What are the two markers trying to agree on?
Feedback consistency means that comparable evidence leads to defensible teaching decisions. Two comments need not use identical sentences. They should recognise the same relevant features of the response and propose next steps that fit the task. A meeting that edits everyone’s wording into one style can miss a much more important disagreement about what the student needs.
AERO’s formative assessment guidance connects feedback with evidence of understanding and decisions about what to teach next. The protocol here applies that principle to an original two-marker exercise. It is a practice routine for examining comments, not an official moderation procedure, a required sample size or a reliability benchmark.
What should each marker see before comparing comments?
Provide the exact task, the criterion being examined, any permitted method and the same version of the response. For a practice meeting, use invented work or appropriately handled classroom material. Ask each marker to note what the response shows, the evidence supporting that judgement and one proposed next action before seeing the other comment.
Keep the task narrow enough to discuss properly. In the example below, the criterion is “calculate the arithmetic mean and explain a valid method”. Assume students have already been taught to add the values and divide by how many values there are. If this has not been taught, the appropriate next step may be an explanation and model rather than an independent correction.
Do not let one marker supply extra facts about the learner that the other does not have. If classroom context changes the decision, state what that context is and reconsider together. The response alone cannot establish that a student was careless, rushed or deliberately avoided the task.
What does this original practice response reveal?
Task: “Find the mean of 4, 6 and 8. Explain your method.” Invented response: “6, because the middle number is always the mean.” Marker A writes, “Correct. Move on.” Marker B writes, “Test your rule on 2, 3 and 10. Then find the mean using the total and the number of values.”
The answer 6 is correct for the first set: (4 + 6 + 8) ÷ 3 = 6. The explanation is not a general rule for the arithmetic mean. In the ordered set 2, 3 and 10, the middle value is 3, but the mean is (2 + 3 + 10) ÷ 3 = 5. That changed case exposes what the first set concealed.
Marker A has recognised the correct number but has not addressed the explanation required by the criterion. Marker B targets the faulty rule and offers a way to examine it. The substantive difference is whether the reasoning needs attention. It cannot be settled by noting that both markers accept 6 as the first answer.
Before sending B’s comment, check whether the student can carry out the comparison. If they cannot yet calculate the mean from the total, model that method with them. A counterexample can reveal that a rule fails; it does not automatically teach the replacement.
How do you resolve the disagreement without voting?
Put the response and criterion beside both comments. Ask A to identify which words establish a valid method. Ask B to identify the claim their new task tests. The phrase “always the mean” is the evidence that matters here. The team can then record that the numerical answer is correct but the explanation needs revision.
An agreed comment could be: “Your answer is correct for 4, 6 and 8. Check whether the middle value is also the mean for 2, 3 and 10. Use the total divided by the number of values, then explain why your original rule does not work for every set.” Follow this with teaching if the student cannot use that method independently.
Record the reason for selecting this action, not just the final sentence. In this case: “A correct result can coincide with an invalid general rule; the criterion requires a valid explanation.” That note helps a colleague recognise a similar problem in a different response.
Agreement is still only a result of the discussion. Both markers could overlook the same error. Check the calculation, source or task requirement directly. If the criterion is ambiguous, keep the disputed comment out of a shared bank until the responsible teacher clarifies the task. A majority preference cannot supply missing evidence.
Which wording differences can you leave alone?
“Add all three values and divide the total by three” and “Find the total, then divide by the number of values” can express the same method for this task. One is more specific; the other names a reusable operation. The choice depends on what has been taught and how much support the student needs.
By contrast, “show more working” leaves the faulty middle-value rule untouched. The response already contains a reason, so requesting more explanation may merely produce a longer incorrect explanation. Classify the difference as a diagnosis or next-action issue before spending time polishing the prose.
The guide to turning a rubric descriptor into a student action can help when markers agree on the weakness but cannot say what the student should do. Keep that translation tied to the criterion and the response rather than adding a stock instruction to every script.
How do you check whether the shared decision carries over?
Give the markers a second invented response: for the set 2, 3 and 10, the student writes, “The total is 15. There are three numbers, so the mean is 5.” This meets the stated criterion. A useful acknowledgement names the valid method. Sending the earlier correction about the middle value would respond to a problem that is no longer present.
For another case, use: “For 2, 3 and 10, the total is 15. I divide by 2 because there are two gaps, so the mean is 7.5.” The addition is correct. The new teaching point is counting values rather than gaps. Ask each marker to locate that decision and propose a check that shows why the divisor is three.
These cases distinguish agreement on a reusable principle from copying one comment. They do not establish the reliability of a whole faculty’s marking. Use further responses that reflect the tasks your team actually assesses, and retain cases that reveal unresolved differences for a later discussion.
What should you keep after the meeting?
Keep a short record of the task and criterion, the response detail, the two proposed actions, the agreed decision and its justification. Add the next case and what it revealed. The table below provides a copyable record; the printable exercise contains the first task and both comments for independent review.
When adding wording to the feedback comment bank, include the conditions for using it and the evidence that would make it inappropriate. The second response above is an example of when to leave the correction out. A shared bank is more useful when it preserves these boundaries.
If your department is exploring JeddAI, get in touch with an invented task and the decisions you want teachers to check. Use the same evidence review for proposed feedback from any source. Contact and meeting links are the main next step here; individual teachers can also create an account through the secondary link.
| Response evidence | Decision |
|---|---|
| Correct 6; says middle value is always the mean | Test the rule with a counterexample and teach or check total divided by count. |
| 15 ÷ 3 = 5 with a valid explanation | Acknowledge the criterion met; do not repeat an unnecessary correction. |
| 15 ÷ 2 = 7.5 because there are two gaps | Focus on counting values, not gaps. Keep the correct total. |
| Field | Your notes |
|---|---|
| Task and exact criterion | ____ |
| Response detail each judgement rests on | ____ |
| Marker A diagnosis and proposed action | ____ |
| Marker B diagnosis and proposed action | ____ |
| Substantive difference, decision and reason | ____ |
| Unresolved question or teaching required | ____ |
| New case and what the review showed | ____ |



