At least where the acceptance rate is 3-5 percent, it seems plausible that there could be something like the “AI Safety Common Pre-Application” that would reduce the time burden for many applicants. In many cases it would seem possible to say, on information not customized to a specific program, that an applicant just isn’t going to make that top 3-5%.
(Applicants meeting specified criteria would presumably be invited to skip the pre-app stage, eliminating the risk of those applicants being erroneously screened out on common information.)
By analogy: In some courts, you have to seek permission from the court of appeals prior to appealing. The bar for being allowed is much lower than for succeeding, which means that denials at permission stage save disappointed litigants the resources they’d otherwise use to prepare full appeals.
This is something that BERI has actually been discussing! If anyone is interested in talking to us more about this (either from a hiring/evaluation perspective or from a candidate perspective, you can send an email to contact@existence.org and I’ll follow up.
I would be excited about a common application. My sense is that the only reason it doesn’t exist is that no one has put the time in to create it; when I’ve talked to hiring managers, most were in favor of the project (though there are some concerns, e.g. the fact that applications are currently a costly signal is helpful for identifying the applicants who actually really want to apply).
The more arbitrary college admissions criteria become, the more the students at elite universities will simply be those who were most determined to get in.
I think “actually really want to apply” is not enough of a correlation to base decisions on. The fact is that even qualified+motivated applicants would need to apply to a dozen+ places, and often EA application questions require a lot of thought anyway.
To give an example, lots of EAs are from top unis, and I’m pretty sure the meta-strategy for applying to selective unis is to not fall in love with any particular one and shotgun a lot of applications. This reduces the role of personal fit.
One thing I always thought was interesting would be to have an actual application limit like UK colleges do. I applied to 20 colleges, and I would’ve been fine applying to 5 with improved odds.
Many elite US universities do this—the Common Application to which my comment indirectly alludes—and law schools did something vaguely similar at least in the mid-2000s (showing my age here). So I am expecting the answer is negative.
Common evaluation would be trickier—e.g., I vaguely remember some universities getting into trouble with allegations that they were divvying up choice applicants rather than competing for them. [Edit: It may have been this—they were apparently colluding about financial aid offers, and reached a settlement with DOJ Antitrust to stop doing this.]
I vaguely remember some universities getting into trouble with allegations that they were divvying up choice applicants rather than competing for them.
This is explicitly the policy in the UK, and (I would guess) almost entirely eliminates offer acceptance uncertainty for Oxford and the other place:
… you can’t apply to Oxford and Cambridge in the same year.
(Disclaimer: I’m not an antitrust lawyer, not that I can give anyone legal advice on the Forum anyway. Also, this is a US perspective.)
The basic principle is that agreements “in restraint of trade” are illegal, with that term interpreted by reference to a “rule of reason” developed through over a century of caselaw. Neither student admissions nor employee hiring are really in the heartland of antitrust, although it has been applied to both in the past.
I don’t see how admissions and hiring are that different when it comes to determining whether use of a common application form would be in restraint of trade (i.e., whether it unreasonably impedes fair competition). I’m also unclear on what a good argument would be for an assertion that using the same first-stage application would unreasonably impede fair competition for employees in the first place. I’d argue that it would promote competition in the market for employees, by making it easier for employees to apply to more potential employers. But I didn’t dig into any caselaw on that.
I’d also worry, though, about the application having only certain kinds of questions which do not bring out the best in (some/many) people. I’ve definitely seen some applications where I thought I wasn’t given the chance to show my worth, and others where I was. This app would have to be drafted with a lot of care.
Likely a reason this hasn’t been done yet (and why we aren’t saying 100% we’ll do this)--it’s very difficult to create a “good” application and process, nonetheless create one that multiple orgs would all find valuable and be able to fit into their workflows. I’m happy to listen to any feedback/input you may have on applications that you thought allowed you to present yourself best.
Yes, I think we would benefit from having organizations run the pre-app and standard app in parallel for one cycle (while compensating applicants for the additional work on the margin!). We’d be looking for a pre-app “score” for each organization at which very few people whose application would have survived the first round of the old process would be eliminated by the pre-app and/or ~no one who was ultimately accepted was screened out.
