As high school seniors wrap up their college admission applications, they find themselves amidst a national discourse on the judicious, permissible, responsible, and traceable use of artificial intelligence (AI). After hitting the submit button on their applications, students inevitably ponder the inner workings of the admission office's enigmatic "black box."
This year, the mystery is more intricate than ever, with questions arising about whether humans or machines are determining their destiny. The use of artificial intelligence by admission offices is not a matter of speculation; enrollment managers have incorporated AI into their processes long before the surge in ChatGPT's popularity last fall. The real question lies in the extent and manner in which this technology will be employed.
What's the Talk?
While numerous articles discuss students utilizing AI to craft college essays, recent pieces have delved into how colleges will leverage tools like ChatGPT. Taylor Swaak, in The Chronicle of Higher Education, explored the use of AI in admission, pondering whether it serves as the remedy for "administrative drudgery" and how it extends into recruiting, predictive modeling, and application scanning and review. Swaak acknowledged caution among her sources but noted that they are indeed moving forward with AI integration.
Liam Knox, reporting for Inside Higher Ed, highlighted a survey conducted by Intelligent, an online education magazine targeting college applicants. The survey indicated that "Half of educational admissions departments currently use AI, and 82% will by 2024," with the majority of schools allowing AI to play a decisive role in applicant evaluations. The survey's credibility has faced skepticism due to its limited sample size of 399 respondents, with 20% not being college admission officers. Admission leaders, like Andy Borst of the University of Georgia, reassured applicants and counselors that AI isn't replacing the human touch in the admissions process entirely.
Engage in a conversation with any college admission leader, and a common refrain emerges: "It's a human process." The intent is to dispel the notion that admissions are solely number-driven and impersonal. It underscores the imperfection of the system, where humans make decisions in as thorough and comprehensive a manner as possible. Holistic admission involves reviewers considering factors beyond test scores and grades, encompassing involvement, leadership, background, context, essays, interviews, recommendations, and more. The decision-making process is inherently human, with educators determining what aspects of an application hold significance and to what extent. But is this sustainable?
Amidst a decline in overall college enrollment, some institutions experience a surge in application numbers. Between 2000 and 2020, the number of students applying to seven or more colleges increased by 184%, reaching 37%. Large universities like New York University and The University of California witnessed applications for fall 2023 admission surpassing 100,000.
Simultaneously, enrollment leaders are stretched thin and under immense stress. The "Great Resignation" has left admission offices, akin to other student affairs departments, severely understaffed. Record vacancies in staff have been reported, with some offices operating with limited capacity. The situation calls for innovation, and colleges are turning to part-time readers to handle the application volume, with some employing over 200 seasonal readers. The question arises: Is this the optimal approach, or could an artificial admission officer provide a more viable, cost-effective solution?
Predicting the Unpredictable
If given the choice, who would you prefer to evaluate your college application? An impartial reader with extensive training by a team of seasoned professionals, armed with a historical perspective on student success? Or a new application reader, trained for a few weeks, possibly not at their best on a rainy day after a family argument? Does the absence of human involvement in that first reader change your preference?
Consider research from The University of Pennsylvania's Uri Simonsohn, revealing the impact of weather on decision-making. Simonsohn's analysis of admission decisions showed that on cloudier days, academic attributes weighed more heavily, while sunnier days placed greater emphasis on non-academic attributes. Despite anti-bias training for admission readers, humans, it seems, are not immune to their emotional nature or even the weather.
RoBERTa: An Admission Superstar or Crossing Boundaries?
A recent study in the journal Science Advances suggests that the era of AI admission officers may not be far off. Researchers from The University of Colorado-Boulder and The University of Pennsylvania employed AI to assess personal qualities in college admission, concluding that an "AI approach to measuring personal qualities warrants both optimism and caution."
The authors emphasized the opaqueness and resource-intensiveness of the holistic assessment of personal qualities in college admissions, criticizing the secretive nature of these decisions. They underscored the potential of AI systems to efficiently identify patterns in data and apply learned models to new cases. The researchers used supervised machine learning to measure personal qualities in students' descriptions of extracurriculars, fine-tuning the RoBERTa model based on human ratings.
The findings indicated that computer-generated likelihoods of personal qualities were similar to human ratings, predictive of college graduation, and demonstrated validity across demographic subgroups. The authors recommended using AI to augment, not replace, human judgment, acknowledging that no algorithm can determine a university's admission goals or the most crucial personal qualities for its community.
RoBERTa Under Scrutiny
Concerns have been raised about the potential overreach of AI in admission, while others argue that these tools offer hope and a means to simplify the process. The key question is whether relying on AI diminishes the human touch in admission or enhances the ability to align with institutional mission and priorities.
Matthew DeGreeff of Middlesex School questions how the humanity and essence of an applicant, the nuances of a high school, and the complexities of diverse communities can be measured by an algorithm. Bob Massa of Dickinson College believes AI can identify personal qualities through college essays and recommendations, but it should supplement rather than replace human evaluation. Tom Bear of Rose-Hulman Institute of Technology sees potential in AI providing a first read for efficiency, but he remains cautious about relying entirely on AI for decisions.
Angel Pérez, CEO of the National Association for College Admission Counseling (NACAC), emphasizes the need to balance administrative improvements with potential effects on equity in the admission process. Pérez acknowledges the study's hope for integrating AI in college admission while cautioning about the importance of a 'do no harm' approach, particularly concerning equity.
Everything's Fine
Kevin Roose, a technology columnist for The New York Times, suggests that rather than competing with machines, humans should focus on becoming more human. While AI may not replace legendary educational leaders like Ted Spencer, it has the potential, when trained and developed ethically, to support successors effectively, sustainably, and equitably. AI admission officers aren't messiahs solving all challenges, but if used judiciously, they could enhance capacity and allow human colleagues to foster more meaningful relationships with applicants.
In conclusion, the debate over AI's role in college admission continues. While some view it as a promising solution to streamline and enhance the process, others caution against losing the essential human touch. The future may lie in finding a balance where AI augments human judgment without overshadowing the intricate and nuanced aspects of the admission process.
