Facebook is expanding a program that grants researchers access to data about movement patterns in an effort to help improve our understanding of the spread of COVID-19, the company said today. Data for Good, which uses aggregated, anonymized data from Facebook’s apps to inform academic research, will now grant access to three new maps for forecasting the disease’s spread and revealing whether residents of a given region are staying at home.
The company will also prompt Facebook users to participate in a survey from Carnegie Mellon University that asks people to self-report any disease symptoms. The responses, which will be anonymized, could help researchers understand new hotspots as they develop or see where the disease has begun to retreat. Carnegie Mellon will not share any symptom information back to Facebook, the company said.
Last week, Google released public reports that use the company’s own location tracking services to reveal the degree to which people have changed their movement patterns in response to the global pandemic. Facebook had already made similar information available to academic researchers.
The moves announced today are designed to improve forecasting and response efforts in the United States and abroad. In an interview, Facebook executives said that the company could aid in disease recovery efforts while still protecting the privacy of individual users.
“We think that Facebook and the wider technology industry can and really must continue to find innovative ways to help health experts and authorities respond to the crisis,” said Steve Satterfield, director of privacy and public policy at Facebook. “But we don’t think that these efforts have to compromise people’s privacy. We think we can assist in the public health response while also continuing to protect people’s data.”
The tools released Monday include co-location maps, which illustrate the degree to which people who live in different areas are mixing; movement range trends, which show the degree to which people are staying home or going out; and a “social connectedness index,” which shows how likely any two people are to become Facebook friends, a measure of the strength of social ties in a given place. Communities with stronger social ties may recover more quickly than others, said Laura McGorman, policy lead for Data for Good.
The disease prevention maps rely on data from Facebook that has been processed to obscure individual identities, the company said.
“Measuring the impact of social distancing policies is absolutely critical at this stage, and aggregated data of this kind provides insights that protect individual privacy but are actionable for policymakers and researchers building predictive models,” said Caroline Buckee, associate director of the Center for Communicable Disease Dynamics at Harvard’s TH Chan School of Public Health, in a statement.
Andrew Schroeder, who runs analytics programs at the humanitarian aid organization Direct Relief, said that the new maps would help public health organizations and aid groups understand the effectiveness of stay-at-home orders and help plan response efforts. The data could inform changes to messaging from public health officials and eventual plans to end the current lockdowns, he said.
Schroeder said mapping efforts from big tech companies are helping researchers understand the effectiveness of social-distancing guidelines in real time, improving models that are tracking the spread of COVID-19. “Three weeks ago it was, ‘is anyone staying home?’” he said. “Now it’s where? How? How much? How much is enough? How does this affect case count? That’s the agenda.”
Facebook will also place a prompt at the top of the News Feed in the United States inviting users to self-report any disease symptoms to Carnegie Mellon’s Delphi research center. Given the continued widespread shortage of tests across the country, reports of symptoms can provide an imperfect but still valuable look at where the novel coronavirus may be spreading before public health officials become aware of it. Facebook will share anything you report with a random identification number, along with a statistical weight value that corrects for bias in the sample. (If different communities respond in different numbers, Facebook says, the statistical weight will account for that.)
Researchers can apply to get access to Data for Good at its website. Facebook is working with more than 150 universities and nonprofits to date, McGorman said.