Despite ethical and privacy concerns, two Stanford scholars say an app is the most effective option before a vaccine hits the market.
The COVID-19 pandemic is the single greatest threat to global health we face today, yet we have no effective national strategy to deal with it, said Michelle Mello, a Stanford professor of both law and medicine. “Instead we have states either keeping people home or letting everybody out – and neither of those is a sustainable solution.”
A nationwide contact-tracing app offers our next best alternative, Mello said. Such an app would help identify people who have been exposed to COVID-19 so that they can be asked to self-quarantine.
In a recent Science paper, Mello and Stanford medical school colleague C. Jason Wang, a Stanford Institute for Human-Centered AI-affiliated professor, weighed multiple ethical issues raised by contact-tracing apps – including privacy, autonomy, equity, and accountability – and reached the controversial conclusion that an opt-out contact-tracing app is the least restrictive way to rein in the pandemic.
“We think the default should be: You’re in,” Mello said. “We’re all in this together, and that is what is going to be needed if we actually care about vanquishing this disease before we have an effective vaccine.”
Contact-Tracing App Basics
In the United States, rather than make a nationwide decision that a contact-tracing app is or is not a good idea, the CDC issued digital tool development guidelines for states and corporations to follow. Thus far, this approach has produced a handful of apps across the country – mostly in small states such as North Dakota, Utah and Rhode Island, but also recently in Virginia. The CDC’s laissez-faire approach means companies and states are scrambling for answers about how to develop an app that will be broadly adopted, does a reasonably good job of identifying people who’ve been exposed to the virus, and is well integrated into the public health system to ensure appropriate follow-up.
In broad strokes, a contact-tracing app that follows the CDC guidelines would automatically share Bluetooth identification between phones that come within a specified distance of each other (possibly 6 feet) for a specified period of time (possibly 15 to 30 minutes) over the course of a few weeks. If no one tests positive for SARS-CoV-2, this data would be destroyed automatically after three weeks. But when a person does enter a positive test result into the app, the data would be uploaded to the health department, which would reach out to exposed individuals and ask them to self-quarantine.
Digital vs. Manual Contact Tracing
In their analysis of the potential benefits and harms posed by contact-tracing apps and other digital technologies, Mello and Wang apply two important principles: First, that digital surveillance measures should be evaluated in reference to the “counterfactual,” or to the approach that would be used without the technology; and second, that for digital surveillance to be deemed appropriate, it should be judged the least restrictive alternative that can achieve the public health goal.
In the case of contact-tracing apps, the counterfactual approach is manual contact tracing. In this approach, health officials interview people who test positive for a disease to identify their relevant contacts, and then call those people to request that they sequester themselves and to offer empathy, support, and relevant health information.
Manual tracing isn’t a good fit for COVID-19, Mello said. The virus is spread primarily through airborne droplets and about 40% of transmissions occur during a time when people are asymptomatic, so the number of potential contacts can be high. And human recall is problematic: Interviews may bog down, memories may fade. “Even if we had deployed manual tracing effectively in the beginning and had had the testing to support it,” Mello said, “just mathematically and due to the nature of the virus, it’s very unlikely that it would have led to containment.”
So are contact-tracing apps the least restrictive alternative?
To be beneficial, Mello and Wang say, a contact-tracing app must be adopted by 80% or more of smartphone users. The problem is, surveys show that somewhere between 17% and 70% of Americans say they would be likely to use one. Moreover, according to a recent Washington Post article, most countries and states that have implemented opt-in apps have seen fewer than 1 in 5 people adopt them. Mello and Wang therefore conclude that an opt-out app is justified to achieve the public health goal of controlling the spread of COVID-19. In other words, they favor a Bluetooth-based contact-tracing app (with appropriate data destruction guarantees) being automatically installed and activated on all U.S. smartphones. Phone users would be able to delete the app, but the hope would be that, due to inertia, few would do so.
The Ethics of an Opt-Out App
Mello and Wang acknowledge that an opt-out app raises serious privacy, autonomy, accountability, and equity concerns. In particular, implementing an opt-out contact-tracing app in a democratic country such as the United States raises concerns about government surveillance and mission creep: If governments are tracking individuals’ contacts for health reasons, will they gradually become Orwell’s Big Brother, surveilling our every move?
Mello said most of these concerns can be addressed through a well-designed Bluetooth-based app with user agreements that include a sunset clause (i.e., the app is automatically deleted from phones when the pandemic is over) and contractually require data destruction at a particular point in time.
“I take seriously the fact that this is invasive of privacy,” Mello said. “But my view has always been, if you’re going to do it, let’s design it in a way that’s actually going to be useful; and let’s design the traditional public health systems around it in a way that they can act on the information.”
Mello and Wang recognize that an opt-out contact-tracing app also raises questions of equity and fairness. That is, the same individuals who don’t have smartphones and broadband for running the app may also belong to communities disproportionately impacted by COVID-19.
To address this problem, they propose directing more manual contact-tracing resources to these communities. For example, if 80% of people have smartphones, “we could rely on the app for them and put a lot more of our human resources into the remaining 20%,” Mello said. Other options, such as providing smartphones to people on the condition that they use the app, could work as well, she said. “If we did have a technological solution that works, I think we would want to make it available to everyone, and I would hope that major tech companies would play a role in making that financially possible.”
Possibly the biggest and most frustrating barrier to implementing a contact-tracing app is political, Mello said. “The response needs to be nationally led, and it’s not going to be.”
Stanford HAI's mission is to advance AI research, education, policy and practice to improve the human condition. Learn more.