The Ohio State University was represented by two members of a panel discussing the ethical use of student data at the Association of Public & Land Grant Universities (APLU) Annual Meeting November 13. TDAI affiliate Don Hubin, director of the Center for Ethics and Human Values and emeritus professor of philosophy, organized and moderated the panel, and Julie Carpenter-Hubin, who oversees data management for research and student success as Ohio State’s assistant vice president of institutional research and planning, joined two other panelists to discuss the issue.
While the software and systems that allow institutions to collect data on students have existed for some time, institutions and humans are still determining how to combine the information and use it appropriately. The panel is part of an effort to discuss the issue as it emerges.
While most universities don’t have “big data” on students per se, Carpenter-Hubin says, the “big-ish data” they do have includes information on high school academics, college course performance, learning management system activity, and student identification card use. These datasets haven’t been merged to create one statistical picture, however. “A lot of this data has been siloed, and we’re just now thinking about using it in aggregate,” she says. “Our ability to pull the data together is on an upward trajectory.”
But as new tools make combining student-related datasets increasingly easy, other dilemmas arise. “When you’re dealing with data about people, there’s always the risk of misuse,” Carpenter-Hubin says. “In a university setting, that it could be pretty harmful.”
One primary issue is privacy and the differences in how “private” is defined.
“People assume college-age students don’t care about privacy,” Hubin says. “In fact they do.” However, their concept of and norms around privacy may be different than for previous generations. He gives an example of students who post pictures online of parties but are upset when their advisors mention seeing the images. “It might violate an implicit norm they have,” he says. “You have to be careful in how you use the data so you don’t violate the implicit norms they expect.”
Securing truly informed consent to use someone’s information requires care as well; a standard acknowledgement of consent is not necessarily enough. “It’s not enough to tell people that you’re using it as agreed to in paragraph 73 of their user agreement,” Hubin says. “It turns out that’s not feasible. They’re not giving meaningful consent.”
“We have to think about how to really protect people’s privacy,” said Carpenter-Hubin. There have been several breaches in which “anonymized” data was linked to individuals because people were able to connect the dots and identify whom the information referenced. The issue involves thinking creatively about what is truly anonymous and which data can be meaningful without providing keys that identify individuals.
Trust can make all the difference. “It’s surprising how often, when people think their privacy is being protected, they’re willing to share,” says Hubin. As a result, the amount and types of information universities have about people makes their relationship with those individuals similar to that between doctor and patient or lawyer and client. “It’s not just a commercial transaction. There’s a need for a fiduciary relationship that arises when there’s an imbalance between the two parties. If we’re collecting this sensitive data on people, we have the responsibility to use this data the way they would want it to be used.”
The perceived unimpeachability of data can be a problem in itself. “You can look at big data and think you know something,” Hubin says. But if the information or assumptions baked into the data are incorrect, the data isn’t as informative as the user may think. This is particularly true when it comes to making predictions or conclusions about an individual based just on data points — for example, whether a student’s academic performance or social activities might indicate a need for some sort of intervention.
Data should be used as a general guide rather than a prophecy of inevitabilities. Carpenter-Hubin and Hubin agree that one way to mitigate this effect is to channel discussions about a student’s information, whether it is high school GPA or learning management system activity, through an advisor who knows the student personally and may be aware of outside factors. After all, the data can only show so much. “It’s great for support,” Carpenter-Hubin says.“For prediction, not prescription,” said Hubin adds.
The two also agree that ethics and user expectations are ongoing topics that belong front and center for everyone who works with data.
“For me, making sure my team and the other people who work with data on campus are aware of the ethical issues is my responsibility,” Carpenter-Hubin says. “I’ve seen some of my colleagues at other institutions pressured to use data in unethical ways. I think that only happens in an environment where people aren’t talking about these types of issues.”
“You can rush the tech part, but there’s a cultural understanding — you can’t rush that.”