Using Adaptive AI to Mitigate Financial Aid Fraud

Using Adaptive AI to Mitigate Financial Aid Fraud

Here at GCOM, we use artificial intelligence and machine learning (AI/ML) to solve a variety of challenges for our customers across disciplines. However, fraud has recently become a pervasive problem for our higher ed clients, in particular. More and more colleges and universities are being targeted by fraudsters who submit phony applications in hopes of stealing financial aid. Our clients are finding that bad actors are even taking seats in classes, preventing real students from registering for what they need to complete their credentials.  

Financial aid fraud has been steadily increasing since the beginning of the pandemic—when the rapid pivot to online classes made it easier for fraudulent applications to fly under the radar—and shows no signs of slowing down. Student aid applications using stolen identities netted about $100 million over the past 12 months, up from $50 million annually during the height of the pandemic. 

While higher ed organizations recognize that fraud ties up seats in course sections and saps financial aid award funds, it also presents additional risks that few institutions consider. First, since the fraudster now owns a school ID and email, they have become an insider threat. Second (and maybe more importantly) there is a reputational risk if the stolen identity is used to cause further damage to other government entities and businesses.

Why aren’t higher ed institutions doing more?  


Generally, they are stunted by a lack of comprehensive, reliable data. Since many schools don’t have a centralized data lake or warehouse, they’re forced to manually cull data together and then attempt to run matching routines against it. Some institutions told us they just pass around spreadsheets of applicants and then eyeball the list manually. You may want to go back and read that last sentence again.  As you can imagine, this is not sustainable or scalable. ​
 

If we want to efficiently identify and prevent fraud, we need to bring data together because the initial crime and the subsequent tell-tale signs cross multiple transactional systems: Admissions, SIS, and LMS. GCOM’s Student Success Analytics (SSA) combines data from all our clients’ disparate on-premise or cloud-based systems. Then, our fraud models for admissions, financial aid, and enrollment use adaptive AI to look for patterns in the data and identify potentially fraudulent applications and actions. 

SSA helps our customers manage the pool of potentially fraudulent applications and tracks the trends in the identified cases over time. This allows staff to investigate these potential fraud cases by various means, including requesting additional verification of identity. Our AI models automatically incorporate the results of clients’ reviews of cases for continuous refinement, keeping ahead of the never-ending evolution of fraudsters’ tactics.  ​ 

Each time we implement SSA with a new client, we train the model to identify characteristics relevant to each institution’s profile. For example, at most community colleges, applications typically come from in-state students, so an out-of-state application combined with other factors may trigger staff review. We display the factors influencing the fraud probability score for each application. We take this approach because we are laser focused on transparency and want to ensure that the AI does not develop unnecessary biases. Providing our clients with the reasoning behind the “why” keeps us all honest and accountable. 

Our clients are seeing tremendous results with SSA. After implementing SSA Cloud Financial Aid with Fraud Detection​, College of the Canyons identified 96% of fraudulent financial aid applications in the first year, preventing about $172k in fraudulent payments. 

What to do next? Many institutions struggle with starting yet another project due to staffing shortages and/or tight budgets. If you are not sure if this is even a problem at your institution, GCOM can quickly assess your exposure through our Fraud Health Check service. This is a slimmed down version of our typical implementation that leverages our standard models to identify the scope of the problem. If you decide to move forward with us, this modest investment will be applied towards your first module of Student Success Analytics.  

-GCOM Leadership Team