Adding a bit of fuel to the fire. This week’s episode (Thu) of the YCBK podcast has Dartmouth’s vice provost talking about how to use AI for reading applications. Starts around 1:05. Check it out.
1 Like
Here is a prediction. Big Data/AI/Machine Learning approaches to normalizing academic qualifications are going to end up making it all the more clear how different high schools do or do not generate the individualized data necessary to discriminate among applicants plausibly within the top 10%, 5%, 1%, and so on.
And it will become clear APs get you so far, but not nearly all the way, in providing such information.
1 Like
Actually, the thought here is that AI will help more with crunching numbers. College Board introduced a SES score a while back and then backed off once there was blow back. Provost Coffin talks about how he has come across kids that are 1-2-3 in their class and get a 1/2 on AP tests and why that is a red flag. Thing is while colleges has enrollment goals, they want to make sure they are filling their classes with kids that can flourish once they are on campus. This will require more and more implementation of big data strategies.
1 Like
Right, but what I had in mind is basically this.
Big Data shows a 4.0 in the very advanced classes at Expensive Private HS has a predictive power as to academic success at Expensive Private College of X (where X is some sort of probabilistic measure).
Big Data shows a 4.0 in the less advanced classes at Underfunded Public HS has a predictive power as to academic success at Expensive Private College of Y, where Y << X.
Finally, Big Data shows a 4.0 in the less advanced classes at Underfunded Public HS plus 5s on a bunch of APs has a predictive power as to academic success at Expensive Private College of Z, where Y < Z < X.
Something like this is what I had in mind when I wrote, “it will become clear APs get you so far, but not nearly all the way, in providing such information.” Given these hypotheticals, adding the AP scores meant Z was better than Y. But I personally suspect AP scores do not including enough information to get Z all the way up to X.
And I am not saying it is wrong for Expensive Private College to prefer applicants with X over Z. But if true, this would mean the best a student could do at Underfunded Public HS would be less than the best a student could do at Expensive Private HS.
Which in fact is something I think these colleges implicitly already believe. But I am thinking Big Data/AI may ultimately make it very clear what is happening.
in discussion with a reader at a highly selective school, there was explicit confirmation that below a certain grade line, unhooked applicants were almost immediately eliminated, with readers essentially searching for something exceptional/world class in the application, failing which its headed to the bin.
Of course, AI can instantly sift through every application against that grade line for unhooked applicants, and assign that deck to a human to scavenge for exceptional traits
2 Likes
If a high school typically has A students in AP courses scoring 1 or 2 on the associated AP tests, that can be a red flag on the school.
2 Likes
I was just listening to an episode of Your College Bound Kid where they mentioned that although the number is still small, the % of not academically competitive applications increased (which is no surprise). They went on to mention that Yale has come up with “some way” to quickly identify those applications so AOs don’t have to waste precious time on them.
1 Like
Yale talks about their new process in their ‘Inside the Yale Admissions Office’ podcast, the Feb 3 episode titled ‘Reading Reloaded’.
Basically a senior AO is going to do a fast review of every app and apps that won’t be competitive won’t go on for full review.
2 Likes
Thanks. Off to listen! Are they basically looking at numbers at that point?
I think so, probably activities too. Maybe LoRs if on the fence? I’m not sure they went into those details!
At that point aren’t you basically doing a full read? I also watched the Live Application Review video Holy Cross posted a couple of weeks ago. While the secret sauce is probably different, they all use the same software. It was interesting to see what the Common App looks like on their end.
Something I learned: checking the test optional box gets rid of the score tab - it doesn’t matter if you put the scores down or not. But not the AP Scores. Those will show if you include them.
1 Like
It is good to get some confirmation of this.
There was always good reason to believe it had to be this way. Basically, the number of unique students with high numbers nationally had not increased (if anything it was going down), so the volume of unique applicants with such numbers likely could not have increased too much absent a massive increase in applications per such applicant. And the applications per such applicant went up maybe a little, but not that much. Indeed, off hand it seemed to be at most that was offsetting the decline in this population.
And then it was sorta obvious from the data in the recent CDS filings. Meaning if the increased volume of applications had been driven by “new” unique applicants with a distribution of numerical qualifications that made them at least as numerically-qualified as the “old” unique applicants, then you would have expected the enrolled numbers to go up more. Test optional complicates that, but between the top numbers not going up that much, and the fact in the last round of CDS filings a lot of the successful applicants at some of these colleges had submitted, it was still pretty obvious that the large increase in applications was not doing much to drive up the enrolled numbers.
But this statement confirms all that in the sense that if that inference that the new unique applicants to these colleges were skewing less numerically qualified than the old unique applicants was true, then the overall percentage of not-numerically-qualified applicants should be going up–which is apparently also true.
I note one of the implications of this is that if you as an individual have high numbers, the sort of numbers that would have been high back in the pre-COVID days, the increase in application volumes and corresponding decrease in admission rates may not really affect your admissions chances much at these sorts of colleges.
Maybe a bit, in the sense that in holistic review, there are probably going to be a few more extraordinary lower-numbers people that get admitted. But that is likely a small effect.
Because it is becoming pretty obvious that at these types of colleges, mostly what is going to happen is almost all of those new lower-number applications are going to be screened onto some sort of fast track from which very few return. And so what will be left after that process will mostly look pretty much like the old pool anyway. Again possibly with a few more truly extraordinary lower-number applications. But not many.
To fold in some of the discussion above . . . I think if it is a human reader doing it, it likely is mostly just the test scores (if submitted) and transcript. They may also use a combination of the school report, internal school data, and regional admissions officer experience to help normalize the transcript at least roughly. I would say this is implied, in fact, by saying a “senior” AO is going to do this at Yale–a senior AO is the kind of person who would plausibly have the experience necessary to at least roughly normalize transcripts from within their region quickly.
But this does seem like an area ripe for application of Big Data/machine-learning/AI approaches. Even if you don’t want to turn the whole process over to such a program, having it give some sort of analysis to that AO to consider would seem to make sense as long as they thought it was working pretty well.
And of course that sort of program could be using anything, in the sense that presumably once it is fully trained it could analyze an arbitrary number of applications extremely quickly no matter how much information it was using as inputs.
1 Like
Like I said, maybe on the LoRs. I’d have to listen to the podcast again for the details…I do think it’s mostly a stats review to throw out the apps that clearly won’t be competitive.
I was going by what YCBK said. He asked schools what % was qualified and he said it ranged from 65 - 90 (I think). 65 being Chicago and that they were the only ones who said that number did not go down. This makes sense bc UChicago has been TO since before the pandemic, which he later pointed out.
How did they define “not academically competitive”? It seems to be tied to TO, so do they mean students with poor grades/transcripts or good grades but low rigor? Or do they mean TO kids with strong grades/rigor but just no test scores?
It’s hard for me to believe that lots of kids are interpreting TO to mean they should apply to a highly selective school with poor grades/low rigor (in the context of their school), but is that what’s happening?
They did not define it. Yale thought they had to find a way to efficiently deal with such applications, so there must be enough of them to warrant a process, right?
At our (top BS) there are definitely kids applying to these schools who would never have bothered before. I think there are lots of kids with good gpas who don’t realize their schedule is not that rigorous.
1 Like
I suspect you are right. This is also precisely where things like school reports, counselor reports, internal tracking data, and AO experience could potentially really help. It is also something that a Big Data/AI-type approach could at least plausibly handle.
Thanks to you both. This makes sense.
This topic was automatically closed 180 days after the last reply. If you’d like to reply, please flag the thread for moderator attention.