This is a complex issue. Harvard grades everyone on a curve, and basically never gives lower than a B. They use "A+" as a system to identify the truly exceptional people.
My point is that a professor should be able to write 20 questions that are relevant and interesting, and just look at the distribution of results to assign grades. I think that makes sense, rather than turning university exams into certification style exams.
My answer to why, it is a horrible idea to not grade on a curve requires and anecdote:
I took an impossibly hard math course with about 15 students where the average grade was a 2, I got a 4, one guy got a 6, and another guy got an 8. The professor gave me an A, but in reality, I was 4 orders of magnitude dumber than the best student. Professors have discretion. The main reason I tell this story, is that the guys that got a 6/10 and and 8/10 have both gone on to have an 8-figure net-worth (fuck-you money). One through finance, and another through a startup. It has been awesome to see the people that I know are insanely smart become insanely wealthy.
The important part of this anecdote is that the professors that wrote the exam wrote two questions that one of smartest humans on earth could not answer. How fucked up would the world be if the professor had only written the 4 questions we all got it right? Asking impossible questions creates greatness. Grading on a curve enables professors to find truly exceptional people.
What you've written here doesn't make much sense; possibly because the reason you wrote it wasn't to support your argument but to tell us you went to school with some people who became very rich.
"How fucked up would the world be if the professor had only written the 4 questions we all got it right?"
This just doesn't make sense. What are you trying to say? Did you mean to ask "What happens if there were only four questions and we all got it right?"
"My point is that a professor should be able to write 20 questions that are relevant and interesting, and just look at the distribution of results to assign grades."
Fine. My point is that this means you're being marked on the basis of how well you did compared to everyone else who took the exam, which means it's not a measure of how good you are; it's a measure of how good you are compared to the others taking the exam. If the purpose of taking the exam is to demonstrate you know the material, grading on a curve subverts and corrupts that purpose. Any exam in which your mark depends on how well other people do in the exam is a nonsense; your mark should only depend on how well you do in the exam. If the professor knows the exam is horrifically difficult and thinks that to score twenty percent is amazing, you should get a good mark for reaching twenty percent, even if everyone else scores fifty percent.
Grading on a curve helps normalize the resulting letter grades based on how well everyone else in the class understand the material, as it was presented. You are correct in this.
I believe it acts as a check to make sure that the class is well taught, and the questions of an appropriate difficulty level.
If every student receives an F for a 40% correct (and they all only get 40% correct), a curve normalizes that. Students who understand the material as it was presented and tested better than others receive a higher grade.
Makes good sense to me. The problems arise in the Harvard situation, where nobody is curved down. If you down curve inflated scores downward, then a curve only serves to inflate grades and is less useful.
"If every student receives an F for a 40% correct (and they all only get 40% correct), a curve normalizes that. Students who understand the material as it was presented and tested better than others receive a higher grade."
If every one in the class was lazy to study/prepare for the exam and every one ended up getting less than 40% correct, then they all deserve to get F. Normalizing with a curve means, it's possible that someone who got only 39% in that exam in a class full of lazy folks who all scored much less than him, can earn a A or A+ ! Compare this to similar class/exam taken by different set of students (all brilliant) at a different time or place, who all worked hard and scored above 80% but still some of them could get B or C's due to curve grading. Wouldn't it be unfair to the 80% scorer who got a C whereas some one who scored a 39% on the same subject got A+ due to circumstances (time and peers he took the exam with) !! How is that fair at all?
This is an invented objection. In reality, in classes of sufficient size, there is always a subset of students who work hard (or are extra smart, or whatever), a subset who is average, and a subset who slacks, or just doesn't understand the material well. The objection that there might be a mass class conspiracy to all score 40%, or that everyone in the class will be lazy, just doesn't make sense.
Who says it's a conspiracy? It's more a matter of culture. I have seen it happen: there's a trend in the class where the homework grades, on average, are high, but the exam grades are low, even though the material is mostly the same. This happens at reputable schools and at "low tier" schools.
It's quite difficult to write new tests that are unique enough not to be trivialized by access to old exams but of consistent difficulty semester after semester. Even if you could it might not give the result you want: if your university moves upmarket or the high school preparation in your subject improves should the average grade in your class increase to reflect these exogenous factors?
It's hard enough writing a good exam -- maximally descriminating, useful pedagogically, not unintentionally ambiguous -- without also requiring it to be mechanistically translatable to a semester grade.
Rampant grade inflation at Harvard and other Ivy leagues does not make grading on a curve a good idea. What does grading on a curve even mean if everyone is getting B's and higher anyway? If the class cannot grasp the material at hand, they should be graded poorly. If they learn it well, they should be rewarded. It is perfectly fine to write difficult (even near impossible) questions on an exam, with or without a curve. The only classes where I saw a grading curve used effectively were the weeder classes at <my well known research university> for high-demand majors. Like ochem for pre-med and electrical engineering prereqs, there were not enough spots in the program for everyone, and a certain percentage would fail the class, therefore assuring that only the strongest students would be admitted to those high priority majors.
We have a fetish for measuring and ordering people. Imagine what would happen if that professor couldn't "find" that one person who scored an 8. We would never be able to find brilliant thinkers.
My point is that a professor should be able to write 20 questions that are relevant and interesting, and just look at the distribution of results to assign grades. I think that makes sense, rather than turning university exams into certification style exams.
My answer to why, it is a horrible idea to not grade on a curve requires and anecdote:
I took an impossibly hard math course with about 15 students where the average grade was a 2, I got a 4, one guy got a 6, and another guy got an 8. The professor gave me an A, but in reality, I was 4 orders of magnitude dumber than the best student. Professors have discretion. The main reason I tell this story, is that the guys that got a 6/10 and and 8/10 have both gone on to have an 8-figure net-worth (fuck-you money). One through finance, and another through a startup. It has been awesome to see the people that I know are insanely smart become insanely wealthy.
The important part of this anecdote is that the professors that wrote the exam wrote two questions that one of smartest humans on earth could not answer. How fucked up would the world be if the professor had only written the 4 questions we all got it right? Asking impossible questions creates greatness. Grading on a curve enables professors to find truly exceptional people.