Quizzes and grading¶
Quiz lessons are central learning tools — they check whether the content was really understood, give learners direct feedback, and contribute to course completion. This page explains how to configure quizzes, what to watch out for with AI-assisted free-text grading, and how to set the minimum score sensibly.
Quiz types at a glance¶
| Type | What for | Grading |
|---|---|---|
| Multiple choice | One correct answer out of 3-5 options | Automatic — points for the correct answer |
| Multiple select | Several correct answers out of several options | Automatic — points for a fully correct selection |
| Cloze text | Complete words in the text | Automatic — points per correctly filled gap |
| Matching | Pair terms | Automatic — points per correct pairing |
| Drag and drop | Place elements | Automatic — points per correct position |
| Free text | Open answer, graded by AI | AI-assisted with your criteria (see below) |
| H5P quiz | Various H5P quiz types | Automatic — see Creating H5P content |
Creating a multiple-choice question¶
- Create a new lesson of type Multiple choice in your course.
- Formulate the question — a clear, unambiguous question.
- Create answer options (3-5 is sensible).
- Mark the correct answer(s).
- Optionally add an explanation per answer — shown to learners after submitting, perfect for a learning effect from wrong answers.
- Set the points (default: 1 point correct, 0 points wrong).
- Save.
Screenshot to follow
Multiple-choice editor with question, answer options and explanations
Tips for good multiple-choice questions¶
- Plausibly wrong options (distractors) — learners should really think, not obviously guess right
- Each distractor a typical mistake — if the wrong option also exposes false assumptions, learners learn even when picking wrong
- No "all answers are correct" trick questions — they confuse more than they test
- One question = one learning objective — no compound constructs
Free-text grading with AI¶
AI grading is useful for tasks that don't fit "A/B/C/D" schemes — e.g. reflections, professional justifications, one's own application examples.
Creating a free-text lesson¶
- Choose lesson type Free text with AI grading.
- Write the task — what should be assessed.
- Provide a model answer — the AI compares against it.
- Define grading criteria:
- e.g. "Professional correctness (40%)"
- "Completeness of the argumentation (30%)"
- "Own examples or reflection (30%)"
- Set the maximum points (e.g. 100).
- Save.
Screenshot to follow
Free-text editor with task, model answer and grading criteria
What the AI grading delivers¶
Per submission, you — and the learner — get:
- Points relative to the maximum points
- Strengths of the answer (3-5 bullet points)
- Suggestions for improvement (3-5 bullet points)
- Overall feedback as a coherent text
The grading is repeatable — learners may attempt a second try.
When AI grading makes sense and when it doesn't¶
| Makes sense | With caution |
|---|---|
| Reflection, application questions, "explain in your own words" | Highly critical assessments where the result has consequences (certificate, certification) |
| Fast feedback in large classes | Specialist knowledge the AI doesn't reliably know |
| Exercises without grade pressure | Answers that should be creative — the AI tends to prefer "normal" answers |
For critical assessments, plan a manual review — the AI gives a tendency; the final word should be yours as a trainer.
Reviewing / overriding the AI grading yourself¶
- In Analytics → Lesson breakdown → Free-text tasks, look at the individual submissions.
- The AI grading is visible per submission.
- If you find a grade unfair: Override — your points + your feedback replace the AI grading.
Learners see that a trainer assessment was made.
Setting the minimum score sensibly¶
For each quiz lesson, and also for the whole course, there's a minimum score at which the lesson or course counts as passed.
| Score | What it means |
|---|---|
| 50% | Very lenient bar — only the essentials need to stick |
| 70% (default) | A solid middle ground, tolerates gaps |
| 80-90% | Demanding, typical for mandatory training with a compliance angle |
| 100% | Perfect — only sensible for very short quiz lessons, otherwise frustrating |
Learners may repeat lessons; the best score counts.
Configuring repeatability¶
Per lesson you can set:
- Unlimited repetitions (default) — learners can try as often as they like
- Maximum number of attempts — e.g. 3 attempts for an exam-like character
- Time window between attempts — e.g. a 24 h wait so there's no spontaneous re-attempt
For mandatory training with a certificate, a limit is often worthwhile; for voluntary exercises, more leniency can be better.
Frequently asked questions¶
How do I stop learners from simply guessing answers in multiple choice? - More answer options (5 instead of 3) reduce guessing success to 20% instead of 33% - Explanations per answer: at least you learn something even from wrong guesses - Set the minimum score higher - Instead of multiple choice, use a free text or a drag-and-drop, where guessing works less well
Learners complain about the AI grading — that it's too strict/too lenient. First step: review your model answer and the criteria. Are they really clearly formulated? Are synonyms or acceptable variations missing? For systematic skew: mark the AI grading as a "preliminary assessment" and additionally review it manually.
How do I see whether a quiz question is too hard/too easy? Analytics → your course → Lesson breakdown: per quiz question, the dashboard shows the distribution of answers. A question that no one answers correctly is too hard (or unclearly phrased). A question everyone gets right is too easy (or obvious).
Can I import quiz questions from an Excel/CSV? Currently not directly in the UI. Pragmatically: for many similar questions (e.g. vocabulary tests) an H5P quiz is worthwhile — it has bulk-entry modes.
A lesson is marked "completed" even though learners didn't reach a full score. "Completed" means: the minimum score was reached, i.e. passed. 80% with a 100% minimum score = passed. If you want to be stricter, set the minimum score higher.