Memuat
Memuat
Want to go deeper?
Join one of our courses or ask us directly on WhatsApp.
Picked from the software, topic, and level of “Claude from Zero to Expert Class #67: Creating Questions, Answer Keys, and Grading Rubrics”.
Other ways we can help with Claude.
Learn with a mentor
Scheduled AI & Machine Learning courses with live guidance and a certificate.

Midjourney Course: Create Stunning AI Images from Text
★ 4.9

DALL-E Course: Creating Professional AI Images from Text
★ 4.9

Stable Diffusion Course: AI Art from Beginner to Professional
★ 4.9
Have it built for you
No time to learn Claude? Our team can do the project for you.

AI Image Creation with Stable Diffusion — Unlimited Custom Visual Production
★ 4.9

AI Image & Video Creation with Leonardo AI — All-in-One Visuals for Business
★ 4.9

AI Image Creation with Midjourney — Cinematic AI Art for Business
★ 4.9
Get notified about new Claude classes
At most one email a week. Unfollow anytime with one click.
The full narration of this video. The current sentence is highlighted while the video plays; click any sentence to jump there.
Hi, and welcome back to Claude from Zero to Expert. Mr. Hendra's fractions module is finished, and now it's time to measure whether students really understand. In this episode we build a test blueprint, multiple-choice questions, a cross-checked answer key, a rubric for essay questions, and we read the quiz results question by question.
Many teachers jump straight to asking AI, make me ten fractions questions. The result often piles up on one type of question, and some learning objectives aren't tested at all. A good question isn't just hard; it's on target, measuring what was actually taught. That's why Mr. Hendra starts with a blueprint. A blueprint also makes it easy to create a makeup quiz or a second version. Just ask Claude for new questions from the same blueprint row, at the same difficulty level. A blueprint is a map of the questions. Mr. Hendra builds it with Claude from the module's three learning objectives. Question one measures finding a common denominator using the LCM, at an easy level. Question two is adding fractions whose denominators are multiples, such as one half and one quarter. Still easy. Question three moves up to medium: the denominators are truly different. Question four is a one-step word problem. And question five is the hardest: an essay word problem where students must write out the steps. Five questions in twenty minutes, and every learning objective gets tested.
Mr. Hendra asks for questions one to four as multiple choice, with one important condition: the distractors must come from common student mistakes. Distractors are the wrong answer options. The questions appear in the right panel. Look at question two: one half plus one quarter. The option two sixths is there on purpose, because that's the answer of a student who adds the numerators and denominators directly. Below is a table just for the teacher: each distractor and the mistake it represents. For example, three twelfths in question three means the student found the common denominator but forgot to change the numerator. With distractors like these, even wrong answers are useful. Mr. Hendra immediately knows which step a student got wrong, not just that they were wrong. Use this list to check the questions Claude writes. First, one question measures only one objective, so the results are easy to read. Second, there's only one correct answer. It sounds obvious, but AI-written questions sometimes have two answers that could both be defended. Third, the answer options are about the same length. Clever students often guess the longest option. Fourth, the distractors are plausible and come from real mistakes, like the ones we just saw. Avoid all of the above options or deliberately tricky wording. A quiz should measure understanding, not test how well students spot traps. Besides multiple choice, Claude can create other question types from the same blueprint. First, true or false with a reason. Students can't just guess; they have to explain why. Second, matching. For example, matching fractions with the right circle pictures. It suits students who learn more easily through pictures. Third, spot the mistake. Students get a classmate's wrong answer and must point out the wrong step. This builds deeper understanding than just calculating. Fourth, everyday context questions: a cake recipe, shopping at the market, or travel distance. Varied question types keep a quiz from being boring. Just make sure every question connects to one blueprint row, so no question is there only to fill space.
Here's the most important habit in this episode: cross-checking. Mr. Hendra opens a new conversation and asks Claude to solve all four questions from scratch, without seeing the key. Claude works step by step and writes the answers in a table. Questions one to three match the original key. But question four is different. The original key says one and a half kilograms, while the recheck gives one and a quarter kilograms. Mr. Hendra asks about it. Claude explains that the original key was wrong. Mr. Hendra still works it out himself: three quarters plus two quarters equals five quarters, which is one and a quarter. Without cross-checking, students who answered correctly would have been marked wrong. Make this a three-layer check. Layer one: Claude writes the key along with the questions. Layer two: a cross-check in a new conversation, without the key. A new conversation matters, so Claude doesn't simply repeat its earlier answers. Layer three: the teacher calculates personally, especially for questions with different results. The teacher is the final checker, not the AI.
Question five is an essay question: Dina pours two fifths of a liter of milk, then one third of a liter. What's the total? For a question like this, we need a rubric. A score of four is for complete, correct steps, with the result eleven fifteenths of a liter. A score of three if the steps are right but there's a small calculation slip at the end. The student understands the concept but wasn't careful enough. A score of two if the denominator was matched but the numerator wasn't changed. A score of one if the numerators and denominators were added directly. Notice that these rubric levels line up with the distractors in the multiple-choice questions. With a rubric, two teachers grading the same answer will give similar scores. A rubric is also easy to explain to students and parents. Before using it, the rubric gets tested. Mr. Hendra writes two sample answers himself. He deliberately doesn't use real student answers, let alone ones with names attached. He asks Claude to score both answers with the rubric, with a one-sentence reason. The most useful part is the constraint: if an answer is hard to score, point out which part of the rubric needs clarifying. From this test, Mr. Hendra adds a row for a correct answer with no steps shown. This pattern is useful for your business too. Dewi could use a rubric to assess intern reports, and Laras to assess a new barista's practice. Clear criteria, fair assessment.
Twelve students have taken the quiz. Mr. Hendra recaps the percentage of correct answers per question, without student names, and notes that the distractor three twelfths was chosen by five students. Claude shows a chart per question. Questions three and five are lowest. Since the distractor three twelfths was chosen so often, the problem is clear: students forget to change the numerator after matching denominators. The suggestion is concrete: review that step for ten minutes in the next lesson. This is why meaningful distractors matter. A quiz doesn't just produce grades; it shows what needs to be taught again. To wrap up, four rules for fair grading. First, keys and rubrics are always cross-checked before they're used for grading. Second, quiz results are sent without student names. Names aren't needed to find patterns. Third, final grades are decided by the teacher, not the AI. Claude helps, but responsibility for grading lies with the teacher. Fourth, quiz results are used to improve teaching, not just to fill in report cards. The app screens in this video are simulations, and menu names may differ slightly because Claude is updated often.
We've built a test blueprint, written questions with distractors from real mistakes, cross-checked the key, created and tested a rubric, and read the quiz results question by question. In the next episode we turn to content: writing short video scripts for Reels and long video scripts for YouTube. See you there!
Try these on your own after watching to lock in what you learned.
Type a question or keyword; we find the part of this episode that explains it and jump the video there.
Notes are saved with the current video time and are only visible to you.
Stuck on a step? Ask here; the instructor and other learners can answer.
No questions yet.

Cerdas Inovasi Digital Team
Software developers, BIM & 3D engineers, and AI practitioners
We build software, BIM models, games and AI automations for clients every day, and turn that hands-on work into step-by-step classes: every step is shown on screen and comes with a PDF guide you can keep.
See our clients →