Finding Middle Ground

Privacy/Sharing Tradeoffs

When you’re online—shopping, paying bills, posting on social media—you’re leaving a digital trail that may have privacy implications.

Sarah Rajtmajer and Priya Kumar, assistant professors and researchers in the College of IST, are examining online privacy in very different ways. But despite their opposing methodologies, they agree that what’s needed is a nonjudgemental understanding of privacy/sharing tradeoffs.

Rajtmajer’s Quantitative Approach

With a background in math and data science, Rajtmajer has primarily used quantitative methodologies to study privacy: The more people share, the more data there is to work with.

“I began modeling online behavior when there was an explosion of social networks that furnished an unprecedented amount of data,” she said. “I had done my Ph.D. work in graph theory, so I immediately tried to understand social networks as graphs.”

But Rajtmajer soon discovered that she couldn’t model peer influence as molecular diffusion, or users’ decisions to share personal information as optimization of a simple objective function.

“I tried to model user sharing in social networks using graphs and algorithms; then I’d collect ‘real’ data and my predictions never lined up with human behavior,” Rajtmajer said.

She began to dive into game theory, another mathematical language that describes strategic interactions among agents. The nodes in the network graph were no longer just abstract and homogeneous objects—now they had goals and strategies. More recently, she's begun to explore nuanced differences among individuals in privacy attitudes and behaviors.

“I kept being confronted by the fact that social contexts are really important,” she said. “My goal now is to leverage natural language processing and other computational methods while also being respectful of context, particularly heterogeneous attitudes about privacy in different populations.”

Kumar’s Qualitative Approach

Kumar conducts theory-driven qualitative research to study how we—society, researchers, educators—think about privacy and digital technology.

“I’m interested in the ways that privacy— an inherently social value—gets framed as an individual preference or ability, like having control over your information,” she said. “The individual dimension is certainly important, but privacy issues are baked into the way our digital ecosystem works, and that’s what needs to change.”

Kumar expressed frustration around literature and broader public discourse presuming that sharing less equals more privacy.

“It’s more complicated than that,” she said, “Oftentimes, disclosing personal information makes sense because of the context. For example, you usually wouldn’t share detailed medical information with your boss. But if you have a condition that’s affecting your ability to work, then it might make sense to tell them. It comes down to determining what is the appropriate flow of information, and that changes based on context.”

For over a decade, Kumar has embraced contextual integrity (CI), a theory that defines privacy as the appropriate flows of information. The theory includes a framework to identify the different components of an information flow and determine whether the flow is appropriate.

“CI has become a popular way to study privacy, though most researchers are only engaging the first half of the framework— identifying components of an information flow,” she said. “This is valuable because it more precisely describes what privacy concerns are, but if we want to actually protect privacy, then we need more research to engage with the second half of the framework and explain how to make information flows appropriate to their context.”

One way researchers can do this is by embracing more qualitative methods when using the CI framework.

“Most empirical research using CI employs quantitative methods, for example, to measure people’s privacy preferences or a system’s information flow,” Kumar said. “But people interpret information differently based on the nuances of a situation, and qualitative methods offer robust techniques for making sense of those differences.”

Kumar encourages collaboration among people from different disciplines to address thorny questions of privacy.

“The second half of the CI framework asks us to examine what are the moral and ethical values at play and how those values shape what we, as society, consider appropriate,” she said. “These questions are inherently subjective, and they are best addressed by teams of people who know how to think critically about technology, human behavior, and social implications.”