Using Artificial Intelligence to Improve Access to and Success in Higher Education

By  - 

Although the term “artificial intelligence” (AI) conjures images of science fiction movies with robots attempting to take over the world, the reality is that AI is a combination of algorithms that can recognize patterns and “learn” how to automate the collection and distribution of information.

While AI is already being used in a wide variety of applications in data management and predictive analysis in corporate settings, it is also emerging as a potential solution to challenges facing higher education — specifically in the areas of admissions, student engagement, and career placement.

When students accepted for admission to Georgia State University (GSU), for example, have questions about when to file their FAFSA, how to register, and where to find housing information, they ask “Pounce.” Named for the university’s mascot, this AI-powered chatbot, or virtual assistant, offers assistance with admissions and can be easily and quickly accessed via a cellphone.

All universities experience “summer melt” — a term that refers to the scenario in which high school graduates enroll in college but end up dropping out before the fall semester even begins. However, this phenomenon is most prevalent among low-income and first-generation students. 

At GSU, Pounce helps counteract summer melt. 

Timothy M. Renick

“Underserved students are tripped up by requirements such as FAFSA verification and submission of immunization records because they don’t have the support system at home to help them,” says Timothy M. Renick, PhD, senior vice president for student success at GSU. “They’ve graduated, so they no longer have high school counselors to guide them.” Because a university can be intimidating to those who don’t know who to call for help, they don’t call and just decide not to attend, he adds.

“We were losing 19 percent of a confirmed class over a summer when we decided to try a new way to engage [incoming] students,” says Renick, explaining that although emails have been used for follow-up, today’s college students generally don’t read emails or understand the information in them. “We needed an interactive solution that could give students an immediate response to a question, so we chose an automated texting platform that is available 24 hours every day.” 

GSU admissions and financial aid staff built a knowledge base of the most common questions asked by incoming freshmen. The school piloted the initiative in the summer of 2016 by giving approximately 3,000 incoming students access to the texting program and using the other half as a control group. When one of them texted a question to Pounce, the tool relied on AI to search the knowledge base for the best answer. If Pounce did not have the answer, the student received a message that a “human” would need to respond to their question and that someone would call within 24 hours.

“That first summer, we received over 200,000 questions in three months,” says Renick. With the ability to answer the most common questions — some of which may have been asked by 200 different people — via an automated texting tool, the admissions and financial aid offices benefitted greatly. They could now spend more time with students who had more complex concerns rather than answering the same simple questions over and over. 

“Students were happy because they didn’t have to be placed on hold or wait for a call back to get answers,” Renick says, “and they could ask their question any time of the day or night.” 

Students in a study center at the University of Oklahoma.

Results from GSU’s initial pilot project indicate that the program is effective at reaching the underserved students who are at greatest risk for experiencing summer melt. On average, first-generation and Pell Grant students sent Pounce 9.4 percent and 31.7 percent more messages, respectively, than others. Engagement numbers by race also show that Asian Americans, Hispanics, and African Americans sent 12.3 percent, 12 percent, and 3.4 percent more messages, respectively, than the average GSU student.

The first two years that the university relied on automated texting, it focused on incoming freshmen, and the service was only available during the summer. In fall 2018, GSU piloted an expanded program that has 4,000 of its 53,000 enrollees using Pounce throughout the school year to ask questions about how to pay a campus parking ticket, deadlines for registration, and a variety of other issues. 

“We used the chatbot to promote our career fair to increase awareness to all students and encourage them to attend this event that features 200 employers with internships as well as jobs,” says Renick. “Even though only 4,000 students have the chatbot, we saw 30 percent increased attendance this year.”

At the University of Oklahoma (OU), the Office of Admissions and Recruitment uses AI to power a chatbot on the admissions website. “The chatbot crawls the content of the university website and uses input from user questions for Q-and-A chat interaction,” explains Bryce Kunkel, senior technology and marketing coordinator for the office. “Having the chatbot look through the site for answers saves time and frustration for website visitors.”

“The bot is configured to answer questions students might have throughout the year, which includes [those] about campus, student life, admission, housing, financial aid, scholarships, and enrollment,” he adds. 

Bryce Kunkel

In the first four months of use, the bot answered almost 30,000 questions and had more than 10,000 unique conversations. But users are still able to speak with a real admissions counselor should they prefer or need it. 

