Hacker News Reader

Comment by Wowfunhappy | Hacker News Reader

While the college admissions process is by no means a perfect system, if you're accepted to Harvard, you're probably a pretty darn good student.

If Harvard graded on a curve, such that half of their students—who presumably got straight A's in high school—received a C or below, would that be more fair? Would it be a better representation of student performance?

I suppose the answer depends on the purpose of giving grades. If the goal is to see who outperformed who, Harvard's average GPA is a major problem. If the goal is to see who learned the material, then Harvard is providing an excellent education.

View on HN

This is always a fun question, and a lot of it depends on the material/subject matter, but why not both?

For things like math, science, and engineering--where the possible depth of knowledge and understanding far exceeds what could be learned in a semester course--I personally think exam grades should center around 50. If you've got 100 points to play with, why not use them?

A lot of my classes did, and I always found the challenge exciting: what stuff did I not learn about this subject? This also gives you the ability to differentiate student performance, should that be of importance.

At the end of the semester, things were usually (non-linearly) curved up to average around B. The instructors (well, most of them) knew what "absolute score" represented enough understanding to move on, and anything below that would not get curved into a passing grade.

This is totally right. There are two kinds of classes: those with a specific set of learning outcomes to meet (eg. a prerequisite course) and those where by the end the in-scope knowledge is essentially unlimited (eg. courses where it's not unreasonable to prove named theorems on homework or exams). The former should be graded with 100% being all the knowledge, the latter in a way that gives a fair distribution and then scales.

Of course, my fellow university students always push back on me when I support the latter :) At the risk of armchair psychoanalyzing, I think a lot of them don't want to reckon with the fact that there's such an overwhelming amount of information out there and so many perspectives to explore it from that "100%" in say introductory topology is almost completely meaningless as a measure of how much topology one actually knows.

UK unis are a lot better at using the full grading scale, in my experience. 70% gets you a 'first class' [0], but there's a lot of room in the 70-100 range to differentiate performance. It was a nice system because there's almost always space for able students to really push the boundary. The idea is that getting >80% is for students doing really exceptional / original work.

[0] A "1st" (A) is between 70-100%, "2:1" (B) is 60-69%, "2:2" (C) is 50-59%, and "3rd" (D, lowest passing grade) is 40-49%. UK degrees are awarded one final grade at the end, rather than a GPA, usually based on your last two years of study only.

Reply on news.ycombinator.com