> But for STEM it is way more clear. Wrong answers on a math test are wrong. Software projects that don't compile should fail. How hard should the problem set

> But for STEM it is way more clear. Wrong answers on a math test are wrong. Software projects that don't compile should fail. How hard should the problem set be? How long should students have to complete it? As you increase the difficulty you will push down the score distribution. How do you propose to square the existence of a score distribution against the quite subjective level of "difficulty" of the work that you're setting for grading?