Tim McKay isn't intimidated by big data. The professor of physics at the University of Michigan, along with his colleagues, analyzes data sets on 200 million galaxies at one time. McKay also spends a lot of time working on an analytics problem back here on Earth: How can a university use data to assess how students are doing in class, and then use digital platforms to provide tailored feedback to keep them on track?
Colleges and universities nationwide are wrestling with how to best use analytics to gauge and improve student performance and retention. McKay characterizes the problem with the status quo this way: "I am now teaching an introductory class with 480 students enrolled. I interact closely with 12 of them."
McKay says the answer might look something like E2Coach. It's a digital platform that delivers messages to students based on their grades and scores coming into a course, their majors, their current study habits and their recent test grades in the course. An engineering student with a deep math background who aced AP physics in high school and got an A on the first exam in a class gets one set of advice; a liberal arts major with a weak math background and a C- on the first exam gets another. E2Coach even predicts the grade a student will get unless he or she does something differently.
The inspiration for E2Coach came from healthcare projects going on at the University of Michigan, such as one designed to help people quit smoking. Researchers found that different types of people -- those who really want to quit but don't think they can or those who could do it but lack motivation -- need different inputs.
Michigan has survey data from more than 2,000 students who have opted in to use E2Coach. McKay focused on "different than expected" students -- A students who received an A-, say, or C students who bagged a B. "The students who outperformed what was predicted spoke about changing their approach during class," McKay says. "If they didn't make changes, they'd end up with what would be expected." Changed behavior could be hiring a tutor or organizing a study group. The more such data the university builds up, the more it can guide students.
The university has been testing E2Coach with four introductory physics courses. Students choose whether to opt in and provide their data. Where McKay once interacted with a dozen of his 480 students, E2Coach lets him interact effectively with half the class, he estimates. "The use of technology enables us to pay attention to students individually and keep track of their performance and enhance interaction," he says.
Such initiatives are happening across higher education, and they naturally raise questions. Are such systems another example of "Nanny U," where professors and administrators baby-sit young adults who should know they'll keep getting D's if they don't do their homework? Can a social process such as learning really be captured in algorithms? Who runs these kinds of systems? Universities are examining these issues, but momentum is building to move ahead. "If we can find ways to make the education at Michigan 10% to 20% better, on a demonstrable and sustained basis, then the gains that accrue to the school are enormous," McKay says.
Download the February 2013 issue of InformationWeek