From wrong answer to root cause — in classroom time.

The problem

At the start of the 2025 school year, the 7th grade math team at Cascade Ridge had a familiar problem: MAP scores from spring showed 38% of incoming students below the grade-level RIT threshold for algebra readiness. But the math department head, who had been teaching 7th grade math for 12 years, wasn't satisfied with that data alone.

"I know what a RIT score tells me — it tells me where a student ranks. It doesn't tell me why they're struggling with variables. There's a huge difference between a kid who's never seen a variable before and a kid who thinks variables are abbreviations for object names. Same score, completely different conversation."

The department reached out to Brainpathio after one teacher saw a blog post about the variable-as-label misconception and recognized it immediately as something she saw constantly in her classroom but had never had a name for.

What the engine found

After an 8-week deployment across four classrooms, the Brainpathio engine had processed 3,800 student responses and generated misconception alerts for 67 distinct student-misconception pairs — meaning specific students with specific root cause hypotheses, not aggregate percentages.

The most prevalent single misconception across the cohort was variable-as-label confusion (24 students), which the engine classified with 81% or higher confidence in 19 of those 24 cases. This misconception — where students treat a variable like x as an abbreviation for a word (e.g., "x means the number of apples") rather than as a quantity that can vary — blocks correct use of algebraic manipulation rules.

The second most common finding was additive reasoning interference in multiplicative contexts: students applying repeated-addition logic to proportion and rate problems (18 students). This finding matched published research from Confrey's ratio and proportion work that Brainpathio's taxonomy references.

Crucially, 11 students who had below-grade-level MAP scores showed zero misconception flags — suggesting their issue was procedural fluency gaps, not conceptual errors. The math team was able to route those students to different intervention resources, saving them from ineffective reteach sessions.

What teachers did with the data

The department used their Monday collaboration period differently for the last 4 weeks of the pilot. Instead of reviewing aggregate test results, they used the Brainpathio alert feed to plan targeted small-group sessions. Three teachers ran "variable meaning" re-anchoring sessions based on a protocol Priya Narayan developed from the research literature on variable confusion interventions.

By week 6, the engine's re-assessment of previously flagged students showed the variable-as-label confidence scores dropping below the alert threshold for 14 of the 19 high-confidence cases — suggesting the conceptual model had shifted.

What we learned from this school

The Cascade Ridge pilot was our first large-scale test of the engine's confidence thresholds. We discovered that our initial threshold of 70% confidence was generating too many borderline alerts that distracted teachers from the high-confidence cases. After this pilot, we recalibrated the default alert threshold to 78%, which reduced alert volume by 31% and increased teacher-rated "actionability" from 67% to 84%.

The problem

Jefferson STEM Academy is a Title I school with a strong emphasis on NGSS-aligned science. The 8th grade physical science teachers were preparing students for the NGSS performance expectations on forces and motion (MS-PS2) and were seeing consistent failures on problems that required students to explain why objects stop moving.

This is one of the most well-documented misconceptions in science education research: the Aristotelian theory of motion, where students believe that motion requires a continuous applied force. Newton's First Law — that objects in motion stay in motion unless acted upon by an unbalanced force — runs directly counter to students' everyday experience and intuition.

The science team wanted to know two things: which students held this belief strongly enough to be blocking NGSS understanding, and which students had already developed a correct Newtonian model but were failing on problem framing or vocabulary, not conceptual understanding.

What the engine found

Over the 8-week deployment, the engine processed 4,100 student responses across three physical science classrooms. The force-motion misconception (Aristotelian motion model) was the most prevalent finding in the cohort, flagging 31 students — roughly 34% of the cohort — with confidence above the alert threshold.

A secondary finding was the agent-causation bias (12 students), where students attributed motion only to visible, intentional agents (people pushing or pulling) and could not model passive forces like gravity or contact forces between stationary surfaces. These students passed problems with obvious actors but failed problems involving falling objects or surfaces exerting force.

Of the 91 students, 18 showed no misconception flags and had correct Newtonian reasoning on assessment items — their test failures appeared to be vocabulary and notation problems, not conceptual. This was a significant finding for the school: nearly 20% of their "failing science students" were not failing science; they were failing the language of science.

The English learner dimension

Jefferson STEM has a high proportion of English learners (38% of the 8th grade cohort). After reviewing the no-misconception-flag group with the school's EL specialist, the team found that 13 of the 18 vocabulary/notation failures were students with identified EL status. The engine had effectively separated misconception-driven failures from language-driven failures — enabling a more precise support strategy.

The science teachers used the no-misconception findings to advocate for EL-specific science vocabulary instruction resources. The misconception findings shaped their NGSS force unit redesign for the following semester, incorporating the conceptual change strategies recommended by Brainpathio's reteach suggestions for the Aristotelian motion category.

What we learned from this school

Jefferson STEM was our first pilot in a NGSS science context (both Cascade Ridge and Emerald Valley were math-focused). Science misconceptions have a different character than math misconceptions — they're often more deeply held, more tied to everyday sensory experience, and more resistant to single-session correction. The pilot confirmed that our confidence thresholds needed adjustment for science contexts, and led us to develop the "conceptual change" intervention flag, which marks misconceptions that research indicates typically require multiple exposures to contradictory evidence, not just direct instruction.