Grading without burnout: how to mark work so it means something

It's Sunday evening. A stack of thirty essays sits on the desk, three pages each. You read the first one carefully, underlining, writing full-sentence comments in the margin. By the tenth, you're underlining less. By the twentieth, you're writing just "watch your punctuation." By the thirtieth, you catch yourself having read the first paragraph, the last paragraph, and filled in a grade some fast part of your brain estimated for you. And then comes the uncomfortable question: did the last student get the same evaluation they'd have gotten if their paper had been on top of the pile?

Almost certainly not. And that's not a failure of your character. It's the normal functioning of a person who, over three hours, made several hundred small decisions — and every single one required matching a specific text against a fuzzy idea in your head. Grading is exhausting not because there's a lot of it, but because it's decision-making with no support. Every time you have to decide "by eye," you pay for it in attention. And attention runs out.

This chapter isn't about grading faster. It's about turning grading into work that teaches students more and costs you less. Most of the energy grading consumes doesn't go into reading the work — it goes into decisions that could have been made in advance, once, for everyone.

Why grading drains disproportionate energy

When teachers complain about grading, they usually talk about time. But time is only the visible part of the problem. If you spent three hours reading a book, you'd come away from the desk rested. When you spend three hours grading, you come away wrung out. The difference isn't length — it's what's happening in your head during it.

Decision fatigue

Every graded paper isn't one decision, but dozens. Is this sentence still acceptable, or already an error? Is it a typo or a gap in knowledge? Should I factor in that this student has improved? Is a C fair compared to the B I gave someone else for something similar an hour ago? How many points to deduct for a missing conclusion? Every answer is a small choice, and human capacity to make quality choices wears down over time. Decision research repeatedly shows that as the number of consecutive decisions grows, people simplify — reaching for cruder rules, repeating the previous choice, or drifting toward whichever option is more convenient.

In practice that produces a well-known effect: the first and the last papers in the stack get graded differently. At the start you're strict and thorough; in the middle, most balanced; toward the end either lenient (because you don't feel like justifying a lower grade) or, conversely, stricter (because the mistakes are wearing you down by then). Both drifts are unfair, and both can largely be eliminated — not through more willpower, but through better preparation.

The contrast effect and the halo effect

The second source of fatigue is that you're not grading against a criterion — you're grading against each other. A weaker text looks worse right after an excellent one than it actually is. After three disasters, an average paper looks like a revelation. On top of that comes the halo effect: neat handwriting, a familiar name, a memory of how the student behaved in class. No one is exempt from this. The only difference is whether you account for it.

The endlessness of assignments

The third source is the sneakiest: grading has no natural ending point. Every paper could get one more note, one more clarification, one more corrected comma. Work with no defined "done" drags on and triggers guilt even when you've already done more than enough. The chapter on the psychology of productivity describes the same mechanism with open loops — an unfinished thing takes up space in your head even when you're not sitting with it.

The practical consequences are boring but effective. Grade in blocks with breaks — the technique described in the chapter on Pomodoro and time-blocking isn't here for output, it's for the stability of your judgment. Grade anonymously wherever you can. Go through the stack twice: the first pass just sorts papers into three piles by overall level, the second reads within each pile and writes comments. And above all: before you start, have written down exactly what you're actually grading.

Formative and summative assessment: two different jobs

Most of the unnecessary work in grading comes from doing two completely different activities at once, the same way.

Summative assessment closes things off. It states where a student stands against some standard — a grade on a report card, a final test score, points on a graduation project. It has to be fair, comparable, and defensible. It doesn't have to be detailed: "not an A, because the argument is missing in the third section" is complete. Summative assessment is a stamp, not a letter.

Formative assessment moves things forward. It doesn't close anything, it says what to do next — a comment on a draft, a correction meant to be revised, three questions in the margin, a one-minute conversation at the desk. Formative assessment doesn't have to be fair in the sense that everyone gets an equally long comment. On the contrary: it should be exactly as detailed as that particular student needs right now. And above all, it doesn't need a grade attached. The moment you attach a number to a formative comment, most students read the number and skip the comment — attention goes to whatever counts as the "real" result.

The mistake that costs the most energy is writing long formative feedback on a summative piece of work. You spend fifteen minutes justifying a grade at the end of the semester, when the student can't do anything with it anymore. That feedback should have arrived three weeks earlier — and at that point it could have been shorter, because it would have had a recipient who could actually use it.

A practical rule of thumb: the earlier in the learning process an assessment happens, the more detailed it should be and the less it should involve a grade. The later it happens, the shorter it can be — but it has to be defensible. Once you settle this for yourself, half the extra work you used to do simply disappears.

Criteria in advance: a rubric as a tool for speed and fairness alike

A rubric is a table: rows are what you're assessing, columns are levels of execution. For an essay, the rows might be "content and argument," "structure and coherence," "language and style," "spelling," and each row has three or four descriptions of what excellent, good, and inadequate looks like. It sounds like extra bureaucracy. In practice it's the single most effective tool against grading burnout that teaching practice has.

