Institutional Innovation: Maximizing student outreach through artificial intelligence
This is the second iteration of our Innovation Column, which will run the third Thursday of every month. If you think your institution is implementing a practice or program that is particularly innovative, please reach out to [email protected] and tell us about it!
With stagnant or declining enrollment, as well as dwindling state budgets for higher ed affecting most public institutions, administrators have increasingly found themselves between a rock and a hard place. How can they maintain their bottom lines, while simultaneously looking for ways to enroll more students, or even retain some of their high-risk enrollees?
One method — or innovation — has been the adoption of artificial intelligence in outreach efforts to scale them at lower costs. Alana Dunagan, higher education researcher at the Clayton Christensen Institute, explains that AI technology is a sustained innovation within the industry:
"AI is a potentially disruptive technology in a lot of different fields. The way I see it used in higher ed is mostly as a sustaining innovation, but that doesn't mean it can't have really break through consequences and impacts on how schools are able to serve students," said Dunagan.
"One of the powerful things about AI ... is that it allows schools to look at where students are at and what they need in a way more efficient and effective way than a staff member could and at scale," she added. "You've always had people in offices following up with students [...] but they don't have the bandwidth to do it so comprehensively; this is a sustaining innovation to these efforts."
A case study: One institution that's adopted AI effectively is Georgia State University. Last year, the Atlanta school experimented with the use of conversational virtual assistants throughout its admissions process, and it won the Technology Association of Georgia’s (TAG) Excalibur Award, which goes to the state's institutions and organizations demonstrating best use of technology.
Lindsay C. Page, a professor at the University of Pittsburgh and Hunter Gehlbach, a professor at University of California Santa Barbara, oversaw the trial of virtual assistants at Georgia State. They said in their study summary they wanted to address the phenomena of "summer melt" — a dreaded period for administrators in which students who have accepted admissions offers fail to actually enroll. Studies show summer melt is an issue that impacts 10% to 20% of college-intending students each year, especially those in the low-income, high-risk category.
Georgia State is a national model for enrolling students from underserved communities. A 2016 New America found nonwhite students make up 60% of the student body, more than half of them are eligible for the federal grant program, and about a third are the first in their families to go to college. With admissions officers so worried about enrollment rates at their institutions, especially for students that can fall off track, summer melt figures certainly make an impact.
The sustained innovation: Page and Gehlbach explained the conversational AI was employed to "efficiently support thousands of would-be college freshmen by providing personalized, text message–based outreach and guidance for each task where they needed support." As part of the methodology, system designers were careful to consider the student-level data. Pre-enrollment needs at Georgia State, student habit and performance analytics, and responses from initial questions posed to students were all part of design process.
What's key is the AI algorithm worked off these areas to answer questions on behalf of students in areas for which it lacked answers. These aspects were combined into a virtual assistant named "Pounce," which was supplied with a knowledge base from university admissions counselors. Any questions texted to the 24/7 virtual assistant that it could not answer were forwarded to officials and addressed immediately.
Pounce acts proactively, seeing when students are falling off track and noting the types of questions students may have but aren't taking the initiative to ask, then offering individualized suggestions rather than generalized advice.
In their article for the Harvard Business Review, Page and Gehlbach explained that other methods of this type exist, including individual counselor outreach or automated text messaging. But, the difference is that these strategies require a great deal of human capital to scale them effectively.
The virtual assistant reduced Georgia State's summer melt rate by 22%, or 4% from the previous year. The results are close to the strategies listed above, but with significantly less staff. Further, the test showed students exposed to the virtual assistant were 3.3 percentage points more likely to enroll on time. The system answered more than 200,000 questions from students; and according to the university's website, the virtual assistant helped the institution enroll 324 students that may have skipped the college experience.
What's the innovative potential? Out-reach at scale: The word automation tends to result in chills for higher education leaders, for fear that they will disrupt the sector's business model. The argument typically goes the rise in workforce automation means institutions may have fewer degree options to offer or will have to reconstruct what type of education is being granted to students.
But, as demonstrated by Georgia State, AI can serve as a complement to sustaining higher education business models if incorporated effectively, especially when there is A/B testing that couldn't happen in an analog environment.
"Of course, everyone is change averse, and higher ed is no different. If you look at how GSU is using the technology, it's not being used in a way that will disrupt the institution; it's being used in a way that enhances the institution's existing model," said Dunagan. "All technologies that have the potential to be disruptive can also be used in a sustaining way. That's why for a disruptive technology to really be like that, it has to be wrapped in a changing business model in a highly disruptive way."
As institutions struggle to find ways of cutting costs while also innovating and enrolling more students, adopting technologies like virtual assistant to enhance existing efforts could pay off. "There are a variety of approaches with AI, and there may not be enough data to say which is better," she continued. "But certainly we have enough data from this experiment to say this is a potentially good approach and one that other institutions should look at."
Scott Burke, Georgia State's assistant vice president of undergraduate admissions, certainly agreed, explaining on the university's website the technology has been a success.
“Every interaction was tailored to the specific student’s enrollment task,” said Burke. “We would have had to hire 10 full-time staff members to handle that volume of messaging without Pounce.”
But of course, every innovation comes with its challenges, which is why leaders should make sure technology adoption aligns with institutional values, goals and existing infrastructure.
Follow Shalina Chatlani on Twitter