I'm curious about what HN readers would suggest for a better approach to college admission.Maybe I'm biased because I code systems for Japanese universities for a living, but I've always really liked examinations.
We have a system we sell as consultingware to many Japanese colleges, installing the base of what is considered acceptable here, with customizations per request. By default, it provides:
1) No way to even tell if a student is legacy, plays soccer, etc
2) Strict auditing of everything. First text in the design document, in 48 point font: OUR OVERIDING PRIORITY IS TO NOT COMPROMISE THE PUBLIC'S TRUST IN THE OBJECTIVITY OF THE EXAMINATION.
3) Affirmative action? Not supported or encouraged (I won't say impossible, but its pretty close -- you'd have to use manual overrides of examination results after turning off the failsafes preventing you from doing so accidentally, which causes BIG RED ANNOUNCEMENTS to appear to the other people responsible for the system).
4) Some schools have particular needs for X number of Y students. (For example, a school with a historical association with a particular church/temple might need X priests per year). They can accomplish it in one very transparent way: say "We are looking for X priests. Please apply to the I Want To Be A Priest exam if you're interested", and then the top X wants-to-be-priest candidates are accepted. Note that, importantly, this has absolutely no effect on the general applicant pool, and is not an effective way to get out of the general admissions standards (since if you elect that sort of treatment, you WILL be tracked straight into the Becomes A Priest major and you will not really be allowed to change).
5) Transparency for everything. The school starts the year by saying "English Department seeking 12 students. History Department seeking 16...." Then, after the exam is over, all scores are published (anonymized) with aggregate statistics: the English Department filled 12 spots out of 127 applicants having an average score of 187 points. The minimum passing score was 214, and students scoring above 212 were added to the wait list, from which 2 were offered admission.
The stats are also broken down by test load: say you're told to take two subjects out of three, your pick. The University will publish the distributions of scores for every possible combination (3 in this trivial case), letting you know if any combination was harder or easier than others (which they do A LOT of work to avoid, incidentally).
And the icing on the cake of full transparency: the university maintains five to ten years of these stats and prints it right in the bloody admission handbook itself. Because that's how confident we are that the process is not corrupt.
I want American universities to be like that. Heck, fine by me if they want to put a thumb on the admissions scales. Just publish how big the thumb was:
"15 black students admitted with average SATs of..."