A computerized early warning system for students in academic trouble

2 Mar 2017 - 09:00

It took 4 years to build and test, but that early warning system now exists. GSU's Vice Provost Timothy Renick debuted its results here Saturday at a session of the annual meeting of AAAS, which publishes Science. "We've done over 200,000 interventions," he says, referring to the meetings between GSU's advisers and undergraduate students. In each case, it was a computer rather than a human that noticed that the student needed help. And the predictions were often based on signs so subtle—getting a B- rather than a B+ in a particular course—that a human might never have noticed. 

The data that power these predictions came from GSU’s 32,000 undergraduates themselves. By tracking the students' progress and comparing the test scores and grades of those who graduate on time with those who don't, GSU created a statistical model of academic success. It became clear early on that certain combinations of indicators, such as difficulty in certain courses that were critical for later ones in the degree, were strongly correlated with the risk of graduating late or even dropping out of school. 

Advisers began to act on those patterns, inviting the students to meet and discuss their academic life. "Sometimes the student has just chosen the wrong courses or taken on too much at once," Calhoun-Brown says. "Or sometimes we found that they needed extra help with writing or math skills. Some needed help with time management." 

But did those early interventions work? Four years ago, GSU had achievement gaps similar to other urban universities with low-income students, with graduation rates about 10% lower for "at risk" students. "Today we have no achievement gap," he says. And the number of students graduating with science-related degrees has doubled among black students. The boost came from better retention of those students rather than changes to the student body, Renick says. "Over the same years that we were seeing these huge increases in STEM [science, technology, engineering, and math] graduates, we only increased the size of our admitted freshman and transfer classes by about 4%."

Other universities are signing up. With a $9 million grant from the U.S. Department of Education, 11 institutions are now launching randomly controlled trials of the GSU system on their own campuses. And administrators from South African universities are now exploring ways to collect similar data to help their students, who face some of the largest racial achievement gaps in in the world. 

During a press conference earlier in the day, Renick faced some pointed questions about protecting the privacy of the students and even about their "grit." An unnamed questioner asked whether GSU's intervention system might be robbing students of the benefit of making mistakes on their own. But Renick was bullish. "That's like asking people to not get help from tax accountants because they should just learn from their mistakes."

Calhoun-Brown says that GSU students are embracing the idea that a computer is watching their progress and alerting their advisers. "They're used to getting recommendations based on data from using Amazon and Netflix. They expect it. And now we have the data to share with them."