In the early days of the coronavirus pandemic, many people were excited by the idea that technology could help us track and contain the virus’s spread. Singapore was one of the first to release a digital contact tracing app. Countries from Iceland to Australia soon followed, while Apple and Google built a platform that allows their phones to be used for contact tracing. The apps track who you’ve come into close contact with, using either GPS or a “digital handshake” between phones over Bluetooth. If a user gets Covid-19, they alert the app, and anyone who has been near them in recent days is notified that they may have been exposed.

So far, contact tracing apps don’t seem to be doing much to stop the spread of the virus. Researchers at Oxford University estimated that 60 percent of the population would need to use a contact tracing app for it to be effective, and no country has come close to that threshold. Does this mean we should give up on these apps?



Kimon Drakopoulos is an assistant professor of data sciences and operations at USC Marshall School of Business, where he researches complex networked systems. He completed his PhD at MIT. He is currently working on a project with the Greek government that combines real-time data and machine learning to identify high-risk tourists and target coronavirus testing resources.

As a data scientist who studies how contagions like the flu spread across networks and how epidemics can be controlled, I believe we’ve been thinking about contact tracing apps all wrong. Telling individuals if they may have been exposed to the virus is important, of course. But the larger value of these apps lies in the real-time data they can provide decisionmakers about people’s behavior, revealing the bigger picture of how many potential exposures are happening in a community every day and where they are occurring.

We don’t just want an app to tell us, days later, if we’ve encountered someone with Covid-19. We want to know how many people each of us encounters over the course of a day—10? 50? 100?—any of whom could have the virus. That one simple data point—the average number of interactions (with anyone, not just those who have tested positive), and therefore potential exposures, a person has per day—can help our leaders make smarter decisions about when and how to reopen.

Network science teaches us that fighting an epidemic is like battling a fire. You have to respond quickly and contain the fire at its edges so it doesn’t spread. If we eased up and left the scene once we thought we had a fire under control, but didn’t know for weeks if it had started spreading again, by the time we had our answers it would be far too late. We need to watch the fire in real time and target our resources to the border areas where it’s about to spill over.

Right now decisionmakers are flying blind as they try to fight this fire. When communities lift lockdowns, they’re basically conducting experiments that take weeks to deliver results, potentially losing lives in the process. States like Florida, Texas, and Arizona started reopening in April and May but didn’t see the fallout from their decisions until June and July. The virus was escalating on the ground for a long time before it began to show up in official case counts, hospitalizations, and now deaths.

How can other states avoid becoming the next Florida? They can use contact tracing apps to gather real-time data about which activities or locations might be responsible for a dangerously high number of potential exposures. Say your state reopens restaurants for indoor dining. Does this double the number of interactions the average person has each day, or increase it by 20? The answer to that question makes a huge difference in deciding whether reopening restaurants is the right move, but currently we have no way of knowing. These apps can help us track the real-time consequences of policy decisions about when and how to reopen schools, parks, shops, offices, and other spaces.

Data from contact tracing apps can also help us better target interventions by revealing where exposures are happening (for privacy reasons, this would be at a neighborhood, block, or zip code level, not individual locations). If a specific public beach or park is a hot spot for interactions, for example, maybe it should be closed while other less-trafficked spots remain open. If the average resident in Town A has three face-to-face interactions a day while the average person in Town B has 25, depending on the local infection rates, maybe it’s time to open up Town A while directing additional testing resources to Town B. This helps us move from blanket state- or citywide lockdowns to a more targeted response.