> I’m imagining “students chosen randomly from the applicant pool, and students chosen based on merit will have indistinguishable academic results”. This isn’t

> I’m imagining “students chosen randomly from the applicant pool, and students chosen based on merit will have indistinguishable academic results”. This isn’t perfect yet but it’s closer to being testable. That's not the point though. Remember, the article proposes to draw a random sample from all applicants who are "good enough", so you would have some sort of aptitude test. However, instead of selecting the best N results from that test, you select a random N. I wouldn't hypothesize that the results are going to indistinguishable or even that they will be better. I would argue that total welfare may improve as a result and that is worth testing. The reason is as follows: If you only take the "best and brightest", there's little room for improvement.