Flawed, incomplete smartphone location data is being used to fight COVID-19

Public health officials are turning to mobile location data to learn how the coronavirus is spreading. But researchers worry the data just isn’t accurate enough.

Flawed, incomplete smartphone location data is being used to fight COVID-19
[Photos: Qim Manifester/Unsplash; Usis/iStock]

For many years, the use of location data emanating from our mobile devices has been viewed in the context of privacy and civil liberties trade-offs. Many people have agreed to trade data showing where we are and where we’ve been in exchange for the convenience of finding the nearest gas station or getting the most relevant weather updates. Debates about the use of mobile location focused on the value of customized ads, big data business insights, or the sheer creepiness of our phones as location-tracking devices.


COVID-19 has changed that. Now, even some privacy advocates see our phones as geographic trackers revealing information that can help us understand the impact of the pandemic and even save lives.

“I have been changing my mind based on evidence that is coming in,” says Irina Raicu, director of the Internet Ethics Program at the Markkula Center for Applied Ethics. “There are arguments saying we are in an emergency situation and we’re facing life-and-death situations and we need all the help we can get.”

But some epidemiologists, civil rights advocates, and researchers who focus on tech’s impact on society worry that location data gaps and flawed methods of analysis could mislead decision-making about the distribution of healthcare supplies and other resources. It could skew the perception of the disease’s impact on vulnerable communities and further entrench distrust in the healthcare system, they say.

The ACLU recently cautioned against using location data for contact tracing, citing its questionable accuracy and spotty coverage of certain groups. Raicu points to ultra-Orthodox communities in Israel who did not receive mobile notifications about the coronavirus because their Kosher phones block internet connections and mobile apps. Some believe that same issue led to a slow uptake of social distancing among Hasidic Jews in Brooklyn.

All the conundrums about giving up privacy to save lives are moot if the data is faulty to begin with.


Fragmented data, skewed conclusions

The Centers for Disease Control and Prevention here in the U.S., the Health Ministry in Israel, and the government of Singapore are now analyzing mobile location data in the hopes of tracing and preventing physical contact that can spread the virus.

“My concern is that given our environment, we’re going to get fragmented data that will be about some populations but not about others,” says Raicu. If entire groups of people could be missed by mobile phone tracking, she says, relying on that incomplete or inaccurate location data in the hopes of tracing contact among carriers of the coronavirus could have harmful implications far beyond privacy invasions.

“We risk getting to the wrong conclusions,” Raicu says.

Cobbling together a truly comprehensive mobile location database would be difficult, says Laura Schewel, CEO of Streetlight Data, which uses mobile location data to gauge transit metrics for municipal governments.

“I’ve seen most of the mobile location data sets and telco data sets, and they’re not complete,” says Schewel. “Let’s say I realize everyone who was at this bat mitzvah three weeks ago was exposed. AT&T has maybe 30% of those people, so there’s no way unless I partner with every app and every telco to trace down that contact spread.”


Another reason for location data gaps: Some people do not have cellphones or smartphones at all. Pew Research reported in 2018 that 9% of Americans aged 65 and over—a group at higher risk for severe illness from COVID-19—do not have a cellphone or a smartphone. Meanwhile, children, who may be carriers of the coronavirus, often do not carry mobile devices.

Despite its drawbacks, Stephen Mooney, assistant professor of epidemiology at the University of Washington, says he’d rather see flawed location data integrated into information used by epidemiologists than not. “In the context like the pandemic, it’s problematic to treat selected data as though they’re the whole truth, but it’s also problematic to dismiss data because it’s imperfect,” Mooney says.

When corporate data methods inform public health decisions

Tech firms that typically use mobile location data for business purposes have joined the effort to end COVID-19, from giants such as Facebook to relative unknowns such as location data analytics firms Cuebiq and Unacast. Last week, Google and Apple announced a project to create a Bluetooth-based automated contact tracing system that will work across iOS and Android.

But data experts and ethicists worry that the mad scramble to limit the scourge of COVID-19 is driving governments and public health officials to gather and analyze data without appropriate care or deliberation.

The Kansas Department of Health is using data from Unacast, a mobile location data firm that typically serves the retail and real estate industries. Media outlets across the country have reported on the company’s scores, which give regions a grade from A to F for compliance with social distancing guidance.


“We are using their data to help gauge how effective social distancing is comparative to the number of cases in a community,” says Kristi Zears, director of communications for the Kansas Department of Health and Environment. The agency is not using Unacast data for contact tracing, she says.

But the company has altered its methodology twice since late March, drastically changing its original conclusions that already may have provided a false sense of how different counties were adhering to social distancing.

Despite the fact that social distancing is about reducing contact with people, the sparse population of around 2,500 residents in Hamilton County, Kansas, was not taken into account until Unacast introduced a new metric last week measuring the likelihood of someone encountering another person in a given area. That change converted Hamilton County’s damning “F” social distancing grade to a passing “C” overnight.

Zears acknowledges that people in the county—where agriculture and food processing supports more than half the workforce—may be traveling to produce food. But she says, “We are hoping to see reductions in all communities, knowing that while many communities have many essential functions being performed, they should also experience a reduction in travel outside of those essential functions.”

A Unacast spokesperson says the firm has regularly updated its methodology for gauging social distancing based on feedback from public health experts.


Biased perceptions

Not only are public health officials incorporating these corporate data metrics into their overall analysis of the impact of the coronavirus. Ordinary people are hearing and reading news stories based on their conclusions.

Several national and local media outlets have cited conclusions from Unacast, location analysis company Cuebiq, and others in the hopes of helping us understand the pandemic. Even Google is now supplying new mobile location data analysis to public health officials showing how visits to categories of places such as pharmacies or residential areas have changed in the past two months.

The failure of Google’s once-touted Flu Trends data, which overestimated the flu’s spread, shows us how relying on select corporate data and analysis can skew our perception of public health reality.

“You have to have context experts, people who understand public health,” Raicu says, “not just the people who have data and want to do something useful with it.”

Using data that leaves people out could further erode trust in our medical system among minorities and vulnerable populations. Already, African Americans, a group with a long history of warranted distrust in our medical system, comprise more than 40% of COVID-19 related deaths in the U.S. “If you don’t care about equity when the stakes are high and it’s a matter of life and death, then I think it’s questionable whether or not you really care about equity,” says Mooney, the epidemiologist.


With the pandemic ravaging communities across the country, effective and equitable use of data is paramount. But Ashkan Soltani, former chief technology officer of the Federal Trade Commission and senior adviser in the Obama White House, says the tech industry should be more humble in recognizing its abilities and limitations.

“It’s not just that they have privacy problems. It’s the misrepresentation by industry to say, ‘This is incredibly effective stuff and it will solve the world’s problems,’ when we know it’s probably just good enough for advertising,” Soltani says. “I’m not sure it’s good enough to make decisions about who to send ventilators to or where to put resources.”