Your marking scheme. Your marking. Your final word.
CORRAICT handles the repetitive work that surrounds an assessment — reading eighty submissions against the same marking scheme, writing eighty feedback comments — and gives you back the part where your judgement matters. Nothing is sent to the student without your approval.
“Marking consistently and giving students genuinely constructive feedback is a permanent challenge. Even with the best intentions, it is not always easy — and so it does not always get done…”
What you ask yourself before trying.
Arguments about time savings resonate with management. For a teacher, the real question lies elsewhere.
"An AI doesn't mark the way I do."
A generic marker applies an average marking scheme. You credit a sound approach even when the answer is wrong, and penalise a correct answer that is poorly justified.
It is your marking scheme that marks, criterion by criterion.
You upload your marking scheme with its weightings and your expectations. Every proposed mark is linked to a criterion and accompanied by the sentence from the submission that justifies it. You can immediately see exactly where the disagreement lies.
"Reviewing automatic marking takes just as long."
Checking line by line work you did not do yourself can sometimes take longer than doing it yourself.
Part of the submission is not assessed by AI at all.
For an exam sat online, numerical answers are compared field by field against your solution, with no model involved: the answer is either correct or it is not. Everything else is adjudicated with the justification shown alongside each proposed mark — you are not redoing the work, you are settling a disagreement.
"My institution will use it to monitor me."
A tool that records marking can be used to measure teachers. The concern is legitimate.
The log traces assessments, not teachers.
The history exists to provide traceability of a grade for the student and for audits. No productivity metrics, no teacher rankings, and no comparison of marking speed are produced or exposed to management.
From the exam to the grade returned.
The same workflow, whether you are marking a scanned paper exam or an online submission.
You set the framework
The exam, your marking scheme, your expectations. If the exercise was created in CORRAICT, the marking scheme is already there.
Once per assessment
The AI prepares
Submissions are read and checked against your marking scheme. Every proposed mark is justified by a supporting extract.
While you do something else
You adjudicate
You browse, amend, override. Flagged submissions come first. A grade only becomes final upon your approval.
The only time that cannot truly be reduced
You return
The grade and teaching notes are sent to the student in your wording. You can reopen a disputed submission.
In one send to the class
What CORRAICT will not change.
Three things that belong to you and will stay that way.
Your way of annotating
The suggested text is a draft. You rewrite it, shorten it, or replace it entirely. What the student reads is your wording.
Responsibility for the grade
It remains yours, and the product is built with that in mind: there is no way to publish an assessment without a teacher having approved it.
Ownership of your exercises
Your exams, your marking schemes, and your students' submissions do not train any model and never leave the boundaries of your institution.
Create an account
Under a minute, no credit card required, even if your institution is not yet a client.
Submit a real assessment
An assessment you have already marked: that is the best way to compare.
Compare with your own marking
Submission by submission, criterion by criterion. One assessment will tell you whether the tool works for you.