Here’s how Georgia State University is using predictive analytics to identify at-risk students and intervene quickly – and how they’re now testing the approach across other institutions.
SPOTLIGHT ON INNOVATION SERIES
The US Department of Education has awarded multi-million dollar “First in the World” grants to 18 colleges and universities that are innovating to solve critical challenges with access, recruitment, retention, and student success. At AI, we have interviewed each of the recipients to learn more about the projects these institutions are pursuing, how their approaches are unique, and what other colleges and universities can learn from these new efforts.
This was the second year of the First in the World grants. You can read our interviews with the 24 institutions that received 2014 grants here.
During the past ten years, officials at Georgia State University have tracked more than 140,000 student records and 2.5 million grades in order to identify mistakes that put a student at risk of dropping out. A decade later, they’ve identified more than 800 different mistakes, and continue tracking all 30,000 of their students so they can quickly intervene. Their next challenge? Using the same approach to track students at 11 different institutions across the country to see if they can replicate those results.
Many institutions are already using predictive data, explains Tim Renick, GSU vice provost and vice president for enrollment management and student success, but no one has done a validated study to show what the impacts are, and if they’re positive impacts, how can they be maximized. Now with the help of a nearly-$9 million grant, officials at GSU will do just that in their project MAAPS: Monitoring Advising Analytics to Promote Success. We talked to Renick to learn more about how GSU is using predictive data, what the scale project will look like, and the benefits they think the data holds for the future.
Replicating a Personalized Approach with Technology
“We know what works in higher education. It’s very effective when students get individualized, personalized attention on a regular basis,” Renick points out. “That’s the model that elite liberal arts colleges have been using for generations, going all the way back to Cambridge and Oxford. You have somebody who’s monitoring what you’re doing, who’s interacting with you on a regular basis. Students learn more, they persist more, and so forth.”
“The problem is that because that kind of support is expensive, it has been inequitably distributed across our student populations,” Renick adds. ” Wealthy students have been able to go institutions that provide that kind of support, and at big public universities like Georgia State, the students haven’t historically been able to get that personalized attention.”
He thinks data is the answer to bringing that approach to some of the biggest institutions in the country — institutions serving 50,000 to 70,000 students or more. Renick emphasizes that using data doesn’t mean GSU will interact with their students less. Instead, they added 40 additional advisors and are having more interactions than in the past. “What it does allow us to do is be much more personalized in our approach,” he explains.
“What we’ve tried to do is leverage technology to give these students more constant and personalized support.”
Tim Renick, Georgia State University
They used ten years worth of GSU data to look at what students have done historically that put them at risk for dropping out. They identified 800 different academic mistakes that put students at risk, such as taking the wrong course for their major, or getting a low grade in an introductory math course but needing to taking an upper-level accounting or chemistry course. Each one of those 800 mistakes now triggers an alert when it occurs for any of GSU’s current 50,000 students, with the goal of preventing them from becoming drop outs.
In the past, no one would notice that a student signed up for the wrong lab sequence, but two semester later someone would say, “Oh, you signed up for the wrong classes and now you need to go back and take this other lab sequence because that’s the one you need for your major,” Renick explains.
Now, any of those mistakes triggers an immediate alert in GSU’s Integrated Planning and Advising Systems (IPAS), and within 48 hours an advisor is contacting students to help them correct or mediate the problem by registering for the correct courses, finding a tutor, or accessing supplemental instruction. They’ve used the system for more than three years, and last year they racked up more than 43,000 one-on-one student interventions that were prompted by alerts from the platform.
“Now the advisors are interacting and can say, ‘Oh, we see you want to be pre-med and you did well in your first chemistry class but we also saw that you didn’t do so well in calculus. Let’s talk about how we can build up your math skills so by the time you take your next chemistry course you’ll have a better chance of succeeding and getting a strong grade,’” Renick explains.
At least half of the interventions are in-person, especially for weightier issues. Part of the grant will also be used to fund additional advisors so they can give the kind of personal attention to the students.
Data Is Key to Closing the Achievement Gap
GSU first started using data and predictive analytics to help their large population of at-risk students ten years ago. Nearly 60 percent of GSU students are Pell-grant eligible, more than half are minority students, and they also serve many first-generation and non-traditional students.
GSU graduation rates are up 22 points, which translates into graduating nearly 1,800 more students each year than was the case five years ago. The biggest gains have been among their at-risk students:
- Degrees conferred to African Americans has increased more than 80 percent
- Degrees conferred to Latino students has more than increased 120 percent
- Degrees conferred to Pell Grant students have increased more than 90 percent
- African American and Latino students are graduating at slightly higher rates than white students
- Pell Grant students are graduating at slightly higher rates than non-Pell students.
GSU has eliminated the achievement gap, making it one of the few schools in the country that can claim that accomplishment. “The students who are benefitting from this monitoring and these interventions seem to be the students who were most at risk before, which is understandable, Renick explains. “They were the ones who had the least family support and the least knowledge about how to make these choices and identify when they are at risk.”
Scaling Up on a National Level
“Can we verify this same approach as effective across other institutions?” is the question that Renick hopes his team can answer. GSU is partnering with 10 other large public institutions, all of which are members of the University Innovation Alliance: the University of Texas at Austin, Ohio State University, the University of Central Florida, Michigan State University, Arizona State University, Purdue University, Iowa State University, the University of Kansas, the University of California Riverside, and Oregon State University.
These institutions will use the same kind of data to track 10,000 low-income and first-generation students, and will give those students the same sorts of interventions when they go “off-path.” They will start with a cohort of Fall 2016 students and track them over the four years of the grant to observe multiyear impacts.
Keys to Success
Scaling up presents a few challenges, such as different advising practices and different technology platforms across the eleven institutions, Renick notes. For example, GSU uses a centralized office for advising and all students interact with that one office. At Purdue, on the other hand, each college and even some departments within a college run their own advising. Schools also have different technology platforms for tracking data, as well.
Renick adds that the use of different platforms was considered a strength by the Department of Education in funding the project. “They would have been less interested if we were just testing one way of doing advising, and just one technological platform. What they want to see is not ‘Does this particular vendor have an impact over another vendor?’ but ‘Does this general approach to try to help students by using predictive analytics and data, reaching out to them, make a difference, especially for low-income students?’”
To ensure that all institutions use a parallel approach to advising and tracking data, they have planned face-to-face training meetings and have a project lead on each campus. There are also plans to convene on a regular basis to ensure common approaches and definitions. GSU also is sharing its current alerts with the other institutions.
Why You Should Watch This Project
“We have to find a way to succeed given the demographic changes of the U.S.,” Renick explains. “Given the economic implications, it’s no longer acceptable for us to have 15- or 20-percentage point gaps between the graduation rates for middle- and upper-class students and low-income students.”
One lesson they’ve learned so far is to avoid preconceptions. For example, one criticism was that using data and predictive analytics would result in students being advised into “easier” majors. GSU has found the opposite to be true because their two fastest-growing majors are computer science and biology. What GSU was doing to support at-risk students was inadequate before, failing to diagnose and address student problems quickly, Renick explains. Now with proactive, timely interventions, they’re giving students a fighting chance.
“Make sure that you don’t get stopped by simplistic criticisms against using data, because inaction can have a costly effect on students.”
Tim Renick, Georgia State University