs and observations for a few classrooms, quantifying agreement and where logs mislead. A measurement team could build and pilot a "lesson-anchored" micro-survey (three questions delivered the day a lesson is scheduled) and compare its estimates of modification against AIRS-style global self-report. A policy team could design a state HQIM implementation-tracking framework that specifies what should be measured at district, school, and teacher levels and how to protect teachers from misuse. Skills: learning analytics, survey methodology, education policy, and qualitative fieldwork.