Eedi + Claude: Diagnostic Questions That Reveal Student Thinking

A multiple-choice item can hide more than it reveals. A student picks C, the key says C is correct, the gradebook moves on — and you still don't know whether they reasoned their way there or guessed between two plausible options. Now flip it. A student picks B, and if B is the specific wrong answer that a specific misconception produces, you have learned something precise: not that they're behind, but where and how. That second kind of item is a diagnostic question, and it is the reason Eedi sits in the Claude for Teachers connector directory.

The point of an assessment item is not the score it produces. It is the thinking it exposes. Eedi's connector does one thing inside Claude for Teachers, and it does it cleanly: it generates diagnostic questions that reveal student thinking, in English and Spanish. On its own that is a useful writing aid. Wired into the rest of the CfT surface — the fifty-state standards graph, the open teaching skills, the class-data workflow — it changes what assessment is for.

Right-or-wrong is a weak signal

Most quick checks give you one bit of information per student per question: correct or not. Aggregate a class of those bits and you get a percentage. A percentage tells you how many students missed something. It does not tell you what they were thinking when they missed it, and that second question is the one instruction actually turns on. Two students can both miss the same fraction-comparison item — one because they inverted the relationship between numerator and denominator, the other because they never converted to a common denominator. Same wrong answer, different repair. If your check-for-understanding can't distinguish those two, tomorrow's re-teach is a guess.

A diagnostic question is designed so that each distractor corresponds to a recognizable line of reasoning. The student who chose the distractor built from "bigger denominator means bigger fraction" has told you exactly which idea to confront. That is the design intent Eedi brings to the table, and it is why "reveal student thinking" is the right verb — the question is instrumented to make invisible reasoning visible.

What the connector actually does

Keep this narrow, because narrow is the honest description. Inside Claude for Teachers, the Eedi connector generates diagnostic questions that reveal student thinking, in English and Spanish. That is the capability. Everything else in this piece is about where that capability fits — not about extending it.

Why does a connector matter here rather than a standalone tool? Because Claude arrives at the request already grounded. Through the Learning Commons Knowledge Graph connector included for every CfT teacher, Claude can see the standard you're assessing, the smaller learning competencies beneath it, and the order students typically learn them. So when you ask for diagnostic items on adding unlike fractions, the request isn't floating free — it's anchored to a specific competency in a specific progression. Eedi supplies the diagnostic-question craft; the standards graph supplies the target. The two together are more than either alone.

Where it lands in the CfT loop

Diagnostic questions are most valuable when they feed something downstream. Claude for Teachers gives them three natural destinations, and this is the one bulleted list this piece will spend:

• Before a unit — placement, not punishment. Use diagnostic items to find where students actually are on a progression before you plan. The results tell you which competencies to assume and which to build, so the k12-lesson-planning skill starts from reality instead of the pacing guide's fiction.
• After a lesson — the exit ticket with teeth. A diagnostic exit ticket doesn't just report a class average; it clusters students by the reasoning behind their answers. That clustering is exactly the input the analyze class data to plan instruction workflow wants when you hand Claude a folder of diagnostics and ask it to build a picture of where each student is.
• Across a class — the map for differentiation. Once you can see which misconception is where, tiered instruction stops being a vibe. The below/at/above split that the differentiation skill produces gets sharper when it's grounded in diagnostic evidence rather than a single raw score.

None of that requires you to leave the conversation. You ask for the questions; you decide what to do with the answers.

Two languages, one instructional signal

The English-and-Spanish detail is easy to skim past and shouldn't be. For a bilingual or dual-language classroom, a diagnostic question only reveals thinking if the student can read it in the language they think in. A misconception surfaced in a student's stronger language is a real signal about mathematics or science reasoning. The same misconception buried under an English-comprehension barrier is noise — you can't tell whether the student misunderstood fractions or misunderstood the sentence. Generating the diagnostic item in both languages keeps the instrument pointed at the concept, not at the reading load. That is a small design decision with an outsized effect on which students you can actually diagnose.

A diagnostic question earns its place not because it grades faster, but because it tells you what to teach next — and it can only do that if every student can read it in the language they reason in.

How to actually use it

Start with a competency, not a topic. "Write me a quiz on photosynthesis" gets you a quiz. "Give me diagnostic questions for the competency where students confuse where a plant's mass comes from" gets you an instrument. Because CfT is grounded in the standards graph, you can name the competency and let Claude align the item to it.

Then close the loop. The workflow that pays off looks less like "make a test" and more like a short cycle you can even hand off on a schedule. Generate diagnostic items for the day's competency with the Eedi connector, in the languages your room needs, and give them as an exit ticket. Feed the responses back to Claude and ask it to group students by the reasoning their answers imply. Then use those groups to adapt tomorrow's plan — the kind of task CfT can run as a scheduled job at 4pm while you drive home. Each pass is short; the compounding effect over a unit is where the value lives.

If you're assembling a broader assessment stack, the marketplace has complementary pieces worth a look: the exam blueprint skill for structuring a full assessment around competencies, and the concept explainer for turning a surfaced misconception into a targeted re-teach. Neither replaces a diagnostic question — they're what you reach for after one has done its job.

The through-line of Claude for Teachers is that good practice — differentiation, mastery-based pacing, small-group instruction — is time-expensive, and the constraint is a teacher's week, not a teacher's will. Diagnostic questions are a perfect small example. Writing genuinely diagnostic items by hand, with distractors that each map to a real misconception, is slow, skilled work most teachers never have the minutes for. A connector that generates them well doesn't lower the bar; it makes the high bar reachable on a Tuesday. That is the whole bet, in miniature.

Want the ecosystem view of how these connectors fit together? Start with the series hub, or browse more builder insights.

---

Part of the Claude for Teachers series. Related: Hand Claude a Folder: Analyzing Class Data to Plan Instruction · Snorkl + Claude: Insights on Classes, Assignments, and Progress. Browse AI tutoring skills or more builder insights.