“When a student needs to talk to a human, there are two ways that a person using the bot can interact with an admissions counselor,” says Kunkel. “During the day, we have staff assigned to log on and handle chats that request a human or that the bot cannot answer.” 

After normal office hours, users are able to submit a ticket, which is delivered via email to a staff member for follow-up.

“Right now, the information base for the bot is static; however, we have plans to allow it to securely integrate with other OU systems,” says Kunkel. “This would allow a student to ask a question like ‘What’s my application status?’ and the bot would return the student’s actual status.” 

He believes that offering a tool that lowers barriers to asking questions is beneficial to all students. “Calling a university to ask for help can be intimidating,” Kunkel admits. “We are continually improving the knowledge base with more answers — informed by the questions that are being asked — so the experience will continue to become better over time.” 

Eliminating Bias
Another potential area in which AI may be helpful on university campuses is the career placement office, says Kevin Parker, chief executive officer of HireVue. His company strives to change the way employers discover and hire talent by removing bias from the hiring process. It does this by combining video interviews with predictive, validated industrial-organization science and AI to augment human decision-making. 

“Employers have used AI at some level as long as eight or nine years ago, but we are seeing increasing adoption of AI tools not just to assess applicants but to also expand the talent pool to identify quality applicants,” Parker says.

Kevin Parker

An AI-driven initial interview reduces bias by asking the same questions of all candidates in the same manner, he explains. The removal of bias in initial assessments of applicants led to a 16 percent increase in diversity for one corporation one year after switching to video evaluations enhanced with AI, according to Parker.

The company, Unilever, not only used AI-driven video interviews as a late-stage step in the interview process for new graduates but also relied on AI-powered search tools to identify the academic background required for the job. In addition, Unilever used AI games to reveal skills and attributes possessed by candidates. 

Parker predicts that AI-assisted interviews through career placement offices at a university may be the next area of opportunity for the technology. This will help students avoid blindly applying to jobs for which they may not qualify by having them take part in video, AI-assisted interviews. “Students would be able to interview from any place and can be interviewed not for a specific job, but for a company that will then match them with opportunities that are right for them,” Parker explains. “Employers will benefit by widening their campus recruitment efforts with minimum investment, and universities benefit by expanding career opportunities for their students.”

This approach not only works for companies reaching out to universities to hire new graduates but can also help universities enhance their own hiring practices by expanding searches and removing bias from the recruitment process. Machine learning can automate the review of résumés to identify skills and experiences required for the position, which frees university staff to prepare for interviews. 

Although there are concerns that human biases may transfer to AI tools, there is a way to normalize the candidate pool, according to Nidhi Gupta, senior vice president of technology at the career-search marketplace Hired. In a Forbes article, he suggests that using an algorithm to analyze similar numbers of résumés from men and women, even if the full dataset includes significantly more résumés from the former, can create an unbiased result.

At the University of Arizona, although the administration has not yet implemented AI tools, the staff is evaluating their potential, says Kasey Urquidez, EdD, vice president of enrollment management and dean of undergraduate admissions. 

Kasey Urquidez

“Students and parents want access to information 24 hours a day because their schedules may not match our office hours. In fact, many students submit applications after midnight,” she says. “I don’t foresee us removing the human factor from our holistic review of applications, but AI offers opportunities to improve engagement.”

With one team focused on exploring AI opportunities as part of the university’s development of a strategic plan to prepare for the Fourth Industrial Revolution, Urquidez and her staff are talking with other institutions about their experiences. “One advantage [of] not being an early adopter is the knowledge we gain from others who share the lessons they’ve learned,” she says. 

In her conversations with peers at other universities, she has heard some caveats about adoption of AI technology. At one institution, a student hacked into the chatbot system and inserted alternate answers to questions, she says. “When a student submitted a question, the response was that if the student had to ask that question, then he or she should not be in college,” says Urquidez. 

As with all information technology systems, she notes, it’s important to ensure some level of security.

Although universities do have to consider security needs as well as how AI-driven tools can best support their goals and mission, the one guarantee is that students are open to new technologies.

College students are comfortable with AI and mobile technology, so it’s important to offer tools that fit their needs, says Renick. The feedback GSU has received has been positive — and not just from students. “I’ve had university staff asking for access to Pounce,” he says, “so they can easily find answers to questions they are [often] asked.”●

Sheryl S. Jackson is a contributing writer for INSIGHT Into Diversity. This article was published in our January/February 2019 issue.