Friday, April 20, 2007

Reworking the Grading System

Chapter 7 of What the Best College Teachers Do talks about evaluating students. While I can't recall any specific passage or thought from the book that got me thinking this way, it did prompt me to think more about grading systems.

What does it mean for a student to have a problem "80% correct"? If I cannot answer this question, then it makes no sense for me have some conclusion that sounds like "therefore, 80% is an B- in my class." This is one of the reasons I am thinking of grading problems out of 5 points (see the last part of this post). I want to have a grading system that is simple enough to be consistent, but diverse enough that there are enough strata to have an accurate gauge on students in the class.

When I look at student scores right now, I see a spreadsheet that only has the final scores on tests. That means in a hypothetical 4 problem exam, I could not tell the difference between an 80-80-80-80 student and a 100-100-100-20 student. I would feel that the 80 student is probably a stronger student than then 100 student because the 80 student has shown that he has an understanding of all the topics, but the 100 student has not developed the same breadth of knowledge. This is also an indication of me not really knowing what it means to be 80% correct. Alternatively, it could also be that the 100 student just made some small error at the beginning of the last problem and that ruined the rest of his otherwise perfect work.

My current thought is data intensive, but sounds like something I might be willing to do with three small classes. It might be beneficial to keep a record of all the grades on the individual problems (homework and exams). This has multiple benefits:
  1. I would be able to see overall trends in the class. If there are specific sections that are giving students more difficulty than others, it will be immediately apparent and give me a chance to cover that material more carefully.
  2. It can get students away from thinking about their understanding in terms of percentages. I could assign grades based on the relative numbers of points they got on their problems, and not the sum of their scores (this would also require me to use a grading system of 4-5 points for every problem so that it is a consistent measure).
  3. It may also provide good feedback for students if this was presented to them in a reasonable manner. It would have to be organized in a nice way and not just a list of numbers.
  4. It will also provide historical data to recognize (for example) that most students really struggle with this particular topic, this problem is extra tricky, or other observations that could potentially slip through unnoticed.
As always, the devil is in the details because I have no idea at this point how such a system could be presented to a class in a way that would make sense to them. I need to think about this a little bit more.