The government’s answer to containing the coronavirus pandemic may well lie in the pockets of its citizens. Smartphones have become so ubiquitous that nearly all of us cannot imagine leaving home without them. How many of us have gone sprinting back to our front doors with the panic-stricken thought: “I forgot my phone!”
Today’s smartphones all have built-in GPS, which means they are not only versatile communication gadgets but also perfect tracking devices. By analyzing the smartphone logs of people’s movements through space and time, governments can identify gatherings, trace contacts with infected individuals, and detect guideline violations, from curfew infractions to improper social distancing to non-essential work travel.
Big Brother Concerns
Before discussing how smartphone tracking can be used to combat COVID-19, we must address the broader issue of mass monitoring. When we call a friend, use an app, or browse a website, we expect not to be snooped on. When we take a walk or ride a bike, we expect not to be followed.
With a virus, our rights to privacy in where we go and what we do outside our homes are in potential opposition with the public’s rights to be protected against infection. What’s more important? An infected individual’s right to privacy? Or the safety of the people sharing his commute?
While many would say that the security of the many is more important than the privacy of the few, we must also recognize that any reduction in personal rights is a slippery slope. Historically, governments have used wars and emergencies to curtail individual freedoms and, often, when the crises passed, those freedoms were not fully restored.
We must not allow the pandemic to destroy our way of life. For this reason, any reduction in our privacy must be limited, transparent, and temporary. The recommendations made here are intended to be a timeboxed solution to a public danger and, if adopted, set to expire within months — not years.
Compliance is Key
There are two general approaches available for slowing the spread of COVID-19. The first is to focus on known infections, quarantining confirmed cases and their recent contacts. The second is to focus on general population safety, with the understanding that many infections cannot be known and traced. The first approach is reactive, since we are literally reacting to a newly infected patient, and the second approach is proactive since we are preventing infections before they can occur.
The two approaches are not mutually exclusive but rather complementary. If we were dealing with a few hundred, or even a few thousand people, we might be able to implement these approaches through traditional methods of awareness, intervention, quarantine, and social norms. However, at the scale of billions of people globally and tens of millions in major countries, we require a more rigorous implementation.
Both the reactive and the proactive approaches have something in common — they are based on the physical distance between people. For a confirmed case, we try to put distance between the patient or the people the patient may have infected and the rest of the population. For preventive measures, we seek to maintain distance between each person at all times.
All the teleworking, shelter-in-place, self-isolation, and quarantine measures enacted across the globe and encompassing billions of people are only as good as their practice. What is the point of bringing social and economic activity to a standstill if people continue to congregate and swap germs? If we are going to accept such draconian measures, we must have a way of enforcing compliance. Otherwise, a few bad apples will continually spoil every bunch.
Data That Matters
Experts agree that if every person maintained a safe distance from every other person at all times, the spread of COVID-19 would be dramatically reduced if not stopped altogether. In theory, if we could see and analyze every person’s location in real-time, we could identify nearly every instance of noncompliance: every time a person got too close to someone else, or traveled to their office as part of a non-essential job, or went to a public location with too much population density.
National mobile carriers have access to real-time GPS information for virtually every cell phone on their network. By aggregating this data from the carriers operating on their soil, governments can have a running log of the movements of every person with a cell phone within their borders.
By combining this location data with business, employment, residential, and mass transit information, governments can determine the nature of a person’s movements, including when they leave home, where they go, and how they get there.
As noted, this data is extremely sensitive and, if not used strictly for the purposes of enforcing social distancing regulations during this time of emergency, can be subject to privacy abuses on a mass scale. With that said, governments alone are vested with the authority and power to access and use this data appropriately. In a time of national crisis, we must be prepared to forego conveniences and even liberties in order to empower our governments to look after the greater good — subject to appropriate checks and balances.
With a full complement of aggregated data, governments can employ data scientists to develop algorithms that will identify patterns in real-time. Even without supporting information, the location data tells a story. For example, the location where a person spends the night hours is likely their home, the location where a person spends the day hours is likely their work. Overlaying map-based data can identify other locations where a person spends their time. Any indoor location that is not considered essential may represent noncompliance with the local shelter-in-place order. Similarly, overlaying employment data can indicate if a person who is not part of an essential business is continuing to commute to their office.
Combining data across individuals can indicate hotspots, meaning both indoor and outdoor locations with unsafely high density. It can also provide information about overcrowded transit lines and times. When the location data is sufficiently precise, it can even flag people who are likely to be standing closer than six feet apart.
Incorporating information on building types and sizes can help further stratify the data, determining whether a high-density location is a residential high-rise with people vertically separated in apartments on different floors, which is safe, or a commercial event venue with people all next to each other, which is not.
Given the advances in cloud computing and big data over the last decade, it is possible to capture and analyze data across the entire population in real-time. With the right algorithms, likely noncompliance can be flagged within seconds. And with machine learning approaches, data scientists do not have to come up with their own algorithms, they can have computers discover the most telling patterns.
In addition to flagging noncompliance with regulations, location data provides the key to tracing transmission and identifying individuals who are likely to have been exposed to the virus. By starting with a new confirmed case and working backward we can find all the individuals whose location data overlapped with those of the confirmed case over the preceding weeks. Each such individual represents a suspected case.