The reason is simple. Without a rubric, you redo the entire decision-making process from scratch on every paper — and that's exactly what wears you down. With a rubric, you decided once, at your desk, with a clear head, before you even assigned the work. During grading itself, you're just matching the text against a description. What was a creative, exhausting judgment call becomes recognition, which is a far cheaper operation.

The second effect is pedagogical and matters even more. A rubric given to students in advance changes the assignment. Instead of "write an essay on topic X," it becomes "write an essay with a clear thesis, three arguments with examples, and a conclusion that doesn't just repeat the intro." Students suddenly know what's expected of them. Work comes in better, which in turn shortens grading. And a large chunk of grade-arguing disappears, because the conversation shifts from "why did I get a C" to "which row was I weak on."

The third effect is legal and human: a rubric is a defense. When a parent asks why their daughter doesn't have an A, you show the table the class received three weeks in advance, and the row where the work fell short. The debate about your supposed arbitrariness ends before it starts.

How to write a rubric without adding to your workload

Stick to three to five rows. A rubric with twelve criteria is a trap — you spend half an hour scoring instead of five minutes, and the student still doesn't know what matters. Describe observable things, not impressions: not "the work is creative," but "contains an original example that wasn't covered in class." Avoid negative phrasing ("has no errors"), because you can't teach toward it. And expect the first version of the rubric to be flawed — you'll know because you'll keep reaching for exceptions while grading. That's information, not a failure: revise the rubric and next time it's better.

Rubrics can also be recycled. Criteria for an essay change only slightly between grade levels and topics — the bar shifts, not the structure. In two years you'll have a ready-made template for every type of assignment, and prep for grading will cost you minutes. Drafting a first version of a rubric, or converting it into a point scale, is also work where AI can serve well as a drafting assistant — with the final version always being yours to approve, and the wording of criteria matching what you actually taught. The tip Tests and quizzes with AI walks through a concrete process.

Feedback that actually gets read

There's a bleak experience almost every teacher knows: you spend an evening writing comments, hand back the papers the next day, and watch students hunt for the number, flip the page, and stuff it in a folder. Your two hours just ended in a binder.

It's not ingratitude. It's that feedback meant to be read has to meet several conditions — and most school feedback doesn't meet them.

It has to be specific. "Weak argument" isn't feedback, it's a diagnosis with no treatment. "In the second paragraph you claim this causes harm, but you don't say to whom or how" is feedback. The difference is that something can be done with the second sentence.

It has to target the work, not the person. "You're careless" is a judgment about the person and triggers defensiveness. "The assignment had three tasks, two are complete" is a description of a state and triggers action. This applies to praise too: "you're talented" is worse than "this transition between paragraphs works great, because it follows on from the previous idea." Praise aimed at performance and process is usable; praise aimed at traits is nice but empty.

It has to be singular. When you write twelve notes on a piece of work, the student fixes none of them, because they don't know where to start, and mostly takes away the feeling that the whole thing is wrong. Pick the one thing that moves the needle most, and write it clearly. You can mark the rest, but don't elaborate on it. The "one thing to improve" rule is also the biggest time-saver you have in grading.

It has to arrive when it can still be acted on. Feedback on work that's definitively closed is academic. Feedback on a draft that will be revised is learning. That's why it's worth shifting the weight of comments from final papers to work in progress — even at the cost of the final paper getting just a grade and one sentence.

And it needs room. If students don't get five minutes in class to read the comment and do something with it, chances are they won't read it. It's worth handing back the work and assigning a concrete task: rewrite one sentence based on the note, write what you'll do differently next time, find the spot in the text the comment refers to.

Form is its own topic. Writing by hand in the margin is slow and often illegible. Spoken feedback is faster, more personal, and students respond to it better — thirty seconds at a desk often lands harder than a paragraph in the margin. For longer pieces, a short voice note or screen recording works well. And for recurring mistakes, it's worth keeping a comment bank: ten to fifteen phrasings you use over and over, stored so they can be inserted in one move. That's not laziness, it's standardization — the energy you save can go into the comments that are genuinely individual.

Self-assessment and peer assessment

There's one piece of work teachers routinely do on behalf of students, even though it's exactly the work the doer learns from: judging quality. When you assess a text, you're the one practicing evaluation. The student practices accepting a verdict.

Self-assessment should therefore not be a closing question like "what would you give yourself," but a structured task before submission. The student gets the same rubric you'll use, and before handing the work in, checks off where they think they stand and justifies it in one sentence. The benefit is threefold. First, they have to read their own work with critical eyes, which alone removes some errors. Second, when grading, you see not just the work but whether the student can recognize quality — often a more interesting signal than the result itself. Third, explaining grades gets a lot shorter: where your assessment and theirs match, there's nothing to discuss, and the conversation focuses on two or three actual differences.

