I think the problem has more to do with continued application of filters based entirely on one's ranking according to some metric. The problem with this is that we never have a measureable metric for the goals we actually care about. For example, universities want the students who will succeed the most at university and thus use some combination of grades, test scores, extracurricular participation, etc. as a proxy for future success. However, future success is a vague goal and the measures chosen would often fail as predictors even if the objective were formalized due to a combination of chance and various biases of the predictor metrics. This gets disastrous in the long term because all these filter processes reinforce each other's biases and due to people being filtered trying to game the system. I think to fix some of these issues, we should stop trying to rank everything precisely and introduce some degree of actual randomness in who passes a given filter. For example, instead of passing the top 10% of applicants in the old system, still keep the top 5%, but do a weighted choice for sufficiently qualified people to fill the other slots. I think this would also help improve diversity as well without the use of true affirmative action.