What sets IB marking apart: criterion referencing
IB assessment is criterion referenced: the internal assessment and the Extended Essay in each subject have a set of published assessment criteria, graded by different strands. The supervisor's job is to judge which level a student's work falls into for each strand, and to write specific feedback on how to improve. The difficulty is that an Extended Essay easily runs to several thousand words, and the supervisor has to check it against the criteria over and over across drafts, under real time pressure.
The AI for Empowering Learning and Teaching Funding Programme asks schools to run AI-assisted teaching in at least three subjects. A school offering the IB can bring the feedback process for the IA and Extended Essay into this, as long as the tool it uses genuinely matches the IB assessment criteria rather than a generic essay score.
How AI supports feedback for the IA and Extended Essay
Drafting feedback against the IB assessment criteria
The reliable approach is for the system to hold that subject's IB assessment criteria, give a preliminary level for each strand, and draft criterion-linked feedback that points out which strand a student still falls short in. The supervisor moves from grading from scratch to reviewing and correcting, first reading the system's criterion-referenced judgement and then deciding whether to agree. The draft feedback is a base copy for the supervisor to edit, not a finished piece for the student.
Every grade backed by the student's own words
Every level should point back to a specific place in the student's work, so the supervisor can see the basis for the judgement at a glance. A score with no supporting text cannot be reviewed, and it is hard to explain to the student. With evidence item by item, the supervisor's review becomes a quick confirmation, handling only the points in doubt.
Supporting criterion checks across drafts
Progress on the IA and Extended Essay comes from repeated revision. When a student revises according to the feedback and submits again, the system re-checks the revised draft against the criteria for a fresh first pass, and the supervisor only reviews the strands still in question. The feedback cycle shortens, students can polish a few more drafts before submitting, and the supervisor's load does not rise in step with the number of drafts.
The line on academic integrity
The IB has clear requirements on academic integrity: the Extended Essay and internal assessment must be the student's own work. This line needs to be set out before AI is used. AI is for giving feedback on what the student has already written, not for coming up with ideas, drafting, or rewording sentences on the student's behalf. Diagnosis points at the problem; the revision is still for the student to do.
A school would do well to write this into its own AI use and academic integrity policy, and to have supervisors keep a record through the guidance process, showing how the feedback was produced and how the student revised on that basis. This protects the student and the supervisor's professional judgement alike. For the final grade on the IA and Extended Essay, the decision must rest with the supervisor and teacher, and AI only provides a reviewable first pass.
What to check when choosing an IB AI marking tool
- Which criteria are built in: whether they are the IB assessment criteria for the matching subject, graded by strand, rather than a generic essay score
- Whether each grade carries the source text: try marking a few IAs or Extended Essays and check that every level points back to a specific passage in the work
- Whether the feedback is framed clearly: the system should draft a first pass for the supervisor to review and help guide the student to revise, not reword sentences for them
- Whether the supervisor's review is quick: editing grades and feedback should take only a few steps, and keep a record for academic integrity and reporting
- How student work is handled: where it is stored, how long it is kept, and whether it is used to train models must comply with the Personal Data (Privacy) Ordinance