Peer assessment works if it has three safeguards. It needs criteria (without a rubric it turns into a contest of either mercy or cruelty). It needs a format that doesn't allow shallowness — it works well to require a fixed number of specific observations, say two things that work and one question about an unclear spot. And it must never result in a grade: students exchange observations with each other, not points. Grading a classmate is relationally costly and pedagogically pointless.

Peer feedback isn't a time-saver in the sense that it grades work for you. Students do it worse than you and sometimes handle it poorly. The saving is elsewhere: a large share of trivial errors get caught before the work reaches you, so you're not reading the same missing punctuation thirty times over. And the pedagogical payoff outweighs that loss — the ability to judge the quality of your own and others' work is a skill students will need in everything they do afterward. This connects to how students organize their own studying; the topic is developed further in the chapter on productivity while studying.

What not to grade

The fastest way to reduce the load of grading is to stop grading things that don't need to be graded. It sounds like giving up. In reality it's the hardest and most expert decision you make in assessment.

The guiding question is simple: what is this grade actually for? If the answer is "to make students do the work," you have a problem somewhere other than grading — and a grade solves it only short-term, at the cost of students doing the work for the grade, not for learning. If the answer is "to have a data point for the report card," ask how many data points you actually need to justify the final result. Most teachers collect far more grades than they need to justify the outcome, and every extra one gets paid for with an evening at the desk.

More concretely: practice work doesn't need to be graded — that's what practice is for; when it's graded, students stop attempting things they don't know how to do yet and learn less. Ongoing homework just needs checking, not grading; the check can be collective and take three minutes. Practicing a new skill gets graded after the practice, not during it. Behavior, presentation, and on-time submission belong in a different category than subject mastery — when they get mixed in, the grade stops telling you what the student actually knows. And a first attempt at anything demanding is better commented on than graded.

Pay particular attention to grading things where you can't tell who actually did the work. Homework written outside of school is no longer reliable evidence of a student's own skill today, and grading it burns energy that doesn't translate into anything. That's not a reason to give up, just to assign differently — the next chapter, on attention in the classroom, covers how to design assignments in an age when every student has a language model in their pocket, and the tip AI in the classroom: how to teach its use covers it practically.

Written narrative comments and parents

Narrative comments have a reputation for endless work. Done badly, they really are: thirty unique essays, each half a page, each written late at night at the end of the semester. Done well, the workload is comparable to grading — and far more useful.

Structure makes the difference. A narrative comment that works has three or four fixed parts: what the student handles well (concretely, with an example), where there's been progress since last time, what the main thing to work on is, and specifically how to work on it. The parts stay the same for the whole class; the content is individual. That means the text gets written fast, reads well, and — crucially — can be compared to the previous period.

Two rules save a lot of misunderstanding. Describe the work and the skill, not the character. The sentence "he's lazy and unfocused" is a judgment of the child, and a parent will hear it as an attack. The sentence "tasks requiring a longer piece of continuous writing are still coming in unfinished" describes the same thing, but it's something you can work with. And don't skimp on specificity in the positive parts. Parents read narrative comments mainly to find out whether someone actually sees their child. One concrete sentence that makes clear you're talking about their kid, not a template, does more for trust than three paragraphs of generalities.

With parents generally, most conflicts around assessment arise from information lag. When a parent learns about a problem at the end of the semester, they're surprised, and surprise brings emotion. When they got a short note in November, what follows is cooperation. A short, timely, matter-of-fact message is cheaper than an hour-long meeting at the end. The tip Grading and parent communication with AI walks through phrasing harder messages and preparing for a difficult conversation — with one important limit that applies to all AI use in schools: AI proposes, the human approves. Text that goes to a parent must be your decision and your responsibility, and it must never contain a student's sensitive personal data fed into a tool without appropriate data protection. Leave specific names, diagnoses, and family circumstances out — work with a general description of the situation and add the name only to the finished text.

The rest of assessment-related admin can be lightened the same way, from paperwork to recurring messages; the tip Teacher admin with AI offers a practical overview.

Key takeaways

• Grading doesn't drain you through reading, but through deciding. Every decision you make in advance, once, for everyone, is a decision you won't make thirty times at your desk.
• Distinguish formative from summative. Detailed feedback belongs where it can still be used; final assessment should be defensible, not long. A grade attached to a comment erases the comment.
• A rubric is the best investment. Three to five criteria, described observably and published before the assignment. It speeds up grading, improves submitted work, and ends most grade-arguing.
• One thing to improve. Specific, about the work not the person, delivered while something can still be done about it — and with five minutes in class to respond.
• Let students grade too. Self-assessment against the same rubric before submission, and ungraded peer observations, remove some errors before they reach you and teach a skill students will need.
• Don't grade everything. For every grade, ask what it's actually for. Practice work, skill-building, and first attempts get by with a comment or a check.
• Communicate with parents earlier, not more thoroughly. A short, timely message heads off a conflict that would otherwise cost an hour. Describe the work, not the character.