Countries Rolling Out Coronavirus Tracking Apps Show Why They Can’t Work
If you think tracking apps will keep people safe as economies reopen, look to South Korea, Singapore, and Australia to see why you’re mistaken.
If there’s one thing we all wish our phones could do, it’s protect us from coronavirus. After all, they seem to be able to do just about everything else. Over the past three months, they have replaced schools, movie theaters, and family get-togethers. Once this is all over, who will even want to go back to the legacy reality of holding business meetings in conference rooms when we can use Zoom instead? And why go to a germ-infested doctor’s office when smartphone-based telemedicine can diagnose and treat many ordinary ailments perfectly well?
But the coronavirus is no ordinary ailment, and despite the techno-optimists’ best efforts, smartphone apps will not get us out of this crisis. Public health authorities in many countries have pinned their hopes on the development of coronavirus tracking apps, which in theory could allow governments to quickly identify and notify people who have come into contact with infected individuals. Those who have been exposed to the virus could then be tested and isolated, potentially breaking the chain of further transmission.
As the United States debates the merits of turning telephones into personal trackers, other countries have already moved forward. South Korea early on created a public database of coronavirus cases that provides extraordinarily detailed information about every infected individual, including their exact movements around the country. The database is constantly updated using location information from payment card transactions, mobile phone signal data, and closed-circuit TV footage. This government database made possible the development of apps such as Corona 100m, which maps infected individuals and provides information about their age, sex, and nationality. It’s hard to imagine many Western countries taking such a cavalier approach to patient privacy, but the app seems to be popular in South Korea.
Singapore’s TraceTogether, released on March 20, is a more limited coronavirus tracking app that keeps a record of close contacts between the phones of registered individuals without registering where those interactions occurred. Unlike South Korea’s database, it is optional, and in fact only about 25 percent of the population has signed up. It uses a phone’s Bluetooth function to detect and log every instance when two people (or at least their phones) come close enough to each other for Bluetooth to recognize the signal.
[Mapping the Coronavirus Outbreak: Get daily updates on the pandemic.]
It is Singapore’s app, not the South Korean ones, that has become the model for other countries. This includes Australia’s COVIDSafe, which was rolled out on April 26. Although take-up of Singapore’s TraceTogether has been poor, Australians have flocked to COVIDSafe, with more than 4 million downloads in the first week—even though the underlying government monitoring system is not yet operational. When it is, users who test positive for coronavirus will be asked (not ordered) to let the government warn others that they have encountered someone with the virus, without telling those notified who the infected individual was. It is unclear how much of Singapore’s source code was built into Australia’s app, but analysts note a few tweaks to the Bluetooth implementation and data encryption.
The relative success in the take-up of the Australian app even before it is functional might have something to do with the fact that the Australian government is effectively holding the country’s economy ransom to COVIDSafe. Prime Minister Scott Morrison has made the widespread adoption of COVIDSafe the key condition for reopening offices, restaurants, and other amenities. How widespread would be widespread enough? The government won’t say. The requirement comes despite Australia’s relatively benign experience with the virus—there are currently only two dozen or fewer new infections a day—and the fact that more than 90 percent of all known cases can already be traced back to an identified source.
If Australia won’t be specific about how widespread app use should be, that’s probably because it doesn’t know. Tracking apps are an unproven solution that rely heavily on “if only” reasoning: If only everyone would install the app, if only people would carry their phones at all times, if only phones wouldn’t run out of power, if only people wouldn’t turn off Bluetooth, if only tracking apps didn’t interfere with other software. Finally and most importantly: if only the virus would spread in ways that can be detected using the app.
Scientists are still unsure how the coronavirus spreads, but the firm consensus is that most people catch the disease either by being in close proximity to a cough or sneeze, or by picking it up from a surface. The first could happen in an instant, say in a supermarket aisle, while the second could occur hours later when you unpack your groceries. That’s why health workers wear face masks and wash their hands: They have to protect against both direct and indirect transmission.
Neither transmission route fits the logic of tracking apps. When a person tests positive for the coronavirus, tracking apps notify other people who have been near the infected person in recent weeks. Singapore’s tracking app is supposed to notify all people who have been within 2 meters of an infected person for at least 30 minutes, while Australia’s app claims to notify people who have been within 1.5 meters for at least 15 minutes. Since Bluetooth can’t actually be used reliably to measure distances, these figures suggest an illusory precision.
The inconsistency between what the apps measure and how the virus spreads puts governments in a bind. Set the time window too narrow, and the app will classify millions of people as possibly infected, requiring the government to track down everyone who has ever passed a coronavirus carrier on the street. Set the time window too wide, and the app will flag too few exposures to the virus. There is no “Goldilocks” zone in the middle of these two extremes.
Set the threshold at 15 or 20 minutes of close proximity to an infected person, and the coronavirus app will identify a moderate number of people for health authorities to contact. But most of the people who have contracted the disease from the infected person casually—that supermarket sneeze comes to mind—will be missed. The other problem is that actual transmission events are rare compared to the number of interactions people have. To find those transmissions, you have to wade through an enormous number of casual contacts, and that means tracking down virtually everyone. Once governments reach that point, they’re no better off than if they had simply relied on effective but labor-intensive human contact tracing in the first place, without the app.
Unfortunately, the coronavirus simply does not spread in the way that epidemiologists model it statistically. That’s not necessarily a problem—for our understanding of broader progress of the pandemic. But the models usually simplify reality in ways that work on average, even if they don’t apply to any particular case. We’re therefore making a grave methodological error when we expect reality at the level of individual cases to reflect our models for the numbers overall, instead of the other way around. Any technological solution to the coronavirus pandemic has to be grounded in the reality of one-on-one transmission, not epidemiological statistics.
The only way to make coronavirus tracking apps really work is to accept the burden of false alarms and track every person, all the time, everywhere. This is only possible in a totalitarian state such as China, yet even the Chinese government doesn’t have the resources to implement such extreme levels of monitoring. With advances in artificial intelligence, it is possible that it soon will—and when that day comes, China will be prepared for the next pandemic. The rest of the world will probably take the virus over the cure.