Office-based employees scrambled to adapt to remote work during the COVID-19 pandemic. Anecdotally, they were successful. But just how did they do it?
Check the emails. Microsoft recently gave researchers access to more than 360 billion work-related Outlook email exchanges. These communications from 1.4 billion accounts at 4,361 companies came without content—but with useful information about their frequency and pattern.
That’s a lot of emails. And Carey Priebe, professor of applied mathematics and statistics at the Whiting School of Engineering, helped find the story in them. His work attacking real-world problems ranging from biomedicine to sex trafficking uses statistics and graph theory to unlock how complex networks function.
Scholars often have questions for which data can’t supply the answers. “I take that as the ‘nattering nabobs of negativism,’” says Priebe. “I turn it around: Can we say something useful with the data that we have, even though we can’t answer your question?”
The analysis performed by Priebe and other researchers required 55,000 computer hours and offered a new window into changes in work behaviors driven by a global pandemic. In a working paper published in Stat.ML in September 2021, the authors observed that already existing silos in the workplace (and communications within them) intensified as employees worked remotely. Communications within those groups also deepened as a result. The researchers also found that while email did not replace the “serendipitous, in-person interactions” of pre-pandemic office life, the silos created in lockdowns were more “dynamic”—with memberships within them becoming less rigid as employees worked remotely.
“The increased siloing we observe need not be feared,” they concluded.
Existing relationships with his former Johns Hopkins student and longtime collaborator Christopher White, PhD ’09, who is the managing director for special projects with Microsoft Research, and with Jonathan Larson, principal data architect at the company, brought Priebe to the project.
“I’m their chief academic egghead when it comes to anything related to statistics on graphs,” he says. Priebe adds that he is eager to bring more powerful quantitative resources to bear on the email data, fashioning “hypothesis tests” from the paper’s larger conclusions to better interrogate and refine them.
“People identify with how things changed when the lockdown happened,” Priebe observes. “So I found that gratifying—and a little bit surprising and very instructive.”