I had a professor in grad school who had studied how admissions inputs correlated to various measures of success (which is of course a fraught subject of uts own). This was a while back though I assume not much as changed. As I recall, standardized tests followed by some sort of adjusted class rank were the best predictors while a lot of the softer stuff like interviews and letters of recommendation were pretty close to worthless. I expect schools could drop some of the more time-consuming softer stuff with very little impact on the admissions process.

In MIT's case it probably is especially important to assess certain types of skills, because someone who doesn't have a reasonable math background for example is going to have a bad time. (Though problems doing math is a surprisingly common theme in higher ed including at the grad level--even with programs you wouldn't think were especially quantitative.)