Setting and marking an assessment

An assessment holds questions, and the question type decides what can be marked automatically later. Choosing it thoughtfully now is what saves the marking time.

Objective types — multiple choice, numeric, checklist — carry a correct answer and can be scored without you. Written types cannot, and the software does not pretend otherwise.

• Marked by the software: multiple choice (single and multiple), numeric, checklist.
• Marked by a person: short answer, long answer, rubric-assessed, file upload.
• A rubric question is scored against criteria you already defined, not against a number you invent here.

Until a student hands in, their work sits as not submitted and nothing is marked. On submission the objective questions are scored immediately, and every written answer is left alone with no score at all.

A submission after the deadline is recorded as late rather than blocked. Whether lateness costs marks is a professional decision, so the software records the fact and leaves the judgement to you.

The objective questions now carry a suggested score. It is not the mark. It sits in its own column beside an empty awarded column, and the gradebook reads the awarded one.

This separation exists so an ambiguous multiple-choice question, or an answer the key got wrong, does not silently become a grade.

Written answers are yours to mark. If AI grading is switched on, it drafts a suggested score with a rationale and the quote it relied on, so you can see what it read before deciding.

A suggestion is accepted, edited or rejected, and which of those you chose is recorded. Over time that gives an agreement rate — the share of suggestions kept unchanged. A rate near one hundred per cent is usually a sign nobody is really checking rather than a sign the AI is excellent.

Marked work stays with you until you return it. That is what lets you mark a whole class over several evenings and release it as one set, rather than students receiving marks in the order you happened to work.