In thinking about how to grade, it is worth understanding that there is no consensus on the purpose or meaning of grades. With an unclear objective, it is understandable that we’ll end up with unclear or contradictory recommended methods for reaching that objective.
TAs and instructors are grade producers; list five different grade consumers:
We have a list TAs gave in previous semesters in a footnote1.
For each of the consumers listed above, why do they care about grades? What do they hope grades tell them?
We have a list TAs gave in previous semesters in a footnote2.
Note that several grade consumers want some form of who’s the best
measure. Best
is a broad topic, which is probably worth its own section.
Many things contribute to student success in courses. For example,
If you could grade each of the four items above independently, how would you combine them to create an overall grade?
In a mailing list of professional CS educators, Arnold Pears once wrote “Learn the skills
should be synonymous with Pass the course
“. That assertion went without challenge, and aligns with many of our intuitions, but there are real reasons to want other components to count. Indeed, there is an argument that your grade should be the maximum of your understanding, time management, and social skills because if you are good at any of them you can provide value to a company, either as a developer or manager or marketer or assistant or trainer or …. Or perhaps it should be the minimum of your understanding, time management, and social skills because if you are bad at any of them then I don’t want you as a co-worker. Or …
B+ understanding?
A common question I hear instructors and TAs ask when trying to evaluate the quality of a grading system is does this student deserve the grade they got?
That question presupposes a definition of the intended meaning of each letter grade. But definitions of grades are surprising hard to come by.
The lore I was told by multiple senior faculty when I was staring in academia, and am repeating here without reliable sources because I have not found them, is this:
Once upon a time grades were given in a fairly piecemeal way and were mostly consumed only internally to the generating institution. Then the central limit theorem and the prevalence of bell curves came into vogue, and a new self-calibrating
grading system came along:
grade on whatever scale is handy
fit the resulting grades to a bell curve
announce grades in standard deviations of the curve:
C= within 0.5 standard deviations of the mean
Band
D= 1 standard deviation from
C
Aand
F= 2 or more standard deviations from
C
This model became known as The Curve
and grading using it called grading on The Curve.
But it had (at least) three problems:
Real grades are not distributed on a perfect bell curve. They are not independent variables, so the central limit theorem does not apply. In practice, they tend to be fat-tailed ditributions, often with multiple peaks.
Not many people are two standard deviations from the mean. People want to tell more people they are good, and to deter more people from continuing by giving them bad grades.
Grading on a curve creates competition, including incentives for academic sabotage: the worse other students do, the more likely your knowledge is to be graded highly.
On the positive side, the wide-spread popularity of this grading system meant that it became possible for grades to be interpreted outside of the granting institution. That had a slow but dramatic impact on the overall perceived importance of grades.
If you define C
as average student
and then tell a university that most of your students are at or below average
that rubs most universities wrong. Clearly we are the exception: almost all of our students are well above average!
Result: grade inflation. Grades are still based on relative performance but now the cut-offs between letters are flexible and can change, and generally move toward higher grades over time. More and more, universities are even allowing each course to pick their own curve so that a student who is reliably at the 80th percentile in every class may get A+ in one class and B- in another.
UVA does not have a definition of the goals of grades. The School of Engineering and Applied Sciences at UVA does not have a definition of the goals of grades. The College of Arts and Sciences does have a standard grading scale
mapping points to letters, but does not define what a point is or how easily they should be given out.
The CS department does have a definition of the intended meaning of each grade, created by a committee of our faculty and approved by a vote of our faculty: http://ugrads.cs.virginia.edu/grading-guidelines.html. As a rough summary:
| Letter | Word | Learning | Prediction for next course |
|---|---|---|---|
| A | Good! | mastered all | you’ll do well |
| B | Okay | competent in most | able to to well, though review recommended |
| C | Caution | sufficient | it’ll be challenging, but possible with review |
| D | Danger! | minimal | you’ll probably fail |
| F | No | insufficient | you’re definitely not ready |
Suppose you have graded a student on three assessments measuring similar ideas: they got 40% on the first, 50% on the second, and 90% on the third. You should give them an overall score of
This is also not a topic of consensus among our faculty, though fortunately most TAs are not asked to weigh in on this. But whether to permit grade replacements or not can be a contentious topic among those who are asked to consider it.
Suppose you are assembling final grades for a course with 4 major topics. Student X earned 75% in all 4 topics. Student Y earned 90% in 3 topics and 30% in one topic.
This is also not a topic of consensus among our faculty, though fortunately most TAs are not asked to weigh in on this. Some of our faculty argue that if a student can pass a prereq without learning all the prereq content, we are setting them up to fail in later courses. Other faculty argue that one bad topic does not make someone a bad student, or even that specialists are more valuable than generalists and that this kind of topical-differentiated grades are a good sign.
smart-personclubs and awards
good at school? Will they succeed if we admit them?
smart-personclubs and awards – This pool all look good; who’s the best of them?