Blog
A rubric AI can follow is a marking sheet a second reader could apply to a photo of a student page and land on the score you would give. Rubric AI grading depends on that sheet. Each criterion names something you can see. Each row has a point value. Partial credit has a boundary. A short example shows a full mark and where a partial mark stops.
On a Sunday pile, a missing unit, a slipped sign, and work on the back get decided again on every paper. Write them down. A method mark needs the work in the frame. A unit mark needs the unit on the page.
Graider is the AI red pen for paper grading, a product of DM Liquidity LLC. You upload the test and the answer key, photograph the student papers, and get marks against the rubric and key you provided. You review, toggle feedback, and hand the work back as a PDF or an email. The public site is graider.org. For why your standards stay in the lead, and why you keep the pen, read grade against your own rubric.
Your rows. Their handwriting. Points a photograph can support.
Write a rubric for AI grading the way you would brief a colleague marking the next section. One row for each thing you score. A point value on that row. Words that point at the student page.
Correct final value, 2 points, is a row a second reader can apply. Reasoning, 3 points, becomes usable once you name it: the equation from the prompt, the substitution, and the arithmetic. Give each feature you look for its own points.
If question 4 is worth 6, the rows under it should add to 6. A diagram worth 1 of those points gets its own line, so a missing label leaves the calculation points in place. Write every accepted form of the answer. Order the rows as you scan the page: setup, substitution, result, unit, then the sentence of meaning.
Write each partial-credit border once, before the photos are on the desk. A usable rule has three parts: the condition, the points, and a short picture of a paper that earns them. Arithmetic error after a correct setup, 2 of 3, can be reapplied on every photo.
Put the borders you already use into lines:
Name what earns zero: a blank, a bare number when work is required, or an answer to a different question. Say whether readable crossed-out work still earns method points, or whether you score only the response left standing. When a missing unit and an arithmetic slip arrive together, name the point each one costs.
One strong response and one partial response are enough for most items. For a 4-point explain item:
Write only the levels you actually award. On a multi-step item, show the setup, the final value with units, and one path you still credit, such as a sign error with the method intact. On multiple choice, record the letter and what a second mark does. A comment you repeat for the same miss can sit beside the rule, ready to toggle before release.
Name the marks that sit in the margin, on the graph, and on the back of the sheet. Meter per second on the final value can be 1 point, with the number still earning its points when the unit is wrong. If a missing unit makes the final answer wrong in your course, write that and use it on every item where you mean it.
Method marks name the required lines: equation, substitution, and result. Say whether a bare correct result earns a different score. Write rounding as a range, such as 2.4 to 2.5 inclusive, or as the figure count from class. A graph can split the same way: labeled axes and a line, 2 points; axes only, 1 point.
Photograph the full sheet, including the back when work continues there. How to shoot the stack is on the AI paper grader page. The rubric assumes the writing is in the frame. A faint line or a cropped corner stays with you in review. When the prompt says to show work, name which work and how many points it carries.
Write the calls you would hate to reopen: points, accepted forms, partial-credit borders, units, required work, and any comment you may toggle on. Save a teacher look for a new method that still reaches the required relationship, a diagram your example missed, a prompt that was easy to misread, and a page too light to trust.
One line can keep that review available. An equivalent method that reaches the required relationship earns the method points. Unusual layouts get a teacher look before release. Rubric AI grading holds up when the ordinary path is specific and the leftovers are named as your review.
Upload the test and the answer key you already use, with this rubric on the key or beside it. Photograph the student papers. Graider marks those photos against the rubric and key you uploaded. You review the marks, toggle feedback on or off, and hand the work back as a PDF or an email. You keep the pen. The score you release is the one you will stand behind when a student asks.
Graider is built for clearing the Sunday grading pile. A class quiz follows the same habit on the AI test grading page. Score a strong sample and a thin one out loud before you photograph the stack. Add any sentence you needed in order to decide, then copy that border onto the key. That revision is part of how to write a rubric for AI grading.
Specific criteria, point values that add up to the question total, a partial-credit rule with an example, and a note on units or method marks when those carry points. A second reader with the photo and the rubric should be able to point at the page for every point awarded.
Give each scored feature its own row: the points, and something visible on the paper, such as the final value, the unit, the setup, or the label. Put that rubric with the answer key you upload. Photograph the full sheet, including the back when a student continued there.
Yes. Write the condition, the points, and a short example of a paper that earns them. Graider marks the photographed papers against the rubric and answer key you upload. You review the marks before a score goes home.
Unusual methods, a diagram in a layout your example did not cover, a prompt that proved easy to misread, and a page too faint to trust. Write the ordinary borders into the rubric. You choose the unusual marks, toggle feedback, and release a PDF or an email.
The test, answer key, and rubric you upload. You review the marks, toggle feedback, and hand work back as a PDF or an email. Graider is the AI red pen for paper grading, a product of DM Liquidity LLC. The public site is graider.org.
A rubric with point rows, partial-credit rules, and a worked example is ready to sit with the answer key. Photograph the papers, review the marks, and hand the stack back when you can stand behind the scores. You keep the pen.