Fog Creek Software
Discussion Board




Curving CS Grades

I'm taking a graduate level CS course, and our final grades are curved.  I can't remember the last time grades were curved for a CS class I've taken, and I think it's ridiculous.  I'd kinda' like my grade to be based on the knowledge I've acquired relative to some objective criteria - not on how good are bad the other nine students happen to be the semester I enroll.  To your knowledge, are a lot of CS courses still curved nowadays, and do you think curving grades is effective and/or fair?

bob
Monday, March 29, 2004

Most of the grad courses I've taken the grades have not been curved.  But almost all of the undergrad courses they were.

I think curves are handy when you have professors that have widely varying teaching styles or work requirements.  In one of the Physics courses I took, the high grades for the quarter were around 50%, the low was about 7%.  They curved it, and appropriately so in my opinion.

Sometimes professors can't judge adequately enough to create an object point scale to grade A work, so curves help them be more fair.

In grad courses where there is less people, I don't think it is fair because if they do it with a strict bell curve, someone is going to fail even though they might have done "A" work.

Andrew Hurst
Monday, March 29, 2004

My CS courses weren't curved and I thought that worked out the best.  I thought it gave me a realistic result in terms of my academic progress.

At the time, I was pursuing a business minor.  Now the Business Department curved the results for all its courses.  One accounting course I took was comprised of three exams only.  They were weighted equally.  A passing grade was 60% which equated to a C-.

Here's where things get weird.  I got 36%, 38% and 41% on each of the three exams.  One would think I bombed the course, right? 

Nope, good enough for a B-.

That doesn't seem very realistic to me.

C
Monday, March 29, 2004

One of my professors once told us that even if we think there's no curve, there always is.  It may not always be made manifest in the same ways.

The typical curve involves taking the scores into the bell curve, and using standard deviations to say who gets what grade.  While still used, most people see the apparent flaws and it gets railed a lot.

However, if an instructor sees that the current scores thus far in the class are overly high, he may give a test that is unnecessarily hard to broaden the distribution.  Similarly, he can give an easy test if it looks like too many people are doing badly.  This equates to an invisible curve.

I was a TA for a hard junior-level class, and I saw people who wouldn't even start lab assignments without wanting me there.  They couldn't write "Hello World" to save their lives.  And yet they passed the course because of such things.

It's not fair.  But fairness is rarely the case, and it's not necessarily the best thing.  If you feel the grading process is underestimating your abilities and knowledge, I found most professors to at least entertain your arguments, and some take it into account.

Walt
Monday, March 29, 2004

I think Walt hit on the real problem:

Bell curves can protect the clueless and worthless.  It's probably not the curve itself you have a problem with; it's some subset of your nine classmates that deserve some grade significantly lower than the one they will receive.

Or maybe I'm projecting, and this isn't what you mean at all.

pds
Monday, March 29, 2004

I took a graduate level computer architecture course where 25% of the final questions were on topics not covered in the assigned readings, lecture notes, or supplemental suggested readings.  Luckily it was graded on a curve.

Had the grading been a straight 90, 80, 70, 60 then the average score would have been a low C or high D.  The lone standout was a student that got an 85%.  He had worked 5+ years in one of Intel's design groups.

Professors can always devise tests tougher than the students can handle, so grading on a curve doesn't necessarily mean that grades are being dumbed down.

anon
Monday, March 29, 2004

Oh, I should add that all my other graduate courses were not graded on a curve.

anon
Monday, March 29, 2004

Curving is all relative, but then again, so is the grading system.  If a firm grading policy is established at the beginning of the semester, that's one thing, but this often is not the case.  Any grade you expect to get when said policy is not in place is just speculation.

Let me clarify, who ever said that 90%+ is an A?  It typically is, but why?  It's just semantics.  The 90% mark is just as arbitray as saying only the top 10% of the class gets an A.  What really matters is ones understanding of the matterial in question.  There is a set of criteria that are laid on the table, and students are supposed to demonstrate their effective understanding of the matterial.  Any percentage or points mark is simply an effort to quantify the subjective qualifier of one's own understanding. 

Some professors may teach a broad range of matterial only ever expecting a narrow understanding of the matterial.  So yes, you may only understand 30% of the matterial, but the professor may have only expected an average of 25% understanding.  On the other hand, a professor may expect 100% understanding.  Thus 80% understanding is hardly demonstration of knowledge gained.  It's all relative.  The system isn't sound, and always gave me high levels of stress.  In the end though, I found that the professor often assigned fair grades.  Maybe I was hoping that I did better, but the grade was usually fair.

Elephant
Monday, March 29, 2004

Grades are useless anyway. They aren't real.
Curve them, plot them, bell them, they don't
mean a thing. You might as well care how
much you got from the tooth fairy on your
very first lost tooth.

son of parnas
Monday, March 29, 2004

Back when I took them, we did exams in cuniform on clay tablets and if the prof. waited more than 2 hours, the clay set up and became hard to curve.

old_timer
Monday, March 29, 2004

I finished my CS degree a couple of years ago. Most grades were not based on rigid criteria which was determined by the CS department consensus and strictly maintained over the decades. If professors did not have a calculated curve, they biased grades to offset the general performance of the class. If you’re in a class too slow, you’ll get through little material beyond what’s in the textbook. If you’re in a class too fast, lectures will be preoccupied with obscure questions by students trying to show off everything they know. You just have to hope that you have a good professor who can read and balance the class.

As a student, it would be great to have an indication as to how you are doing in your educational journey. Unfortunately, that’s just not going to happen. Like many benchmarks in the IT world, earning a 4.0 gpa or having xyz certification only passes mustard with management and HR folks. Now there are sites out there like www.pickaprof.com which disclose grading information about classes with student feedback. So, a stellar gpa might just mean that you did some research and avoided all of the difficult classes.

jfm424
Monday, March 29, 2004

Although I didn't have the best GPA, and it didn't bother me, I know that many students are afraid of having low GPAs because they think it will hurt them in the job market.

