Gains:
- Ability to generate multiple choice, open-ended and higher-level thinking questions in line with achievements with artificial intelligence support
- Ability to control question quality by applying the principles of good distractor (wrong but plausible option), single correct answer and item difficulty.
- Understand the risk of artificial intelligence producing wrong answer keys, non-objective or biased questions, and be able to see that the teacher must verify each item.
Exams and quizzes are the primary tools by which the teacher measures learning; But writing good questions is harder than you think. An ambiguous root, similar options, and clue-giving language detract from the measurement value of a question. AI rapidly generates a large number of outcome-related question outlines; You monitor quality, verify the answer key, and maintain attainment alignment. In this unit, we will cover good article writing and quality quiz production with AI. The limit is clear: AI can produce incorrect answer keys, non-objective or biased questions; The teacher verifies each item.
Matter anatomy and principles of good questions
A multiple choice question has two parts: the stem (question stem) — the main issue being asked; and options — one is the correct answer, the others are distractors. A good distractor, even though it is wrong, seems plausible to the student who has not fully mastered the subject; It is generally based on a common misconception. A common misconception such as "The Moon produces its own light" is a powerful distractor. A nonsense or irrelevant distractor ("The moon is a cheese") makes the question easier and makes it less distinctive.
Principles of good questions: (1) The stem asks something single and clear; (2) there is only one correct answer; (3) the options are of similar length and are grammatically compatible with the root (the longer option is generally correct, do not give a clue); (4) “all/none” options are avoided if possible; (5) if the stem is negative (“is not”), it is emphasized; (6) questions measure the level of attainment—asking recall questions to analysis attainment misses the mark.
Item difficulty is how difficult the question is; Discrimination is how much the question separates those who know from those who do not know. A good exam is a mix of items of varying difficulty.
Tip: When asking questions from the AI, name the target misconception: say "distractors should be based on these 3 common misconceptions." This way you get instructional distractors, not random ones.
Question types
Multiple choice: Quick measure, automatically scored; but measuring higher-order thinking requires careful design. Open-ended/short answer: Shows the student's own statement; It is scored with rubric (6th unit). True-false: Fast but high luck factor. Matching: Good for concept-definition relationship. Higher-order thinking questions: Scenario-based items asking "why/how/evaluate"; It measures analysis-evaluation gains. AI produces blueprints for them all; Your job is to choose which type measures which outcome.
Step by step: Quiz production with AI
- Give the gain and Bloom level. The question should measure that level.
- Specify type and number. "6 multiple choice + 2 open ended".
- Name the distractor strategy. What misconceptions should it be based on?
- Ask for keys and justification. The correct answer to each question and why the others are wrong.
- Verify the key manually. Solve each question yourself; Is the line marked by AI really true?
- Alignment and bias checking. Does the question measure gain? Is there cultural/gender bias?
- Balance of difficulty. Adjust the mix of easy-medium-hard.
three mini cases
Case 1 — Wrong key. A science teacher took a 10-question quiz from AI. The key to the 3rd question was marked "B", but when the teacher solved it, he saw that the correct answer was "C" — the AI thought its own distractor was correct. If the teacher had not corrected him, he would have considered the student who knew it wrong. Solving the key by hand saved the entire quiz.
Case 2 — Weak distractor correction. In a teacher's first quiz, the distractors were nonsensical and students had to eliminate and find the correct answer. He had the AI reproduce "distractors based on these common misconceptions: X, Y, Z." Discrimination increased in the new quiz; class average better reflected actual learning.
Case 3 — Gain alignment. A history teacher's outcome was at the level of analysis ("relates causes") but the AI produced recall questions ("tell the date"). The teacher updated the prompt as "at the acquisition analysis level; ask scenario questions that establish a cause-effect relationship." Questions have become a measure of achievement.
Copiable templates
1) Multiple choice set based on achievement:
Your role: [subject] teacher. Gain: "[gain]" ([Bloom level]),[X. class]. Write 6 multiple choice questions that measure this outcome. Let each question have 4 options; Have the distractors be based on these common misconceptions: [write the misconceptions]. For each question, explain the correct answer and why EACH option is true/false in a separate key section. Mixed easy-medium-difficult.
2) Answer key verification:
Solve the multiple choice questions below one by one. For each question: (a) the correct answer you found, (b) the reason, (c) does it match the marked key. Mark the inconsistencies clearly as "CONTRADICTION". Questions and key: [paste]
3) High level (scenario) question:
"[subject]", [X. class], write 3 scenario-based questions at the analysis-evaluation level. Let each question give a brief situation and ask the student for justification/comparison. Don't let it be solved by heart. Add sample answers and evaluation tips.
4) Bias and compliance audit:
Check the following questions for: gender/culture/region bias, unfair advantage/disadvantage to a particular group, inappropriate context. Flag problematic items and suggest a neutral alternative. Questions: [paste]
Weak prompt / Strong prompt
Weak prompt:
Write 10 questions about fractions.
Level, outcome, type, distractor strategy, no key; Its quality and alignment are left to chance.
Powerful prompt:
Your role: math teacher. Learning Outcome: "The student adds two fractions with different denominators" (application). 5th grade. 6 multiple choice questions, 4 options. Distractors should be based on the following misconceptions: adding both numerators and denominators directly, keeping the numerator constant and adding the denominator. Give key and option description for each question. Mix easy-medium-difficult.
Question type
The level it measures
strength
Attention
multiple choice
Any level (depending on design)
Fast, automatic score
Distractor quality
open ended
understanding-creating
original expression
Need a rubric
true-false
remember
very fast
Luck factor is high
Scenario based
Analysis-evaluation
prevents memorization
Spelling is difficult, AI helps
Common mistakes
- Not verifying the key. The most frequent and most unfair mistake; AI produces the wrong key, the student who knows it will be victimized.
- Nonsense distractor. The implausible option simplifies the question for free.
- Gain-level mismatch. Remembering question for analysis acquisition; misses the target.
- Language that gives clues. The fact that the longest/most detailed option is always correct teaches the student patterns.
- Overlooking bias. The familiar context of a particular culture/group gives some students an unfair advantage.
Caution: AI sometimes produces ambiguous questions where more than one option can be argued as "correct". He answered each item with the question "Is there a single and indisputable truth?" Test it by saying; Otherwise, clarify the root.
In summary
A good question measures the level of attainment with a clear stem and plausible distractors. AI rapidly generates a large number of question outlines, distractors, and keys that depend on acquisition; If you tell the target misconceptions, distractors will be instructive. However, AI may produce wrong keys, weak distractors, out-of-acquisition or biased questions. Solve each item to verify the key, check alignment and bias; The final test is the teacher's.
Application task
For an outcome, generate 6 questions with the "Multiple choice set based on the outcome" template; Specify the distractor strategy yourself. Then solve each question yourself and compare it with the AI's key before using the "Answer key verification" template. Correct any mismatches. Finally, write a version of the scenario that elevates one of the questions to the level of analysis.
checklist
- [ ] Did I ask the questions according to the Bloom level of the learning outcome?
- [ ] Did I base the distractors on actual misconceptions?
- [ ] Have I verified the key to each question by solving it myself?
- [ ] Am I sure there is a single, unquestionably correct answer?
- [ ] Have I checked for bias and age appropriateness?