Tuesday, December 20, 2005

A teacher's dilemma

I am a "formula" teacher, when comes to grades, which means I pre-decide the formula of grade calculation before I start a class. This is because I tend to get very personal with things and I don't want my personal attitudes get in the way of fairness.

What happens from time to time, is that good students I have known for a semester fail the final exam. This troubles me. On one hand, I know they studied hard, enjoyed the class and understood the material well. On the other hand, they didn't perform in the final.

Exams are not perfect but they are regarded as common means of evaluation in modern education. I know relying on fair exam is actually only partially fair. But, relying on my personal impression of the students would never be fair. So I choose to rely on the exam scores and my formula even though I am really really bothered by that.


Andrew Gelman said...


I think the best way to test, for the goal of evaluating student abilities, is to have many short questions. Basic psychometric models suggest that this is the best way to differentiate individuals given a fixed amount of testing time.

Tian Zheng said...

According a small simulation study of mine, giving partial credit to each correctly executed step has the same effects of many short questions. And that was my grading principle.

What troubled me in this post was that sometime (by a not very small chance), good students under-perform. There is just not so much we can do about it.

Andrew Gelman said...


My problem with partial credit is that the scores on the different parts of a question will be correlated (I think) leading to less discrimination in the total score.

Regarding your second point, is this just an instance of measurmenet error (also called "regression to the mean") that will inherently occur in any measurement setting?

Tian Zheng said...


Partial credit could be designed to have less correlation. I agree with your point on the "regression-to-the-mean" effects. It is definitely noticable. What I felt strong about is the evaluation limitation of exams. Exams sometime can't show you how great a student is. It is just like presentations sometime can't show how great a researcher is.

Anonymous said...

I suspect it is your own DEFINATION of "good students" that might be bothering you much instead of contingent low grades they obtained.

As you mentioned, "I know they studied hard, enjoyed the class and understood the material well.", then they are "good" which is irrelevant with their grades. Students might feel "slightly" uncomfortable when teachers distinguish or quantify them by numbers though it seems that everything in the modern world is quantifiable.

SEO Firms said...

Great formula teacher! I really appreciate your fairness.. anyway " Grades do not measure intelligence"