The Paw Print

By Annya Kong, Staff Writer

The prevalence of artificial intelligence has seeped into almost every corner of every industry, and the college admissions process is no different. According to a recent survey by Intelligent, “For 2024, 80% of officials in higher education said they would integrate AI into their review process this year.” A majority of the 300-plus survey respondents whose schools used AI said that it had the final say on student admittance.

Algorithmic technology has long been used to review transcripts, recommendation letters, and personal statements. Automatic academic record systems like the one used at Rutgers University create academic profiles for students based on quantifiable data like grades and SAT or ACT scores. For prospective graduate students, the Graduate Management Admissions Test, or GMAT, has been used for years to evaluate the writing level of an essay (grammatical correctness, use of vocabulary, etc.).

Implementing machine learning and generative AI into college admissions could further streamline and refine the process going forward, potentially surpassing human admissions officers when it comes to making informed, consistent decisions on which students qualify as a good fit.

Of course, this depends on how officials are implementing AI. AI trained on human decisions will learn to replicate human results, only faster and more consistently. In this case, admissions officials will have to be wary of the potential concentration of human bias; if historical data on a certain college’s admissions shows a preference for wealthy white applicants, for example, AI will mimic that tendency.

On the other hand, if the AI model is based on data of admitted students’ later performance, we might get very accurate admissions in a few years. Research in this area is ongoing. At the University of Miami, AI is being used to compare applications from the 2022 admissions cycle, exploring whether or not certain indicators in a student’s essay – such as keywords or common themes – correspond to their likelihood of persisting in college. The university will be piloting their AI application for use this fall.

In either case, colleges should be transparent about their use of AI in admissions, so that they can be held accountable for the ethical concerns that arise.

Generative AI is changing the landscape on the application side of admissions as well. According to research by foundry10, roughly one in three high school students who applied to college last spring used an AI tool for aid in writing their essays. Around six percent of students had an AI write their final draft instead of doing most of the writing themselves.

AI detection software can be – cyclically, unsteadily – employed to combat this kind of cheating (although what and what doesn’t constitute cheating can be its own struggle, for students and schools alike). But past the cheat-and-detection “arms race” (aptly phrased by one Atlantic article), AI development in admissions is great news for colleges. Switching to an AI-heavy admissions process can greatly expedite the admissions process, which, according to Professor Diane Gayeski in an article from US News, is already “very algorithmic” when done by humans. On top of that, if AI is able to outperform human AOs, the rate of reward – the proportion of accepted students who go on to academically succeed at that college – will also improve.

Will students receive the same level of benefit?

AI can help students in the college application process by summarizing information on the extracurriculars, degree programs, and other aspects of a school to determine good fit. This is what a majority of students from the foundry10 study used AI for. Note that it’s still recommended to cross-check AI-generated results with college websites.

Other than that, naturally, students are discouraged from using AI in their applications in any heavy-handed way. Colleges push for authenticity and uniqueness, and AI-written essays showcase neither. Getting into college by cheating (AI means or otherwise) is, by most standards of philosophical good, a detriment.

However, this only raises the question of whether uniqueness will continue to be a major factor at all in AI-heavy admissions. A distinctive personality is important to human admissions officers who, amongst piles of applications to manually review, naturally seek students who stand out from the rest. This needn’t be the case for an AI trained only for the purpose of accurately forecasting potential – especially when even simple algorithms suffice (as in, outperform humans). An AI might look at just a few key requisites to determine if a student is a good fit for the school. At the University of Pennsylvania and University of Colorado at Boulder, a research group modified the Facebook language model RoBERTa into an AI tool that evaluates student admission essays across just seven variables. Students who scored positively on the ‘leadership’ trait were more likely to graduate in six years than those who did not. These kinds of models, based on relatively few self-contained criteria, disincentivize – or at least, do not encourage – uniqueness, as students gunning for a certain college need only match the criteria of their AI, regardless of individual depth.

Most likely, however, human admissions officers will still continue to play a role in combating oversight, spot checking for individuality, character and any other human essence deemed necessary for a compelling application. After all, an AI that can forecast academic performance is one thing. Reliably predicting long-term real-world success from individual to individual is an impossible other, a probability in which the opinions of other people play an enormous role.

Still, if AI takes up more and more of the admissions process, application strategy may turn into a guessing game over the preferences of a school’s AI. College counselor Rob Lamb, in an article from Inside Higher Ed, expressed concern that keyword searches and other AI scanning tools could cheapen the college application process, prompting students and counselors to focus on “What are this year’s keywords?” or “niche extracurriculars” and giving wealthy applicants with dedicated counselors even more of an advantage.

In turn, colleges would likely get exceedingly cagey about those preferences, a tension that could conflict with the aforementioned need for transparency.

Ultimately, however, nothing fundamental changes in college admissions with the introduction of AI. The process was always gamey, cheating has long been an inevitable likelihood and any issues of ethics, inequity or bias are only exacerbations of pre-existing fault lines. If uniqueness diffuses out of admissions then it was never a core concern to begin with. The potential efficacy of AI in replacement of human admissions officers will be a nod to a reality where ‘holistic admissions’ is, or perhaps ought to be, a (somewhat obscure) test for a few desirable qualities rather than an understanding of a student’s full individual character, as is often claimed. Students may feel angst to have to advertise their life to a computer, or relief to perceive a decrease in the prevalence of subjectivity and luck found in today’s college admissions process. These are all major what-ifs to look out for in the coming years. When the dust of the AI boom settles, students and colleges alike will find out if anything truly changed.