IB students are moving to screen-based exams, and plenty of them will practice the interface. The mark scheme that decides their results is meant to stay the same. In May 2026, the International Baccalaureate ran its first live session of digital exams for the Diploma Programme and Career-related Programme with a selected group of schools, and it says digital exams will initially mirror existing paper-based assessments to ensure fairness and comparability. Both formats run from 2026, and schools can keep administering paper exams during the transition.
Interface practice helps, but it doesn’t answer the harder question of how work earns marks. A correct answer doesn’t always carry full marks, in either format.
Four ideas run underneath that gap. In IB mathematics, marks go to working and method, not just final answers. Applying criteria is a trained skill, not something people absorb. Easy proxies can mislead both AI graders and students. And seeing examiner reasoning question by question changes what students get out of past papers.
In IB Mathematics, Working Earns the Marks
On many IB mathematics components, a student can understand a concept and reach the right result, yet lose marks because the working doesn’t show what the mark scheme credits. In the IB’s Mathematics: analysis and approaches specimen papers, the candidate instructions say: “Full marks are not necessarily awarded for a correct answer with no working.” Examiners are looking at the working on the page as well as the final line.
The same instructions say marks may be awarded for a correct method even when the final answer is wrong. A student who checks only the final answer misses both sides of that rule: the marks lost when a correct answer isn’t supported by working, and the method marks a wrong answer can still earn.
The rollout is phased. From November 2026, all IB World Schools will be able to opt into digital delivery for Language and Literature and Language Acquisition. The IB plans to add more subjects each year through 2029 and to phase out paper exams by the early 2030s. It says exam content is identical across formats and that grade boundaries are set so students can achieve the same results either way. Matt Glanville, the IB’s Director of Assessment, bases the case for digital exams on how students engage with assessment, through accessibility, control and interactivity. His stated rationale is accessibility and student control, including videos, interactive tools, moving images and simulations in subjects like the sciences.
For students, that adds up to a long transition in which delivery keeps shifting, subject by subject and year by year. The thing they can plan around is how the mark scheme allocates credit, and that preparation carries over whichever format they sit.
Rubric Application Is a Trained Discipline and Not an Absorbed One
Written criteria don’t apply themselves. In a study published in Language Testing, Weigle (1998) applied many-facet Rasch measurement, a statistical method for modeling rater behavior, to a writing assessment to examine the effects of rater training. The paper reports that training changed how raters behaved, although differences between individual raters remained.
ETS shows what structured training looks like in practice. Educational Testing Service (ETS), a nonprofit testing organization, scores constructed-response tasks with trained raters. Its guidelines for these assessments advise giving scorers specific instructions and clear examples of responses that define the score categories, rather than leaving criteria implicit.
One of its programs, the Performance Assessment for School Leaders (PASL), is scored by ETS-trained raters who use rubrics describing qualitative levels of evidence, not a right-or-wrong key. Before scoring, raters must show they understand the relevant standards, task directions and rubrics. They practice scoring in sessions run by expert trainers, and they must “take and pass a certification test” before they score operational responses. The aim of all these steps is consistency: scores that hold steady from one rater to the next.
Catherine A. McClellan, Director of human constructed-response scoring in the ETS Research & Development Division, makes the underlying point in an ETS R&D Connections brief on constructed-response scoring. She writes: “… content knowledge alone is not sufficient—a rater must be trained in the specific procedures needed to score the responses to the particular test or item he or she will be scoring.” PASL’s practice sessions and certification test follow the same principle. Students face a version of it too: they can know the subject well and still lack the marking procedure, the skill of applying criteria the way the mark scheme does. A student who only checks whether the final answer matches the key gets no training in how to apply those criteria from that check alone, because the key is all they see.
What Is Easy to See and What Earns the Marks
If trained raters still need training to apply criteria, it’s little surprise that students reach for something quicker. A student finishes a question, sees that the final answer matches the key, and moves on. That feels like success. It also leaves a blind spot between what the student can see, a matching answer, and what the mark scheme rewards, which is working and method.
Recent research on AI marking shows how an easy proxy can pull a grader away from the criteria. A University of Cambridge-led study under the OpRaise project, reported in the paper “AI in University Assessment: Evaluating the Opportunities and Risks of Automated Marking,” tested three frontier generative AI systems, including recent versions of Claude and ChatGPT, on 761 undergraduate psychology essays from the University of Cambridge, Manchester Metropolitan University and the University of Nottingham. Across those essays, the systems matched human-awarded UK degree classification bands only 35–65% of the time: about 63% for Cambridge essays, 53% for Nottingham and 35% for Manchester Metropolitan. Their marks drifted toward the middle of the scale. Essay length, vocabulary range and sentence complexity could lift AI marks even when they did not reflect academic quality. Length and vocabulary are easy to count. Whether the argument holds up takes actual reading.
The authors recommend that final grading decisions remain with human assessors. Their concern is that polish and length can earn credit that genuine quality should have earned.
I’d argue students can fall into the same habit. Whether the final answer matches the key is quick to check, and it can crowd out the criteria. Panadero and Jonsson (2013), in a review of formative uses of rubrics in Educational Research Review, report that explicit criteria can clarify what is expected and support self-assessment and self-regulation. They also report that the benefits depend on how rubrics are used, not simply on handing them out.
Students also need to see those criteria applied to actual work.
Making Examiner Reasoning Visible Question by Question
Checking answers alone leaves examiner reasoning and mark allocation hidden from students. Atkinson, Derry, Renkl, and Wortham (2000) reviewed the research on worked examples. They found that worked solutions can be especially helpful for novices, and that the structure and sequence of examples affect how much learners gain. Applied to past papers, a worked explanation can show why each step earns marks.
Revision Village pairs each question in its Questionbank with a written markscheme and a step-by-step video solution produced by experienced IB educators, so each practice question doubles as a window into how examiners actually award marks. Take a student who gets a question wrong. The markscheme shows where method credit sits, and the video solution shows how a complete response is built, step by step. That connects directly to the specimen-paper rule that method marks may be awarded even when the final answer is wrong.
There’s a small shock in realizing that marks attach to individual steps, not just the final answer. Revision Village puts that realization in front of students question by question: they can set their own working beside the markscheme and the video, and the question shifts from whether the answer matched to which steps carry the marks.
A Different Question to Ask of the Same Past Paper
The usual diagnostic question is “how many did I get right?” A better one adds a second: “can I explain what earns marks and why?”
That second question works on paper or on screen. Revision Village gives students a view of how marks are earned while they work through each question. ETS’s rater training is a reminder that applying criteria is a skill even professional scorers have to be taught.
However the exam arrives, the mark scheme is waiting at the end of it. The student who can trace how their own working earns credit walks in already knowing what is being graded.
