What "flagged for review" means, and why we never guess

Every automated grader eventually meets a paper it cannot read with confidence: a digit that could be a 1 or a 7, two answers written and neither crossed out, a blank where a bubble should be. The question is what the tool does next. Most guess. SuperGrader flags.

What a flag is

When the system is not confident enough to score a question on its own, it leaves the decision to you. The question shows up marked in the results, with the reason it was flagged — "two answers written, neither crossed out" — next to the photo of the student's actual work. You look, you decide, you set the score. Nothing is scored silently and nothing is invented.

Why this is the right trade

A guessed score looks finished. That is exactly the problem: a wrong guess buried in a class set of thirty is the kind of error nobody catches until a parent does. A flag costs you ten seconds per uncertain question and costs nobody a grade they did not earn. Across a whole class set the flags are rare — a handful of questions — and they are the questions you would have wanted to see anyway.

Your correction sticks

When you change a score, that change is yours. Later recalculations leave a mark you have changed alone, and your corrections are never used to train any model. The review screen is not a suggestion box; it is where the grade is decided.

It is the same rule everywhere

The flag is one expression of a principle that runs through the whole product: the AI reads, your answer key decides, and when the reading is uncertain a person is asked. That is also why the same paper gets the same score every time, and why a score can always be traced back to the question it came from.