At least where the acceptance rate is 3-5 percent, it seems plausible that there could be something like the “AI Safety Common Pre-Application” that would reduce the time burden for many applicants. In many cases it would seem possible to say, on information not customized to a specific program, that an applicant just isn’t going to make that top 3-5%.
(Applicants meeting specified criteria would presumably be invited to skip the pre-app stage, eliminating the risk of those applicants being erroneously screened out on common information.)
By analogy: In some courts, you have to seek permission from the court of appeals prior to appealing. The bar for being allowed is much lower than for succeeding, which means that denials at permission stage save disappointed litigants the resources they’d otherwise use to prepare full appeals.
This is something that BERI has actually been discussing! If anyone is interested in talking to us more about this (either from a hiring/evaluation perspective or from a candidate perspective, you can send an email to contact@existence.org and I’ll follow up.
I would be excited about a common application. My sense is that the only reason it doesn’t exist is that no one has put the time in to create it; when I’ve talked to hiring managers, most were in favor of the project (though there are some concerns, e.g. the fact that applications are currently a costly signal is helpful for identifying the applicants who actually really want to apply).
Paul Graham had a nice take on this:
I think “actually really want to apply” is not enough of a correlation to base decisions on. The fact is that even qualified+motivated applicants would need to apply to a dozen+ places, and often EA application questions require a lot of thought anyway.
To give an example, lots of EAs are from top unis, and I’m pretty sure the meta-strategy for applying to selective unis is to not fall in love with any particular one and shotgun a lot of applications. This reduces the role of personal fit.
One thing I always thought was interesting would be to have an actual application limit like UK colleges do. I applied to 20 colleges, and I would’ve been fine applying to 5 with improved odds.
Are there antitrust concerns with multiple orgs (even if nonprofit) using a common screener?
Many elite US universities do this—the Common Application to which my comment indirectly alludes—and law schools did something vaguely similar at least in the mid-2000s (showing my age here). So I am expecting the answer is negative.
Common evaluation would be trickier—e.g., I vaguely remember some universities getting into trouble with allegations that they were divvying up choice applicants rather than competing for them. [Edit: It may have been this—they were apparently colluding about financial aid offers, and reached a settlement with DOJ Antitrust to stop doing this.]
This is explicitly the policy in the UK, and (I would guess) almost entirely eliminates offer acceptance uncertainty for Oxford and the other place:
Hm I don’t obviously see the analogy with the common app—hiring employees and admitting students seem quite different.
(Disclaimer: I’m not an antitrust lawyer, not that I can give anyone legal advice on the Forum anyway. Also, this is a US perspective.)
The basic principle is that agreements “in restraint of trade” are illegal, with that term interpreted by reference to a “rule of reason” developed through over a century of caselaw. Neither student admissions nor employee hiring are really in the heartland of antitrust, although it has been applied to both in the past.
I don’t see how admissions and hiring are that different when it comes to determining whether use of a common application form would be in restraint of trade (i.e., whether it unreasonably impedes fair competition). I’m also unclear on what a good argument would be for an assertion that using the same first-stage application would unreasonably impede fair competition for employees in the first place. I’d argue that it would promote competition in the market for employees, by making it easier for employees to apply to more potential employers. But I didn’t dig into any caselaw on that.
This would definitely reduce the time cost.
I’d also worry, though, about the application having only certain kinds of questions which do not bring out the best in (some/many) people. I’ve definitely seen some applications where I thought I wasn’t given the chance to show my worth, and others where I was. This app would have to be drafted with a lot of care.
Also thank you @elizabethcooper for taking initiative on this!
Likely a reason this hasn’t been done yet (and why we aren’t saying 100% we’ll do this)--it’s very difficult to create a “good” application and process, nonetheless create one that multiple orgs would all find valuable and be able to fit into their workflows. I’m happy to listen to any feedback/input you may have on applications that you thought allowed you to present yourself best.
Yes, I think we would benefit from having organizations run the pre-app and standard app in parallel for one cycle (while compensating applicants for the additional work on the margin!). We’d be looking for a pre-app “score” for each organization at which very few people whose application would have survived the first round of the old process would be eliminated by the pre-app and/or ~no one who was ultimately accepted was screened out.