To keep college students happy and engaged in their coursework, administrators at Georgia State University turned to a school fixture with blue fur and ferocious teeth: Pounce, the school mascot.

In the real world, Pounce is a fuzzy game-day presence, rooting on the university’s athletes. In the virtual world, the mascot is a chatbot enhanced with artificial intelligence.

The virtual version was introduced in the summer of 2016 to incoming freshmen, who could text questions to Pounce 24/7 and in just seconds get answers about financial aid, registration, housing, admissions and academic advising. Three years later, Pounce was rolled out to the entire student body, with broader capabilities—able not only to answer questions but also to initiate interactions on its own. For example, the chatbot can intervene when a student is determined to be at risk of failing a class or dropping out of school.

“Using predictive analytics, we can identify early risk factors rather than waiting for students to find solutions for themselves,” says

Timothy M. Renick,

executive director of the National Institute for Student Success at Georgia State in Atlanta. Because the chatbot converses in natural language, “students have come to see Pounce as a friend,” he adds.

Graduation guides

Georgia State has been at the forefront of the growing use of apps and analytics by colleges to help attract students, guide them through graduation and launch their careers. These digital initiatives are aimed at supporting all students, but they are particularly useful in helping Black, Latino and indigenous students earn their degrees, says

Amelia Parnell,

vice president for research and policy at Naspa-Student Affairs Administrators in Higher Education, a nonprofit based in Washington, D.C.

Naspa recognized colleges across the country last year with its inaugural Virtual Innovation Awards, for notable efforts designed to help students during the pandemic and beyond. One of the award winners was San Diego State University, which mines its data to identify students who haven’t registered by March or April for the fall semester, an early-warning sign that they may be in danger of dropping out. Teams of university staffers and peer mentors then reach out to these students via phone calls, text messages and emails. Early in the pandemic, they discovered that access to computers and stable internet connectivity were common obstacles these students faced with remote learning. In response, the university issued hundreds of laptops and internet-signal boosters.

The school also analyzes data from its frequent check-ins with students via an online communication and resources portal, to identify those who are unable to cover food and housing costs and provide them with emergency financial assistance.

Colleges’ AI use for certain tasks

Full or mostly full usage

Planning, piloting and initial usage

Tracking for potential use

Using chatbots and digital assistants for student success and support

Identifying students who are at risk academically

Sending early academic warnings

Tailoring instruction and remediation to student interactions and performance

Identifying students who are struggling with nonacademic issues

Full or mostly full usage

Planning, piloting and initial usage

Tracking for potential use

Using chatbots and digital assistants for student success and support

Identifying students who are at risk academically

Sending early academic warnings

Tailoring instruction and remediation to student interactions and performance

Identifying students who are struggling with nonacademic issues

Tracking for potential use

Planning, piloting and initial usage

Full or mostly full usage

Using chatbots and digital assistants for student success and support

Identifying students who are at risk academically

Sending early academic warnings

Tailoring instruction and remediation to student interactions and performance

Identifying students who are struggling with nonacademic issues

Tracking for potential use

Planning, piloting and initial usage

Full or mostly full usage

Using chatbots and digital assistants for student success and support

Identifying students who are at risk academically

Sending early academic warnings

Tailoring instruction and remediation to student interactions and performance

Identifying students who are struggling with nonacademic issues

Tracking for potential use

Planning, piloting and initial usage

Full or mostly full usage

Using chatbots and digital assistants for student success and support

Identifying students who are at risk academically

Sending early academic warnings

Tailoring instruction and remediation to student interactions and performance

Identifying students who are struggling with nonacademic issues

Many schools are still behind the technical curve. An estimated 20% to 25% of higher-education institutions don’t have data-analytics applications in place, due to such factors as budget constraints or lack of buy-in from faculty and staff, says

Charag Krishnan,

a partner at consulting firm McKinsey & Co. who primarily focuses on higher education. Other schools are data-rich but implement their initiatives inconsistently, he says.

Still, the onset of the pandemic accelerated the use of analytics as a tool to enable schools to broaden their support for students beyond the classroom. For example, the Georgia State chatbot is set up to recognize “trigger words” and immediately deliver the student’s text to a person experienced in crisis management, Dr. Renick says. “Students have told the chatbot about feeling depressed or suicidal,” even though they hadn’t previously gone in for counseling, he says.

