Common questions

Rubrica is a Mac app that helps instructors grade scanned, handwritten exams against their own rubric. It proposes a score for each question, writes feedback on each question for each student, and flags answers for a person to check. The instructor reviews the results and decides what goes to students.

Instructors and teaching assistants who grade large stacks of handwritten exams. It has been used in university economics, finance, and data courses.

Yes. Rubrica uses Anthropic's Claude to read handwriting and score each answer against the rubric. Grading requests carry a random code, the answer pages, and the rubric, and do not include stored names or IDs.

No. Rubrica proposes scores and flags answers for a person to check. The instructor can change any score or comment, and Rubrica sends nothing to students on its own.

For grading: a random code, answer pages without the cover page, and the rubric go to Anthropic. Stored names, student IDs, and cover pages do not. A name written on an answer page goes with the page, so instructors should check first. Other features send other data when the instructor uses them: rubric tools send sample answer pages, the class diagnostic sends samples of feedback, and reports go to students by Gmail with their name. See every path out.

Rubrica is designed so that stored names, student IDs, and cover pages are left out of every grading request. Compliance depends on how a school uses any tool and on its own policies, so using Rubrica does not by itself make a course or institution FERPA compliant. We are glad to answer questions from IT and compliance teams.

Early studies are small. On one midterm's structured section, Rubrica differed from human scores by 1.26 points per exam on average. On a 25-point essay it scored lower than the human, by 3.64 points on average. See the studies and their limits.

Our conservative estimate is about 5 minutes per exam, or about 17 hours on a 200-student exam. That assumes about 13 minutes to grade an exam by hand and about 8 minutes to review and correct it in Rubrica. It is an estimate, not a measurement, and it varies with the exam and with how closely each one is reviewed.

We have not measured whether Rubrica's feedback improves learning. The app records whether a report was sent, not whether a student read it. Research on feedback in general is summarized on the Evidence page.

Exams with clear questions and a rubric that spells out partial credit. In the one comparison with human scores, Rubrica differed more on an essay question than on structured ones, and scored the essay lower on average, so instructors should review essay scores closely.

Each student receives a PDF report with their score, a card for each question, and written feedback on each question, based on the rubric. Reports are sent from the instructor's own Gmail account, only when the instructor sends them.

An Apple Silicon Mac on macOS 12 or later. macOS 14 or later includes the name-reading model; older versions need Ollama installed separately. Instructors also need an approved Google account and a Gmail account to send reports.

Yes. Each version has its own rubric in the same workspace.

Not directly. Grades export as spreadsheets for you to upload.

Rubrica has been used on economics, finance, and data exams. Results on other subjects are unknown, so try a small batch first.

Yes. Support is available during the pilot. Write to hello@tryrubrica.com.

Rubrica is in a pilot. If you would like to use it in your course or department, contact us.