In The News
Can Better Data Keep Students From Dropping Out of College?
The Atlantic | | May 22, 2012
One of the biggest challenges to college education in the U.S. is the staggering number of students who don't, in the end, earn degrees. Attrition doesn't just mean a bunch of Americans with half-finished college careers; it also means huge sunk costs for students. Which often translate to student-loan defaults.
Much of that, says Charles Thornburgh, a former senior executive at Kaplan, is preventable. "Students are operating in an information vacuum," he says. It's very easy to be spectacularly non-strategic when it comes to selecting the courses and majors that will inform what students learn -- and, in turn, what career paths they take. (Or don't take.)
Higher education, in other words, has not been terribly data-driven.
Thornburgh is hoping to change that. Today, he's launching Civitas Learning, a digital-education platform that uses predictive analytics to help guide educational decision-making. Civitas works with partner institutions -- and also, more specifically, with the data those partner institutions have gathered about their students -- to identify trends about the classes students enroll in, the majors they take on, and other factors that can determine career courses and overall success in their post-collegiate lives.
Particularly as the college experience becomes less about setting four years aside to explore, and more about fitting education in with jobs/kids/etc., making smart decisions about courses will have a direct bearing, Thornburgh argues, on students' ability to stay in school.
Civitas' current and potential partner institutions include a wide range: everything from community colleges to four-year universities to proprietary schools. Diversity here, as in other aspects of higher education, can be a benefit. Civitas takes the data those schools have gathered about their students -- demographic, behavioral, academic -- and then anonymizes, normalizes, and analyzes it. And: combines it. The company then uses the data to identify trends and takeaways: which courses and tracks are most beneficial, and for which kinds of students; which courses and tracks tend to have high attrition rates; which instructional approaches tend to be most effective at ensuring good educational outcomes. Civitas then translates those insights into real-time recommendations for students, instructors, and administrators through a customized platform.
This is digital education with a (literally) old-school twist. One of the most exciting elements of the new explosion of online learning is the data that all the new platforms will be gathering about students' demographic backgrounds, learning styles, and general progress through course materials -- all of which can be used to create better, more effective learning environments. But while the most notable digital learning services suggest all the new data that can be generated on their platforms, Civitas is making use of data that are already there: administrative and demographic and behavioral records, the kind that have always been collected by schools -- but which have generally been locked away in the filing cabinets of administrative offices.
And huge amounts of data, Thornburgh points out, translate to reliable analytics. For a given course, "once you get good at predicting students' engagement patterns," he says, "for every starting group of 100 students, you can predict how students will interact with the course." And while the Civitas platform is purposely granular -- it allows students and instructors to track student progress not just on a course-by-course, but test-by-test basis -- the overall point is how course performance translates to college performance in general. One of the key insights Civitas offers is its insistent connection between present and future. The minor choices students make, it argues, are not minor choices at all: They have a direct and crucial bearing on students' ultimate success, in academia and in what will follow. Civitas' models provide a real-time view of which students are at risk of dropping out of school; which courses and degree paths might be contributing to attrition; and which resources and interventions might be most successful at preventing attrition.
The point isn't to take the serendipity out of college education, Thornburgh is quick to note. Rather, it's to help students and teachers and administrators make better-informed decisions when it comes to class selection and other choices that can have a bearing on a student's overall academic record. Basically, "we give them Netflix-style recommendations," Thornburgh says. And they can do with those recommendations what they wish. The Citivas platform suggests classes that have provided avenues for success for similar students in the past; and "we can warn them about taking combinations of classes that have shown to be toxic for students."
Civitas is hoping to do well by doing good. The company, based in Austin, has raised $4.1 million in venture capital funding. And going forward, Thornburgh says, it will charge membership fees of its partner institutions -- with its value to those institution increasing, the thinking goes, the more of them that sign on to the service. More data means more insight. In education, as in everything else.