AI-Assisted Feedback

Artemis can ask Athena to write feedback on a submission. It does this in two quite different situations, and they are worth keeping apart because they serve different people.

	Grading feedback suggestions	Formative feedback requests
Who asks	Nobody — they are waiting when a reviewer opens a submission	The student, on a draft
Who reads	The reviewer, before they mark	The student
When	After the due date, while marking	Before the due date
Effect on the mark	None until a reviewer submits the assessment	None at all
Exercise types	Text, programming, modeling	Text, modeling, and programming where the exercise is assessed semi-automatically

Neither ever marks anything. A suggestion is a draft a person accepts, edits or discards, and a formative response is guidance the student reads. The mark is always written by a reviewer.

Switching It On

It is configured once, for the course, in the Athena AI Feedback panel on the course overview: Grading feedback suggestions and Formative feedback requests turn on the two kinds independently, and each takes effect as soon as you set it. See Course Configuration.

There is no per-exercise switch: an exercise gets AI feedback because its course has it on and because its type supports it. Two restrictions follow from the type rather than from a setting.

• A programming exercise takes formative feedback requests only where it is assessed semi-automatically — that is, where manual assessment is enabled alongside the tests. With purely automatic assessment there is nothing for Athena to be formative about.
• An exam exercise is outside this entirely. Exam feedback follows the exam's own review and publication dates.

ℹ️

Information:

Because it is a course-level setting, turning it on reaches every eligible exercise at once, including exercises that already exist. Turning it off does the same.

What a Reviewer Sees

Where grading suggestions are on, opening a submission loads Athena's proposals into the assessment editor, which takes a moment. Each is a feedback block with points attached, sitting where a reviewer would have written one.

The exercise pages show this per type, because the editors differ: see text exercises and modeling exercises.

⚠️

Warning:

Suggestions are drawn from assessments already given in the same exercise, so they inherit whatever the first reviewers did — including their mistakes. Mark a first batch deliberately before leaning on them.

What a Student Sees

Where formative requests are on, a student can ask for preliminary feedback on a draft before the due date and get a response with a provisional score. It is explicitly not their mark, and Artemis says so where it appears.

Students can also shape how that feedback reads — how detailed and how formal — through their feedback preferences. See Adaptive Learning for those, and AI Feedback on a Draft for the student's side of the whole thing — how they ask, what comes back, and the limits on how often.

ℹ️

Information:

A student cannot complain about preliminary AI feedback: there is no assessment to dispute yet. See Assessment and Grading.

Related

• Iris — the other AI a student meets, which tutors rather than assesses.
• Artemis Intelligence — every AI service Artemis integrates with, and who owns what.