Moderation and Comparable Judgement
Two teachers at the same school each teach a Korean beginners class. Teacher A's students receive school assessment marks averaging 82, while Teacher B's average 68. Both used the same assessment plan and the same performance standards, but Teacher A's tasks included a heavily scaffolded dialogue with sentence starters provided, while Teacher B's tasks required unscripted interaction. During a moderation meeting, the panel compares samples of student work against the standards descriptors. Several of Teacher B's students wrote accurate but simple responses; several of Teacher A's students produced polished dialogues closely following the provided frames. The panel must decide whether the marks are comparable and what adjustments, if any, are warranted.
1. Identify the main threat to comparability and recommend a moderation decision.
Read the worked answer
The main threat is task equivalence. Teacher A's scaffolded dialogue supplies much of the language, so a polished product may reflect the frame rather than the student's independent control of Korean. Teacher B's unscripted task demands more spontaneous production, so similar-looking work represents different levels of achievement.
Moderation should compare annotated samples against the standards descriptors, focusing on evidence of independent language use. If Teacher A's tasks allowed students to demonstrate less than the standard requires, the panel should not automatically raise Teacher B's marks to match. Instead, it should judge each sample on the evidence and consider whether Teacher A's marks overstate achievement because the task was too supportive. A defensible outcome is to adjust Teacher A's marks downward where the work does not evidence the standard, or to note the task limitation and require future tasks to reduce scaffolding. The decision must be recorded with reasons.
2. Explain why simply averaging the two teachers' marks to 75 each would be an inadequate moderation outcome.
Read the worked answer
Averaging treats the discrepancy as a calibration error shared equally by both teachers, but the evidence points to a task-design difference, not random drift. Forcing both sets to 75 would move students whose work genuinely met the standard and students whose work did not, producing new errors rather than correcting existing ones.
Moderation is about comparability of judgement against the standards, not equalising distributions. A class can legitimately have a higher or lower average if its achievement differs. The panel should examine individual samples, confirm that judgements align with descriptors, and adjust only where evidence shows misalignment. Averaging also removes the incentive to design demanding tasks: a teacher whose rigorous task yields lower marks would be rewarded by having them raised. The recorded rationale should explain the task-equivalence finding and any adjustments made.