This story is from The Hill's Changing America publication.
The job of a college admissions officer is not an easy one. For any competitive higher learning institution, the admissions process used to hand-pick each incoming student has also come under increasing scrutiny in recent years.
To ensure the ongoing success of an institution, admissions officers are tasked with the nearly impossible task of efficiently evaluating thousands of applications each school year, with the expectation that their choices will reflect the institution’s standards, grow diversity and inspire enough students to enroll.
The process is a balancing act, and one that is expected to proceed without gender-based or racial bias. The problem? Humans are inherently biased, and schools are now beginning to realize the faults in their traditional approach to admissions — one that has placed an outweighed emphasis on test scores and transcripts, and often fails to find the human factor in their applicants. The flaws in this system also tend to leave underprivileged groups behind and keep underrepresented demographics as anomalies.
Surprisingly, the solution to this issue — to this lack of humanity — might be found in artificial intelligence.
“The mission of the organization is to bring a human aspect back into the admissions process,” said Andrew Martelli, the chief technology officer at Kira Talent, a Canadian-founded company that works with learning institutions around the world in hopes of delivering a more holistic approach to reviewing candidates.
Hopeful students applying to institutions that partner with Kira undergo a video interview process in which they will not encounter another live person. Instead, video- and text-based prompts lead applicants through a series of questions. Their answers are then used to evaluate things like leadership potential, verbal and written communication skills, comprehension of key concepts, drives and motivations, and professionalism.
Martelli said artificial intelligence has entered the picture in a beta phase, and one way it is used is not to evaluate students but rather the admissions officers and their possible biases.
“It's almost more of a science experiment, to understand things like: Are people accidentally or inadvertently introducing bias?” he said. “When schools express interest in it, they are presented with an AI-based tool that takes video data, and analyzes personality traits and behaviors. We take the very same footage that you view as an admissions person to get a sense of the applicant, and we have them run it through a series of algorithms. Schools are then able to run the algorithms, which give them AI-based data to then compare to what their human reviewers said.
“The idea behind the technology is to help the human reviewer ask questions of themselves: Did I see these traits or qualities? Am I missing something? So the emphasis is not on using AI to replace the human aspect of the process. Our whole focus is on helping the human be a better evaluator of other humans.”
It’s those principles that the company used last year in its partnerships with schools such as California State University Fullerton. Members of the admissions committee were able to pre-record questions for the students to answer through video interviews.
“Kira allowed us to bring our own personality,” said Deanna Jung, assistant professor of nursing and coordinator of pre-licensure programs. “We have a diverse faculty, so there was a diverse group of individuals reading the questions. Students were able to watch those videos and think, ‘OK, there are faculty who teach here who are like me.’ ”
Not all AI systems are created equal, nor are they without unconsciously programmed biases. At the end of the day, data scientists are still human, meaning many of the subjective choices they make as they create and refine training data can lead to racial bias in machine learning systems.
Human bias is an issue that pervades nearly every industry and facet of life, certainly not just in the process of college admissions. Over the past few years, society has become acutely more aware of how human prejudices can affect people’s lives. These biases can slip into AI systems creating what is called algorithmic bias, taking various forms from gender bias to racial prejudice and age discrimination.
Researchers are working to figure out how to mitigate the possibility of introducing racial bias into AI-based systems. One postgraduate researcher at the Massachusetts Institute of Technology, Joy Buolamwini, founded the Algorithmic Justice League with the objective of highlighting the social and cultural implications of AI bias using both art and scientific research.
One AI program that’s proven valuable addressed what’s known as “summer melt,” when enrolled students drop out during the summer, before their first fall semester even begins.
Georgia State University a few years ago used an AI chatbot called Pounce, which was designed to combat summer melt.
According to the university, Pounce was able to reduce the occurrence of summer melt by an impressive 22 percent one year, which translated to an additional 324 students showing up for their first day of classes in the fall.
Realizing the power of communicating with their students through text message but not having the human power to implement it, Georgia State partnered up with the Boston-based education technology company AdmitHub.
Over half of the university’s students hail from low-income backgrounds and many of them are first generation college students — a demographic that has shown the need for individual attention and financial aid, both of which aid enrolled students in showing up ready to start classes once the semester starts.
The admissions team worked with AdmitHub to identify these obstacles and fed information and answers into Pounce, which students could then direct their questions to at any time of the day or night by text message. In the first year of implementing Pounce, the AI-based system, had answered more than 200,000 questions by incoming freshmen.
“Every interaction was tailored to the specific student’s enrollment task,” Scott Burke, assistant vice president of undergraduate admissions at Georgia State, said on the university’s website. “We would have had to hire 10 full-time staff members to handle that volume of messaging without Pounce.”
What experts seem to agree on is that the sole use of artificial intelligence won’t be best practice for college admissions decisions anytime soon, if ever. Nevertheless, AI-based systems can serve an increasingly important purpose for schools, not only streamlining teams and processes but also promoting education about unconscious biases amongst admissions officers.
“I do believe that schools continuously look for ways to adjust their practices. I think COVID has also caused people to take a hard look at the processes that they use — to try to find ways to make them more convenient, to make them more accessible, to make them safer because of the social distancing and other requirements,” said Martelli. “I also think a lot of the social movements that we see in place today have asked for schools to take a harder look into their practices and the processes, and the ways they make these admissions decisions.”
As far as the future of AI-based systems, Martelli preaches cautious optimism, saying that it has to be implemented in the right ways. He said that along with the promise that AI shows, there is a lot of danger as well. Experiments over the years have shown just how easy it is for algorithmic bias to make its way into an AI-based system, and Martelli said a biased sample could perpetuate some of the problematic decisionmaking of the past.
“When you think about using those kinds of tools, we still think it needs a person at the heart of the whole system to make the judgment about another human,” he said. “Do I think there's promise there? For sure. Do I think we have to be careful about how we apply it? 100 percent.”