Grading on a curve is not without problems, particularly when class sizes are small, but it has a lot of benefits. It makes it easier to compare students across schools when the schools use the same curve.
Why should someone else's ability influence my grade???
Grades only make sense to compare within the same class at the same university during the same semester.
The grading in my course is not relative. I set a rubric that I assess each individual on. Sure, within the class you can compare grades, but I wouldn't say it is particularly meaningful. Comparing between classes or universities is completely meaningless.
People just want a simple number that can accurately represent someone's knowledge/learning/skill/effort, but it doesn't exist.
It's fairly easy to boost your GPA by mixing in easy courses
For starters, the quality of students is not uniform across universities. Middle of the pack at Harvard is a very different thing from middle of the pack at a party school. To get a useful ordering, you'd need to give every CS graduate the same standardized test, like the bar exam.
With few exceptions there is no mastering the material, or at least there shouldn't be. Not in college. There's always more complex material you can use to differentiate students. When you grade on a curve the test should be difficult enough that even the best students will typically get at least one question wrong, with most students falling along a nice bell distribution. (I say most because it's probably better to err on the side of a handful of students clustering at the top rather than at the bottom, especially if the bottom means flunking out.)
This can be more difficult for smaller, seminar style classes. One solution is giving the professor leeway to shift up and tighten the curve so he's not forced to give Fs or Ds. These classes are usually in the latter years of a program, particularly in undergraduate, where it's less important to winnow students out and ensure challenging curriculums.
That latter point is important. If you don't enforce some sort of statistical distribution how can you gauge the quality of your curriculum? If everybody is getting As is that because all your students are smart and disciplined, or because the material is too simple? If you require a bell distribution, then the curriculum will by necessity be a good fit for your cohort of students. In this way a university can maintain a quality curriculum without having to resort to external metrics or comparisons with other schools.
Instead the differentiation was between departments and courses. It was well understood that X majors were Y majors who couldn't hack it, and that within a major, certain classes were for high-achieving masochists while others were schedule padding. (This was crucial information, not just dick-measuring. You could easily put yourself in crisis by taking a class beyond your abilities, or multiple hard classes at once. Friends helped each other avoid such nightmares). If you asked me to evaluate a classmate's transcript, I wouldn't even look at their grades, only at the classes they allowed themselves to have grades in.
No one has ever asked me that, or to see my transcript.
If a class is filled with C and D students (objectively), the curve will spread them out and those C's become As. If a different class is filled with A and B students, they are spread out and those B's can fail or become C's. Now the first group's curve A's are not as good as the second group's A's, and are, in fact, worse than the second group's F's or C's.
I often set the curve in my classes. I was offered money to not do my best on exams because of how it affected other's grades. My ability or lack there of should not affect another's grade which should reflect their demonstration of subject mastery.
Perhaps methods to decouple evaluation scores from student grades should be better explored without antagonizing anyone? Perhaps scores after the first week should be compared to scores mid-way and then last?
Maybe all of that temporal basis is flawed too and we need to try crazier things. A weirder option could be to randomize the evaluations given back to teachers and to have the teachers select which ones seem to be theirs and some consideration is given to professors that are accurate for themselves. Basically it’d be a test of whether a teacher could identify their own teaching methods’ strengths and faults in contrast with other teachers’ and encourage diversity in styles among faculty. If a teacher wants to grade hard, I think they should be free to do so if it’s shown to be in students’ best interests. This is vaguely similar to how classifiers in machine learning training works but with a really different goal in mind.
If 45% of students are failing, then either the class needs to be changed, or more effort needs to be made to prevent students from taking a class that they are likely to fail.
...I don't really accept that.
The problem with grade inflation isn't just that it becomes more difficult to judge student competency; similarly, it becomes more difficult to judge curriculum quality. If everybody is getting As when the curriculum is challenging, they'll continue getting As if it degenerates. Enforcing a distribution helps you to maintain a stable relationship between the abilities of your student body and the quality of the curriculum.
In my experience, they aren't, for a number of reasons.
People trying to judge students based on grades would then have to have a rough sense of how the institution works and how a grade should be interpreted.
But this is what they have to do anyway, curve or not. For example, employers already know that a B- at an Ivy League institution is a somewhat poor grade. (At Harvard, 75% of all grades given are B+ or above.)
Admittedly I think the Ivy League system (which effectively is grading on a curve, just a very narrow one with the mean set to an A-) is pretty good. There's certainly little incentive for students to want easy classes, because in terms of grading there is basically no difference between an easy class and a hard class. It's as close as you can come to just not giving grades at all, but doesn't mark you out as a weird hippy school.