An algorithmic analysis of the data can be used to automatically stratify suspected cases as high, moderate, or low risk based on proximity to the confirmed case and the length of time of the exposure. A person who passed by the confirmed case on the street may be considered low risk, a person who shared a transit commute may be considered moderate risk, and a person who sat in the next cubicle for hours at the same office would certainly be considered high risk.
Once the high-risk cases have been identified, it becomes possible to inform them that they have been exposed and direct them to a self-quarantine and testing protocol. This entire process can be automated, with standard SMS messages containing all the necessary information and links auto-delivered to the smartphones of high-risk cases. The messages can include automatically scheduled appointments at local testing centers and require a “YES/NO” confirmation.
The high-risk cases can then be tracked more closely to ensure self-quarantine compliance until testing confirms whether they have the virus. Additional safeguards can be put in place such as automatically mailing high-risk cases a package containing face masks and gloves to minimize their danger to others in their households and communities.
The ability to track noncompliance and contacts is only as useful as the ability to enforce compliance and testing. In turn, transforming data into action requires careful consideration of the accuracy of the data and the appropriateness of the action. Algorithms can flag likely noncompliance and suspected cases, but these flags represent probabilities, not certainties.
Because we are dealing with people’s freedoms, we must exercise extreme care in how we enforce on the basis of probabilistic analysis. One approach could be an enforcement regime based on intensity tiers that are tied to the quality of information and the nature of the response. Stated more simply, the more evidence we have that someone is knowingly not complying with local regulations, the more pressure we should apply to force behavior change.
An illustrative framework for an escalating enforcement regime is SMS messages, followed by calls, followed by citations and fines, followed by visits from law enforcement officers. For example, a suspected case can receive a text message on their smartphone directing them to proceed to a testing facility on a specified date and time, and to remain in self-quarantine until then. If they continue using public transit, they receive a follow-up text message directing them to remain in self-quarantine or face fines. If they still ignore the direction, they receive a series of phone calls from a government or law enforcement agency asking them to comply. The escalation continues to fines and, potentially, a visit by officers of the law.
The vast majority of people would comply in the early stages of such a regime, based on automated SMS messages or calls. As a result, governments can drive a significant level of compliance with the same instrument used to detect noncompliance: the smartphone.
The approach outlined here requires iterative testing locally before it can be considered for large-scale rollout. Collecting data, developing algorithms, identifying patterns, tracking actions, and automating enforcement requires the implementation of new software capabilities and operating models that must then be field-tested and refined to ensure they are effective in practice.
However, the key pieces, including all the necessary data, already exist in government, technology companies, and telecommunication provider databases. Consequently, a cross-functional task force comprised of data scientists, clinicians, network and systems specialists, and software engineers can likely build a working prototype within weeks.
The prototype can be piloted in a specific geographic region. In addition to refining the technology, a pilot would allow healthcare organizations, mass transit authorities, and law enforcement to explore various approaches to coordinating response and to identify best practices. Importantly, the effort must be focused on managing confirmed and suspected cases. Government agencies should be directed to use this capability specifically for combating COVID-19 and not any broader mandate to avoid any possible misuse. This level of mass surveillance opens the door to a police state and extreme vigilance must be exercised to prevent stepping through that door.
Critically, a geographically limited proof-of-concept would demonstrate whether the approach is effective in reducing rates of transmission and infection. If the piloted region shows a statistically slower spread, then we can have confidence in rolling out nationally.
We know what we have to do to combat COVID-19, we have the data to determine if we are doing it, and we have law enforcement and military personnel to enforce what has to be done. All that we are missing is a centralized command-and-control that combines these elements into a working monitoring and compliance system.
What about people who either do not have a cell phone or who leave it at home? It is true that this approach would not account for these individuals. However, it would encompass a vast majority of people. More than 90% of Americans carry a cell phone and more than 80% carry a smartphone. Nearly 70% of the people around the world have a cell phone. A person would have to be highly motivated transgressor to leave their phone at home for the sole reason of not being cited for unsafe behavior toward the health of others.
Ultimately, the goal of the proposed program is to support the vast majority of people who struggle to comply with shelter-at-home regimes and preventive testing requirements due to a lack of information or direction. If this system gets most people to behave more safely and more consistently, it will give us the best possible chance of bringing COVID-19 under control while we wait for the development of treatments and vaccines.
Using smartphones to track movement need not be an overreaching invasion of privacy if it is limited in scope. The objectives of this program are to verify that individuals are following COVID-19 regulations and to flag noncompliance. Since noncompliance would require transit through public spaces, that would be the focus of the surveillance. Arguably, this would be less invasive than the footage from the millions of CCTV cameras that dot our cityscapes and record our activities or the tracking data from the web beacons and cookies that cover our websites and monitor our browsing. But unlike those other types of recording, this has the potential to prevent tens of thousands of infections and save thousands of lives. Subject to appropriate oversight and expiration provisions, a phone-based approach may get us closer to the public safety required in these extraordinary circumstances.