> why should the grade distribution of an exam be determined beforehand?
because we know beforehand that population density over repeated trials of some measure will fall in a normal distribution, especially when constructing such a test is more straightforward than 90% gets A, 80% gets B, etc
So, you are suggesting that if we have a class of 100 people, and we want to measure their height, we should just order them by length, and then the tallest one we call 2.1 meters, the one in the middle 1.8, and the smallest one 1.5, no matter how tall they actually are?
Wouldn't it be better if we measure their actual size?
because we know beforehand that population density over repeated trials of some measure will fall in a normal distribution, especially when constructing such a test is more straightforward than 90% gets A, 80% gets B, etc