When a whole class reads I denne verden er alt mulig by Klaus Hagerup, the essays that follow tend to arrive all at once. Each one needs a reading that is attentive to interpretation, not just spelling and structure, and that kind of reading is slow when you have thirty or more submissions. Many Norwegian teachers end up spending an entire weekend on a single novel unit. AI grading can shorten that first pass considerably without replacing the teacher's judgment about what a student really understood.

The most useful thing an AI tool can do for a novel essay is apply your rubric the same way to every paper, including the one you read at eleven at night and the one you read on Saturday morning. Human graders drift as they tire, and a strong essay read after several weak ones can look better than it is. A consistent first pass gives you a stable baseline to compare against, which makes your own adjustments more deliberate and easier to explain to students.

Before you grade anything, decide what the essay is actually testing. Hagerup's novel invites responses about imagination, possibility, humor, and the way young characters make sense of the world around them. If your assignment asks students to interpret a theme, the grading criteria should reward interpretation backed by specific moments from the text, not a retelling of events. Writing that distinction into the rubric is the single step that makes AI feedback useful instead of generic.

Set up the rubric around interpretation, not summary

Literary essays often fail because students retell the plot and call it analysis. A rubric for this novel should name the difference explicitly, with separate criteria for claim, evidence, explanation, and language. When the AI scores each criterion on its own, a student who writes beautifully but never interprets anything gets a clear signal about where the gap is. That is far more instructive than a single overall grade with a vague note about depth.

• A clear central claim about the novel rather than a general statement about books
• At least two specific scenes or passages used as evidence
• Explanation that connects each example back to the claim
• Awareness of tone, humor, or narrative voice where relevant
• Organization and language that fit the grade level of the class

Review the AI's first pass like you would a teaching assistant's

Treat the AI output as a draft from a capable assistant, not a final verdict. Skim the scores for outliers, such as an essay rated high on evidence that you remember being thin, and open those papers first. Most of the time you will agree with the tool, and the few disagreements are where your attention adds the most value. Over a unit, you will learn which criteria the tool handles well and which ones need your closer reading.

Keep an eye on essays that take an unusual or creative approach to the novel. A student who reads the book through an unexpected lens may not match the patterns the tool expects, even when the argument is genuinely strong. In those cases, your own reading should override the score, and it is worth adjusting the rubric wording so similar essays are handled better next time. The goal is a process that gets sharper with each unit.

Turn comments into revision, not just justification

Feedback that only explains a grade gets read once and forgotten. Feedback that tells a student which paragraph to rework and what a stronger version might include gets used. For a novel essay, that might mean asking the student to replace a plot summary sentence with one that explains why a scene matters to their claim. Comments written at that level of specificity are exactly what AI can produce at scale for every student.

Once the feedback is out, build in a short revision window before the final mark counts. Students who rewrite one paragraph in response to a comment learn more about literary analysis than students who simply receive a number. This approach also lowers the pressure on the first draft, which often leads to more honest and more ambitious interpretations. Over time, the habit carries into every other text the class reads.