LO-1
Evaluate risk-adjusted return
In scope: Week 3 · Returns.pdf
QUIZ STUDIO
How does Lectora draft exam questions from my own course material?
Quiz Studio drafts questions against what your course actually teaches. Before it writes one, it searches your course knowledge base under an instruction that every question must be answerable from that material, and it records the files it read. Each question is tagged to a learning goal your course declared, using only the codes the course itself supplies. Nothing enters an assessment until a person signs it out.
Quiz Studio
LEC207 · ILLUSTRATIVE
LEC207 / Studio / Canvas
0 questionsNo learning goals selected3 questionsSelected goals with drafts: 1/1
MULTIPLE CHOICE
Draft candidate
A portfolio returns 11% with a standard deviation of 14%. The risk-free rate is 3%. Which statement about its Sharpe ratio is correct?
Easy
SHORT ANSWER
Draft candidate
Why must the Sharpe ratio be computed on the same return frequency as the risk-free rate before you compare two funds?
Medium
EXTENDED RESPONSE
Draft candidate
A colleague reports a Sharpe ratio of 0.57 using monthly returns and a 3% annual risk-free rate. Explain what is misspecified, and how you would restate the ratio.
Hard
Choose a learning goal.
Select a learning goal to open a lane. A category sets scope. Course material focuses the draft.
A question enters an assessment only when a person signs it out, one at a time. Writing a question and signing one out are two different permissions, and an assessment is pinned to the exact question versions it was built from.
What Quiz Studio drafts
Three question types: multiple choice, short answer, and extended response. Multiple choice is drafted as single-select or multi-select, and you decide which. One to thirty questions in a run, at one of five difficulty levels, in Bokmål, Nynorsk or English.
That list is complete rather than illustrative. True/false is not among them, and neither is any interactive or drag-and-drop format. Where a question type is not there, it is not there.
How a question is grounded
Before the model writes anything, it searches your course knowledge base: the material you uploaded, split into passages and indexed. It works under an explicit instruction that every question must be answerable from what the course actually teaches, and every run records which files it consulted. Those sources are stored alongside the questions the run produced.
If a course has nothing indexed yet, Lectora does not pretend otherwise. The prompt switches to an honest branch: draft from the instructions given, and do not claim sources you cannot see.
Two limits worth stating plainly, because the interface can suggest more than the retrieval does. Naming specific files influences the model as a hint, but the search still runs over the course knowledge base as a whole rather than filtering to your selection. And source attribution is recorded at the level of the run, not the individual question, so a question carries the run's source set rather than its own citation.
Tagged to the goals your course declared
Each drafted question is tagged to a learning goal, using only the goal codes the course itself supplies. The model may not invent a code, and a code it cannot resolve is dropped rather than guessed. That is what lets an approved question count toward coverage the moment it enters an assessment, and it is why the coverage view on the Curriculum page can say which goals an exam does not test.
The marking scheme comes with it, split by audience
A drafted question arrives with its guidance already separated by who is meant to read it: what every student sees, what a student who answered correctly sees, and what a student who answered incorrectly sees. Each wrong option can carry a separate examiner-only note naming the misconception that option is built to catch.
The answer key is examiner-only by construction. A student-facing projection of a question cannot carry the key, the examiner guidance, or the internal notes.
Nothing counts until a person signs it out
A drafted question is a candidate. It enters an assessment only when someone approves it, one question at a time, and writing a question and approving one are two different permissions; a department can separate the two roles.
Every question keeps its history. Submit, approve, reject, archive and restore are each recorded as an immutable event with a content hash of the question at that moment. An assessment is pinned to the exact question revisions it was built from, so editing a question later cannot quietly change an exam that already used it.
Approving a question is an editorial act. It carries no statutory force, unlike the examiner approval that a grade requires, and we keep the two apart deliberately.
Three: multiple choice, short answer, and extended response. Multiple choice can be single-select or multi-select, and you choose which. That is the complete list of what the generator produces today, not a subset of a longer menu.
Between one and thirty in a run, at one of five difficulty levels, in Norwegian Bokmål, Norwegian Nynorsk or English.
Yes. The drafting model searches the course knowledge base and works under an instruction that every question must be answerable from what the course teaches. Every run records which files it read, and those sources are stored with the questions it produced. When a course has nothing indexed, Lectora says so and drafts from your instructions alone, under an explicit instruction not to claim sources it cannot see.
Not reliably today, and we would rather say so than imply otherwise. The interface lets you name files, but the retrieval step still searches the course knowledge base as a whole; the filenames act as a hint to the model, not as a filter. Source attribution is also recorded per run rather than per question, so a question carries the run's source set, not its own citation.
We do not claim they are. There is no study of question quality, and the score-agreement result from the grading validation does not transfer to it. What we can describe is the process: grounded in your material, tagged to your declared goals, and reviewed question by question before anything counts.
Not today. An approved question lives in Lectora's question bank and can be pinned into an assessment, and that is where it stops. Writing Lectora-authored quizzes back into Canvas is not an operation Lectora performs at present.