Using AI to help more college students graduate

When I started as Undergraduate Dean of John Jay College, our senior graduation rate was 54%. I knew we could do better for our students.

John Jay serves many first generation Americans and nontraditional students. Many are working while studying, some are parents and the main source of income for their families, and have taken different paths to their degrees. We understand that they are dealing with competing priorities in a way that the traditional undergraduate student is not, which often means that many students are unable to complete their degrees on time.

To address this problem, we partnered with DataKind, a nonprofit that works to tackle the world’s toughest challenges using data science. For two years, we collaborated on a predictive Artificial Intelligence (AI) model to identify students most at risk of dropping out. We looked at indicators including years of enrollment, grades, and number of hours passed to create a risk score for every student, while controlling for bias. Students identified as high risk of dropping out received extra, proactive support from advisors such as one-on-one coaching.

Hundreds of students benefited from this program, which ultimately drove the College’s senior graduation rate up to 86%.

In 2021, Google.org gave Datakind a grant to help expand the model to transfer students, a group of students who are at higher risk of not completing their education. I'm proud to announce they are reinvesting in this work. With this new grant funding from Google.org, Datakind will continue the work to scale our AI-powered intervention program to 6 additional CUNY campuses, and build a roadmap for any school across the US to replicate our success.

During the building process I dedicated as much of my time toward getting to know the tool as I sought feedback from the community that would be impacted by it. This meant prioritizing conversations with students, academic advisors, and other administrators and faculty and answering questions about the ethical use of a predictive model to support student success. This transparency and collaboration with our community made space for the culture shifts that we needed to successfully integrate with AI.

Starting with one college within the CUNY system gave us time to iterate and ensure we got the model right. We had the time to bring the right historical data to the table and to keep a laser focus on results, while closely monitoring for risks like bias in the model. As our focus shifts towards scaling this impact beyond John Jay College and CUNY, we’re able to share the questions we asked ourselves, and encourage others to be ambitious in the goals they set for their own students’ success.

AI has the potential to simultaneously deepen and scale the institutional capacity for care of our students. Each year, John Jay’s academic advisors are assigned upwards of 750 students. The AI model helps identify the ~200 students who could benefit most from a bit of extra support to succeed and make informed decisions about where and how to deploy our limited resources, including the time of advisors.

The biggest barrier to scale is often cost, particularly for smaller institutions or those with limited funding. We’re often scraping together a patchwork of solutions. Support from Google.org and organizations like DataKind, not only helps make innovation more accessible, but it also unlocks the potential to scale our impact. The playbook we’re developing to share our learnings will ensure that these learnings not only benefit students at CUNY, but also help students across the country as well.

When you bring data and AI into the mix, you reshape the fundamental way that institutions see these students. The data itself can’t tell us who these students are. But what it can do is tell us a story about their barriers and reveal pathways to unlock their success. This project has fundamentally reshaped how I think about building a culture of belonging – informed by data, and powered by community.

Today, I can confidently say that John Jay is in a different place for student success and advising. From the front lines all the way to the top of the College, people have seen the impact of AI in higher education firsthand. Transformation can be slow when you have limited resources. But we've found a way to accelerate it by using AI, collaborative leadership, and thoughtful partnership.

Learn more about DataKind’s work to advance economic empowerment in communities across the U.S, with support from Google.org.