Grading (Mid-Semester Feedback)

by Mike Gleicher on March 23, 2017

Update 3/25: My Python Script should have given you feedback as a comment in the gradebook for the assignment for the mid-term evaluation. The first version has some formatting problems making it hard to read, so I will  try to repost.

I promised people feedback in return for doing the mid-semester eval for me. Since a lot of you did the eval (thank you!) there’s a lot for me to do. The upside is this is good practice for final grades and to nail down some of the things that have been left undone.

Here is the information I have:

  1. Grades (on the silly 50 point scale) for the discussions
  2. Grades (on the silly 50 point scale) for the seek and finds
  3. Attendence estimates from Chih-Ching
  4. Quantitative stats (number of posts, length of the longest post) from all discussions (including the seek and finds)
  5. Some rough qualitative sampling of the quality of the discussions (I read through at least 1 assignment and 1 seek and find for everyone)

I have no information on the Design Challenge (it hasn’t been graded yet)

The measures we have are rough:

  1. The grading of discussions gives most people 50 on most assignments. This doesn’t distinguish the “great” (A) from the “good” (AB). There is some noise (some 50s should be 40s, or vice versa, sometimes late grades don’t reflect otherwise good work, …). But overall, if you apply a robust statistic to it (e.g., drop the lowest scores) it says that a person’s performance.
    Most students are in the “usually get 50” category – which will be an AB or A, as determined by closer examination. Others are “usually 40 or 50” or “usually 50 except when sick” – the former is likely to be a B or AB depending on #2 below.
  2. The quantitative measures are not necessarily measures of quality – but they do seem to be well correlated with the actual subjective reading (until someone tries to game the system by padding their posts – don’t try it). This will be how we distinguish “good” from “great” for the discussions (with some manual checking for validation). We’ll adjust within the grade bands (see #1) based on this information.
  3. Checking for attendance is a proxy measure for actually participating in class. However, it seems that most people who come to class do participate in the in-class exercises, and I’ll assume that you’re listening enough to the monologues. There’s also some noise in seeing whose there (we don’t check every day, and Chih-Ching may not be perfect at knowing everybody). Most people are there “almost always” – so we’ll assume that’s not a problem. For a few people attendance is a problem – in most cases, we’ve discussed the situation. But if you’re a significant non-participator, expect a penalty
  4. Grading the Design Challenges is tough too. It’s pretty subjective. I expect people will generally do well.


  1. The 3 design Challenges will be graded on an A-F scale and averaged. They will count for 1/3 of your grade.
  2. The 15 weekly assignments and 1 Design School Assignments + the peer critique assignments: we’ll robustly average (drop the lowest N ~= 2-3) your “grades” to determine a grade range. We’ll use the quantitative metrics (robust average of [# posts, length]) with a qualitative check to adjust (usually, this will put people to the top of the range). They will count for Y% of your grade.
  3. The 15 seek and finds: we’ll robustly average (drop the lowest N ~= 2-3) your “grades” to determine a grade range. We’ll use the quantitative metrics (robust average of [# posts, length]) with a qualitative check to adjust. This will count for 1/4 of your grade.
  4. If your participation is problematic we will penalize your final grade.

What you’ll get in your “feedback” (which I will try to post into the Canvas comments for the mid-term eval “assignment”):

  1. numerical scores (robust averages of canvas assignments)
  2. quantitative metrics (robust averages of max post length, robust average of number of posts – which are “normalized” by subtracting the median # of posts on that assignment for the class)
  3. our estimate of how many of the 10 time chih-ching checked attendance you were there (we did not count the 1st weeks when there were enrollment problems)
  4. brief comments interpretting 1-3
Print Friendly, PDF & Email

Previous post:

Next post: