AI’s pattern recognition abilities can be a game-changer for schools. However, even as more schools turn to new tech like AI, experts say implementing these solutions at the individual school level can only do so much.
A SURGE IN FRAUD
“With the rise of AI-powered spambots, which mimic human behavior, can complete assignments, participate in discussions, and even submit essays ... [we're seeing] a new and serious threat to academic integrity and the proper allocation of financial aid,” Chad Bandy, managing director of higher education advisory at the professional services company KPMG, said in an email to Government Technology. “It's a digital cat-and-mouse game, with fraudsters constantly finding new ways to slip through the cracks.”
Jory Hadsell, vice chancellor of technology at Foothill-De Anza Community College District in California, said the fraud typically follows a three-step pattern. First, bots or bad actors submit fraudulent applications, sometimes using personal information gathered from data breaches or online sources. Once accepted, they enroll in online courses and use their enrolled status to apply for and receive financial aid. Then, they withdraw from courses before they’re required to pay back the funds.
At Santiago Canyon College in California, President Jeannie Kim said the explosion of online learning and financial aid programs, particularly during and after COVID-19, made higher education institutions an enticing target. She said fraudsters even populated waitlists, ensuring fraudulent enrollments remained in the system as the institution identified and removed some fake students. Recently, one anthropology class — Kim noted classes toward the beginning of the course catalog alphabetically are often targeted — was filling up quickly in a way it hadn’t before. As a test, the dean and instructor raised the student maximum a bit at a time.
“Every time they up the cap, it filled immediately,” Kim said. “Upped it again, filled immediately, until it got to 120, but 110 were fraudulent.”
She said the financial toll of fraud is severe, diverting financial aid funds from real students and inflating administrative costs. Beyond financial strain, staff can be overwhelmed by manually verifying applications and enrollments and attempting to recover misallocated funds.
A surge in fraudulent enrollments can also undermine trust within an institution and from the public’s perspective. For Kim, the biggest impact of classes filling up with fake students is that they block legitimate students from enrolling in classes they need to complete their degrees or advance with plans to transfer.
“That is the worst kind of barrier that we could erect,” she said.
HOW AI IS TRANSFORMING FRAUD DETECTION?
Both Santiago Canyon College and Foothill-De Anza Community College District turned to AI to combat these rising threats through a tool called Lightleap.
Until recently, the schools relied on a combination of manual processes and data analysis to identify fraudulent enrollments. These traditional methods, while somewhat effective, were time-consuming and often failed to catch more sophisticated schemes.
The California Community Colleges application portal uses an identity verification tool called ID.me, but it’s voluntary, Kim said. At Foothill-De Anza, administrators supplemented this with an internal fraud detection dashboard that flagged suspicious applications based on a set of predefined criteria, such as out-of-state ZIP codes or duplicate phone numbers. Financial aid officers reviewed applications, and faculty looked for fraudulent students after classes started. Santiago Canyon College developed a similar homegrown system that Kim said placed a heavy burden on faculty.
The schools were able to use successful detections from their previous systems to train the new AI system from Lightleap, and saw staggering results. At Foothill-De Anza, they caught double the cases of fraud just at the application level.
“Now at that front gate, we’re catching the bulk of it and now we can really focus on what’s happening in registration,” Hadsell said. “If your overall enrollment is up, but some of that's not real, then that's a big financial risk for the institution.”
Similarly, at Santiago Canyon College, the system detected and removed 8,000 fraudulent students. The following semester, 7,500 real students were able to take those spaces in class.
“We gave our students their seats back, and that's the big piece,” Kim said.
According to Kim, Santiago Canyon's fraud detection system has a 99 percent accuracy rate. Both schools contact suspected fraudulent students over email or phone to ask for identity verification, and Kim and Hadsell said real students haven’t complained about the process. Faculty haven’t complained, either, as the detection takes place largely at an administrative level and reduces their workload significantly.
UP NEXT: A COORDINATED RESPONSE
As AI-driven fraud detection continues to evolve, experts say the next step is collaborating. California Community Colleges uses a unified application system with no application fee, meaning bad actors can easily apply to all 116 colleges for free.
“If we catch a fraudster, is there a way that other schools can have access to that data, so that if they have an application with that same data come through, like you catch it in one place, you catch it everywhere,” Hadsell said.
Currently, schools could report fraudulent applications to CCCApply, and the state could notify the schools, but it is not an automated process, he said.
“Collaboration among educational institutions to share data and insights will be pivotal, creating a united front against fraud,” Bandy said. “It will be like building a digital fortress that gets smarter over time.”
