Partial Order Curved Grading
For some assignments grading is more subjective. When appropriate, we use partial order
curved grading as follows:
- Scan through a few submissions to find one that appears particularly well done. This is
our starting point.
- Read the next submission, and based on the general grading rubric place it either above
the existing paper—or below it—in "the stack," based on whether it is of higher or lower
quality than the existing submission.
- Read the next submission and place it correctly in the stack, sorted top to bottom, highest
quality to lowest quality.
- Repeat the previous step until all the submisisons are in the partial-order stack. (Partial
order allows for ties.)
- Scan through all the submissions one more time, adjusting as necessary.
- Starting at the top, work down in the pile until you reach the lowest full-credit
submission. The quality level of this submission is used as the 100% comparator.
- For those submissions further below in the stack, assign percentages based on the
comparator. For example, if the only thing we cared about was word
count (and that is never the only factor!), then if the
lowest full-credit submission had 2,000 words in it, then a 1,000-word submission would be
worth 50%, a 500-word submission 25% and so on.
Subjective-assessment, partial-order grading is necessarily somewhat imperfect. For this reason
we typically artificially slightly boost all non-100% grades. In this way, your
grade will not be lower than it "should have been," but it might be higher.
We cannot engage in grade negotiation based on subjective criteria, which would bias the grading.