In my experience, your GPA may influence the number of first interviews you get.  As interviewing continues, it becomes less of a factor.

My belief is that once you get your first job in the real world, your GPA doesn't mean squat, unless your employer wants some pride factor.

Walt
Monday, March 29, 2004

I've always wondered if you can really consider the distribution of intelligence in a university setting to be a proper bell curve anyway.

The problem is that any attempt to properly normalize, standardize, etc. intelligence, change in intelligence due to a class, etc. is pretty much going to suck in at least a few major ways.

Flamebait Sr.
Monday, March 29, 2004

The professor for my capstone class said, on the first day of class:

I don't know what the difference is between an A and an A-, or an A- and a B+.  So knowing that I can't give you a fair grade on each assignment, I'll just give you comments, lots of them.  A number would be meaningless, although if you want I'll give you one.  At the end of the semester I'll give you a grade based on all of your work.  If you're prepared and "pass" all of your pop-quizzes and don't do so well on my tests, you'll probably get a B.  If you do well on all of the tests but aren't prepared for the quizzes you will fail, because life is about being prepared every day.  Oh, and there's a final project - 250+ pages of documents and analysis - it's a big group project and you're being graded on more than just the paper you hand in.  You're always being graded, and everyone's scores will contribute differently to their total score.

So basically, he created a grade for each student based on careful observation of work and preparation throughout the semester.  There was no (this quiz is worth 10% of your grade) from him.  It was a bit scary not knowing where you stood (though if you just asked him he would tell you "You're doing just fine" or "You need to really step it up").  That was an interesting way to grade the class, and probably the best one I've seen - though it clearly took a lot of effort on his part.

Lou
Monday, March 29, 2004

A friend use to teach entry level programming classes in the night program of a local university.  At that level he observed that many students would start out doing very poorly - they just didn't get it - but at some point in the semester their work would suddenly improve as things suddenly started making sense.  So he instituted a policy, which he told every class on the first day: if your work at the end of the semester showed that you knew the material, poor grades from early in the semester would be ignored.

Which, for entry level programming, made a lot of sense to me.

RH
Monday, March 29, 2004

I'll add my mostly uninformed opinion to the mix - even _if_ you scored well, all this means is that you scored well on the test that the professor decided to give.  It has very little bearing on your ability to apply what you've learned to real-world problems.

Case in point: calculus.  I scraped by my calculus courses in the final years of University, and I know for a fact that if somebody presented me with a real-world problem that required calculus, I'd be dead in the water.

Would the problem here be that all courses are taught in isolation from each other?  It'd be interesting if there were an uber-course which presented problems and challenges which required the synthesis of the knowledge you (supposedly) gained from all of the other courses you took.

Maybe I'm just lazy, but by the time I reached the last year of my degree, all I cared about was knowing what was going to be on the exams, and learning that and only that.  Memorize, regurgitate.

Yeah, ok.  I'm lazy.

a
Monday, March 29, 2004

To answer Bob's original comments, it depends.  My wife's undergrad CS classes were all graded based on the policy given at the beginning of classes.  Mine (at a different University) were all "curved" in some form or another.  It all depends on the school and/or professor.

Since this is one of my hot buttons, I will continue.

As a couple people have mentioned, a good teacher will give appropriate grades and there can be many ways of doing this, including giving tests that result in a 90 = A, 80 = B, etc. breakdown, or in knowing how tough a test/project is and grading appropriately, or just in knowing the students through things like class participation.

A bad teacher (bad at evaluating, not neccesarily teaching) who tries these techniques will misjudge the difficulty of a test, or aptitude of students and, in general, will give students non reflective grades.  Everyone knows of a professor who is an easy A/B and another who no-one gets an A with.

A bell curve (give top A% and A, next B% a B, etc.) has some serious flaws (like getting stuck in a class with super-geniuses), but will result in the majority of the students getting the grades they deserve, independent of the teachers ability to evaluate, assuming the percentages are set properly.

Anyway, all this just frustrates the idealist in me.  And don't even get me started on grade inflation...

madking
Monday, March 29, 2004

The main problem I see is that university passing grades are set in stone irrrespeictive of the subject.

How the hell do you get 90% + in an English compostion class? Yet if there is a rigid grade system the top student in English will be discriminated agaiinst compared to the top student in a subject where it is feasible to get 90%.

When they do the English GCSEs and A Levels they have a final meeting when they decide what percentage should be given as the pass mark, and what as the grade A and so on. The decison is based on a seat of the pants decision -" are the students exams better than last years/" "This person got 65%? Is her exam wiorth an A?" Much more honest in my opinion than so-called scientific methods.

Stephen Jones
Tuesday, March 30, 2004

I've just finished a Software Engineering masters (anyone hiring in San Francisco?) and I don't think any of the classes were graded on a curve.  The good instructors looked at the body of a student's work over a semester, the bad instructors just gave a numerical grade, but I don't think anyone, ever, said "this percentage of the class will get A's."

The best instructors said (at the beginning of class) that 10-20% of the grade would be class participation and that folks who added to the discussion would be rewarded and that people who sat like lumps would not be getting A's, period.  I always felt that that gave the instructor enough leeway to give students the grades that seemed appropriate to their knowledge, not just to the "points" they earned.

Boofus McGoofus
Tuesday, March 30, 2004

I honestly don't think that universities have much to offer.

888

anon
Monday, April 05, 2004

*  Recent Topics

*  Fog Creek Home