Iris, the AI Tutor
Iris is a tutor built into Artemis that you can ask questions of while you work. It already has the exercise and your latest submission in front of it, so you can ask about this task without explaining it first.
It is built to help you get there rather than to get there for you. Artemis does not allow it to give complete solutions or full code blocks, whatever your instructor has configured — so asking for the answer will not produce one, and asking why your answer is wrong usually will.
Choosing Whether to Use AI
The first time you open anything AI-powered in Artemis, you are asked to choose. This is your decision, not your instructor's, and it applies to your account.
Cloud uses current cloud models on Azure. Artemis states what that means: processing is GDPR-compliant, your data is stored on university datacenters, and your personal details — your registration number, your email address — are never sent to a model.
Where your Artemis instance offers it, On-premise, marked as experimental, uses open-source models that run only in the university datacenters. Artemis lists its limits: lower response quality and slower responses than Cloud.
No AI leaves the rest of Artemis working exactly as it does now. You lose Iris, the automated AI feedback on modeling and text exercises, and AI-powered guidance; you lose nothing else.
The chat panel puts the same decision in front of you in words: accepting confirms that you are at least 18 and that you agree to your interactions and the context around them — including your code — being processed, with nothing stored in the cloud. Choosing an option settles both; Decline leaves the chat closed and changes nothing else about Artemis.
Asking a Question
On an exercise, the Iris tab sits in the panel beside the problem statement. Opening it gives you an empty chat that already knows where you are.
The chip above the message box is the topic — the exercise Iris will answer about. The + button beside it changes that: point Iris at a different exercise, at a lecture, or at the course as a whole when your question is not about the task in front of you. The × on the chip removes the topic entirely, which is what you want for a general question.
Type the question and send it. Iris answers in the panel, and you can keep going: it remembers the conversation, so a follow-up like "why does that matter here?" works.
Under each answer are three controls: one copies the text, and the thumbs up and down rate it. Those ratings feed an aggregate that administrators can see, so they are how the people running Artemis find out that Iris is unhelpful. They are worth using for that reason, and they do not reach your instructors as individual conversations.
Answers That Cite the Lecture
When Iris answers from lecture material, it places source chips beside the claims they support. Iris gives each chip a short label and summary; its icon identifies a slide deck, a video transcript, or one of the course's FAQs.
Hovering a chip shows its summary. Lecture sources also show their Unit: and Lecture:. Where several sources back the same sentence, their chips sit together as a group, and the arrows beside the summary step through them.
Clicking a lecture citation opens its unit, with the slides at the cited page or the video at the cited moment. That lets you check an answer against the material: the chip shows where Iris found support, not whether its answer is right.
When a Citation Cannot Open the Exact Place
New citations to a lecture page or video moment carry an internal material version. When that version is available, Artemis checks that the slides or video still match what Iris used before opening the exact position. A citation whose version could not be verified opens the unit without jumping to that position. If a checked position cannot be opened, Artemis tells you why:
- The cited material has changed since this answer was written — the slides or the video were replaced, so the page or the moment an older answer names may no longer hold the same content.
- The cited material no longer exists — the file or the video has been removed from the unit.
- The cited material cannot be checked right now — for example, while a video's transcript is being regenerated.
In all three cases the lecture still opens and only the exact position is left out. Once updated material has finished processing, you can ask again for an answer based on it. If the whole unit has been deleted rather than its material replaced, Artemis says that the linked content no longer exists and opens the lecture alone.
Asking About What You Are Viewing
On a lecture page, Iris can use the slide pages visible in your open units as context for your next question. Once you have played a video, it can also use the current video moment. In the combined view / fullscreen, Iris uses the one unit open there. This viewing context is sent with your question, so you can ask "what is this right here?" without naming the page. It reflects what was visible when you sent the message.
When Iris Shows You the Place
Instead of describing where something is, Iris can take you there in the lecture unit's combined view / fullscreen. It can move the slides to a page or the video to a moment, then answer with that material in front of you. With synchronization on, the other viewer may follow. The topic selector is hidden in this view; your next message uses the lecture that contains the open unit.
When Artemis saves an acknowledged move, a chip appears in the conversation, reading Navigated to page 5 and timestamp 0:56 in lecture unit Concurrency. The chip stays, so the place is one click away later on, and clicking it reopens the view if you have closed it. A point-out that was checked against the material opens at one position only, the page or the moment that check covered. Where an older saved point-out still names both a page and a video moment that cannot stay synchronized, Synchronize slides and video switches off so the next movement does not undo it, and Artemis names the page and the moment it is keeping.
Iris can move the material only while that combined view / fullscreen is open. If you click a saved point-out after the material has changed, the unit still opens, but the old position is withheld just as it is for a citation.
The screencast below shows both halves together: Iris answering from the unit open in front of you, and moving the material to the place its answer is about.
Where Else Iris Appears
The exercise chat is offered on programming and text exercises. On other types — modeling, quiz, file upload — there is no Iris tab, and in an exam there is none either, whatever the course settings say.
Beyond an exercise, Iris can answer about a lecture you have open, and about the course as a whole — which is the one to use for "where do I find" and "what is due when" questions. Your instructor switches Iris on or off for the whole course rather than place by place, so the difference between one exercise and another is its type, not a setting.
Iris is also available in Visual Studio Code, where it can read the code you have not submitted yet and offer a hint when you seem stuck; see Iris in VS Code.
Your conversations are yours: they are part of the data export you can request of everything Artemis holds about you.
Getting Better Answers
Iris is as good as the question. Three things help.
Say what you tried. "I expected the list to be sorted after the first pass but it is unchanged" gets you further than "my code does not work", because the first tells Iris where to look.
Set the right topic. A question about last week's lecture asked with this week's exercise as the topic gets an answer about the wrong thing. Change the chip first.
Ask about the reasoning, not the answer. Iris cannot give you the solution, so a request for one wastes a turn. "Why is a list of observers better than calling each display directly?" is a question it can answer completely.
Related
- Iris for instructors — the settings behind the tutor you are talking to, including the support level that decides how much it gives away.
- Artemis Intelligence — where Iris sits among the other AI services and what each is responsible for.