How do you mark a VCE SAC against VCAA performance descriptors?

What actually makes a VCE SAC hard to mark?

The hard part of a SAC is not the reading. It is that two separate obligations run at once. You have to apply one standard consistently to every response in the cohort, and you have to be able to say the work is genuinely the student’s own. Miss either and the mark is not defensible, whatever the feedback says.

The consistency problem is specific to how VCE works. VCAA publishes performance descriptors alongside its study designs, and they are advisory. They exist to support holistic judgements about a student’s demonstration of the key knowledge and key skills for each outcome, not to hand you a mark. You build the assessment tool. That freedom is useful, and it is also where standards drift, because two teachers can read the same descriptor row and land a band apart.

The authentication problem is separate, and it does not wait until marking. VCAA expects assessment tasks to be part of the regular teaching and learning program and completed mainly in class time. Where any part of a SAC is undertaken outside class, teachers must monitor and record each student’s progress through to completion, with regular sightings noted on the Authentication record for school-based assessment form. The record you build while the task runs is what you stand on if a mark is later questioned.

How do you turn performance descriptors into a marking tool you can actually apply?

Start from the outcome, not from the descriptor grid. The study design sets the outcome and its key knowledge and key skills, and all criteria in any assessment tool must be drawn directly from the study design. VCAA’s advice is that teachers use professional judgement to decide which parts of the descriptors to select. So take the rows that match what the task actually asks, and drop the ones it does not test.

Then fix the language. Descriptors describe quality in general terms so they can cover every cohort in the state. A usable marking tool converts that into evidence you can point at in a script. A row about analysing complex cause and consequence becomes something closer to: identifies more than one cause and weighs their relative importance using specific evidence. Same standard, but now two markers can agree on whether it is there.

Sanity check the range before you mark anything. The lowest level should describe something most students in the class can do, and the highest should stretch your strongest. If the top band is unreachable, everyone compresses into the middle and your rank order carries much less information than it should.

• The outcome and the key skills this task assesses
• The descriptor rows you are using, and the ones you deliberately dropped
• What counts as evidence for each criterion in this specific task
• Marks or weighting per criterion, agreed with anyone else teaching the study
• Two or three annotated benchmark scripts, kept once you have marked a sample

Why does your rank order matter more than the marks you award?

This is the part teachers new to VCE most often get wrong. VCAA statistically moderates School-assessed Coursework against the external examination so that assessments from different schools sit on a common standard. Statistical moderation does not change the rank order of students as determined by the school’s school-based scores. Only the numbers move.

The practical consequence is blunt. Marking your cohort generously does not help them, because the generosity is adjusted out. Marking harshly does not hurt them, for the same reason. What survives moderation is the order you put students in and the size of the gaps between them. The useful question about a mark is therefore not whether 32 out of 40 is fair in the abstract, but whether this student is correctly placed against the others.

That reframes where your effort belongs. Chasing absolute calibration against an imagined statewide standard is mostly wasted. Applying one internally consistent standard across your own cohort is not. It also tells you which scripts deserve a second look: the ones sitting on a boundary between two students, rather than those comfortably in the middle of a band.

What does authentication require while the SAC is still running?

Authentication is a process you run during the task, not a declaration you collect at the end. The rules students must observe are plain. Work submitted for assessment must be their own, they must not plagiarise or cheat, they must acknowledge all resources used, and they must not receive undue assistance from another person. VCAA is explicit that unattributed or disallowed use of generative AI in assessment may be a breach of academic integrity, and that it has to be investigated under the school’s policy.

Your side of it is the record. Where a SAC or part of one is completed outside class time, you monitor and record each student’s progress through to completion, which means regular sightings and notes on the authentication record. Plans, drafts, redrafts, observations in class. This is why marking and authentication are hard to pull apart. The process evidence tells you whether the finished script matches the student you have watched all term.

The awkward cases are usually about preparation rather than the task. Students rehearse with practice pieces, past papers and study guides, and none of that is a problem. General prompting and advice that leads a student to refine or self-correct is allowed. The line sits at undue assistance on the assessed work. Being explicit with a class about which side of that line each form of support falls on removes most of the trouble before it starts.

How do you hold one standard across a whole Units 3 and 4 cohort?

Mark by criterion, not by script. Take criterion one across all thirty responses, then criterion two. It is faster, and it stops a strong opening paragraph from quietly lifting every other criterion in the same script. Halo effect is one of the easiest inconsistencies to design out.

Then deal with ordering effects. Shuffle the pile so it is not in class-list order, since the same students should not always be marked while you are fresh. Take a break at the point where you notice your comments getting shorter. When the pile is finished, re-mark the first five scripts against the tool without looking at what you gave them. If they move, your standard drifted, and the answer is to re-mark rather than talk yourself out of it.

If more than one teacher runs the study, cross-mark before anything is released. Pick four scripts spanning the range, mark them independently, and compare more than the totals. Compare which criterion each of you used to get there. Two markers reaching the same total by different routes is a warning sign, not a result, and that is the disagreement worth having in a faculty meeting rather than after reports go home.

Where does AI-assisted marking sit within these rules?

Two questions get muddled here. The VCAA rules described above govern student work: whether the response submitted for assessment is the student’s own. They are not a ruling on how a teacher drafts their own comments. You are the assessor either way, and any tool that produces a mark is producing a suggestion you accept, edit or reject. Your school’s assessment and AI policy governs your practice, so read it before you change how you mark.

Within that, the sensible application is the criterion-by-criterion drafting that eats the evening. A tool holding your criteria, the descriptor rows you selected and your usual comment bank can draft a first-pass mark and comment for every script in one run, applying the same wording to the same evidence across the cohort. Jeddle is built for that kind of criteria-based drafting, and most teachers start with expert essay marking on a practice task rather than an assessed SAC.

Practice tasks are the safer and more valuable starting point. Detailed feedback on a rehearsal piece two weeks out changes a result. The same feedback on the SAC arrives after the mark is fixed. Whatever you use, keep the human decisions where VCAA and your school expect them: on the mark, on the authentication judgement, and on the record.