READ: 4 MIN
Sometimes lost in all of the talk about the huge increases in applications at top colleges, is just how the colleges are operationally handling the influx. Last year, Columbia University had over 60,000 applications for the class of 2025, up from just over 40,000 the year prior, an increase of 50%. So how did Columbia handle the extra 20,000 applications? Did they hire 50% more admissions officers? Ask their team to put in 50% more hours? It’s actually hard to know for certain, but it’s likely a combination of these two strategies. But they also likely got more “efficient” with how they reviewed applications, which is a way of saying that they spent less time on each application than they have in previous years. Colleges have not been keen to admit to this, but how else to handle the volume, which we predict will go up again this year?
How did Columbia handle an extra 20,000 applications last year?
In the old days, colleges would assign piles of applications to single readers who spent the winter months holed-up at home or in on-campus offices, reviewing an application every 20 minutes or so. Then each application would get a second read from another admissions officer. Some schools still review applications this way, but others, especially those with heavy volume like Emory, among others, have moved to a paired reading process. In this new process, a pair of readers review an application together, often splitting focus (one person on grades and scores, the other on essays) and then coming together to make a decision. Others, like Penn, have deconstructed the process into even finer parts. Time spent? About 7-8 minutes per application, and sometimes much less. This doesn’t feel great to students who have spent months putting together their applications, only to get a decision in minutes. But it’s simply not operationally possible to do the old method for many top schools. Even Yale, which touts that every decision is made in committee, gives most decisions a quick yes/no in committee based on a sorting that prioritizes discussion for only some applicants.
Which leads us to the topic of Artificial Intelligence (AI) in college admissions. This was the subject of a recent article in Diverse Issues in Higher Education. Author Rebecca Kelliher approaches the impending use of big data from a diversity standpoint - will the data reinforce the biases inherent in the current process, or will it eliminate them from consideration? Her finding is that it could go either way depending on how colleges use the data that they have. Will colleges focus in on zip codes where students can pay full-price, or will it seek out students in far-flung locations that may have the ability but not the means? How colleges will ultimately use AI is not yet known, but we can say with certainty that we believe it is coming. We know because this kind of work is already being done in the enrollment management departments at colleges where they use larger data sets obtained from the College Board or Naviance to target their marketing dollars. Even the top colleges - the ones that get more applications then they can reasonably handle - pay third-party companies to use specific targeting criteria for finding students that fit their needs. They are already using similar models to predict financial aid and determine merit aid awards that maximize yield while minimizing award dollars.
These same kind of targeting and predictive capabilities could be applied to large data sets of incoming applications. Rather than approach each application as a stand-alone entity worthy of a complete read (even if the only read is in a few minutes by a paired admissions team), they could be sorted by quality against some criteria. AI could be applied to the data set, for example, to pull out all students with a certain course rigor score (a notoriously difficult factor to standardize given the varying curriculums at high schools). This list of “high rigor” kids could then be sorted by GPA and given priority reading status. We can picture a process whereby complex algorithms that characterize students is used to build classes that are statistically diverse. Colleges say they don’t have quotas by demographic (although the statistics challenge that stance). But as AI evolves, colleges with large enough data sets could indeed target a diversity mix that better reflects their institutional priorities. AI could also help colleges use their human readers more judiciously, giving them more time to review applications that meet a minimum set of criteria, while auto-dening the mass of undifferentiated applicants. Colleges might say this is going too far, but if applications continue to climb in the face of test-optional admissions, and Columbia eventually ends up with 100,000 applications or even 200,000 ten years from now, how else would they manage?