After receiving a supportive chatbot message in the fall, one Georgia State sophomore replied: “Yeah it’s been stressful especially with deteriorating mental health and no will to live lol.” Within minutes, a university staffer followed up with the student and referred her to on-campus counselors as well as her academic adviser. “I needed to talk to someone, but didn’t know who. The chatbot was just handy,” says the student, who is Nigerian and a first-generation college student.

Many students have reported that they prefer the impersonal nature of the chatbot when revealing their problems, Dr. Renick says. Students are “revealing personal details about themselves, such as ‘I just lost my job and can’t afford a textbook,’ ” he says. “Ninety-five percent of the time, we can come up with a solution.”

Grades show results

Georgia State is now testing a messaging system for specific classes. Last fall, students in a required American-government course were randomly assigned to groups that either received coursework-support text messages or didn’t. The messages reminded students when their assignments were due, offered study tips and practice exams, and even solicited direct feedback for the professor.

Overall, students who received chatbot messaging were more likely to earn an A or B in the class than those who didn’t receive messages, the university found. First-generation students receiving messages earned final grades about 11 points higher than similar students not receiving messages. “That’s a full letter grade better,” Dr. Renick says. “It’s leveling the playing field.”

One of the students who received the texts is

Zul-Qarnain Hossain,

a 20-year-old sophomore from Duluth, Ga., who is studying computer information systems. He says he wasn’t struggling in the government class, but that he appreciated the chatbot support. “It gives us the feeling that the teacher is reaching out and trying to help us,” Mr. Hossain says. “It’s a big morale booster and keeps us on our toes.”

Similarly, San Diego State uses data analytics to identify classes that have higher rates of students who earn a D or F or withdraw entirely—typically introductory classes for freshmen and sophomores. Using the college’s messaging system, students enrolled in these classes are encouraged to attend supplemental instruction sessions, which are led by students who previously took the class.

Move-in day at San Diego State University, an award winner for virtual student services.



Photo:

Bing Guan/Bloomberg News

For her freshman chemistry class,

Deegan Roecker,

a San Diego State student from Redondo Beach, Calif., attended several supplemental sessions that she learned about through the school’s messaging system. “It was a cool way to meet other students and have people to study with,” she says. “As a freshman, it’s hard to put yourself out there and meet other people.”

Ms. Roecker, who is now a 22-year-old senior majoring in kinesiology, says supplemental instruction was a “huge factor” in her final letter grade—an A. In addition to attending classes, she works as an administrator for the college’s supplemental-instruction program, which continues to grow because of its effectiveness.

An analysis of grades in an introductory calculus class last fall found that 100% of students who attended four or more supplemental instruction sessions passed the class. Of those who attended one to three sessions, 75% passed. Among those who didn’t attend any sessions, more than half earned a D or F or withdrew.

More to come

For the next phase in data analytics, some schools are exploring or implementing advanced machine-learning applications that can analyze 150 or more attributes of their student body going back years into historical data. Understanding how these various attributes are related and how they are associated with student outcomes allows schools to offer individualized support based on the students’ needs.

For example, Georgia State has found that students majoring in economics who get below a B-minus in their first math course have a very low chance of graduating. “This isn’t new data, but we’re using the data more proactively,” Dr. Renick says. “Let’s try to get you some support in math before you take killer classes in economics.”

Analytics can also play a role in recruiting students in the first place. McKinsey published a report in April that detailed the efforts of a small college that wanted to prioritize its marketing resources. Using machine learning, the university (not identified in the report) was able to create archetypes of high-school seniors who were most likely to apply. In the end, using those archetypes to target the 10% of students likeliest to apply accounted for about 90% of eventual applicants, the report found.

All this data analysis raises concerns about student privacy and data breaches. In a survey of more than 16,000 undergraduate students at 71 U.S. institutions, 49% of the respondents agreed or strongly agreed with the statement, “I trust my institution to use my personal data ethically and responsibly,” while 17% disagreed or strongly disagreed. However, about half said they didn’t understand how their personal data was being used. The findings were published in a 2020 technology report from Educause, a nonprofit that focuses on information technology in higher education.

In its recommendations, Educause advises institutions to inform students about what data is being collected and how it is protected. It also says students should be allowed to opt out of data-collection initiatives that aren’t mandated by accreditation organizations, which certify whether a school meets minimum academic standards.

Ms. DeCarbo is a writer in South Carolina. She can be reached at reports@wsj.com.

Copyright ©2022 Dow Jones & Company, Inc. All Rights Reserved. 87990cbe856818d5eddac44c7b1cdeb8

Leave a Reply

Your email address will not be published. Required fields are marked *