Link of the day: Contact-tracing apps are not a solution to the COVID-19 crisis
Update: I recently also wrote an article in favor of contact tracing apps as a response to this one.
A lot of people put a lot of hope into contact tracing apps. I think these hopes, unfortunately, are somewhat misguided. There are a couple of categories of arguments against these apps:
False Positives – what is a contact?
Most current proposals are based on Bluetooth. There are a couple of technical reasons why Bluetooth is not really good at estimating distances. Two of them:
- Bluetooth chips don’t send with the same strength in all directions. And there is usually only one chip per smartphone. It is therefore very hard to calculate distances based on signal strength
- Signals can be reflected and bounce off or penetrate walls
Also “distance” in itself is not really meaningful:
- What should the app do if two people are 2m apart, but separated by a wall?
- How should the app now whether you are next to a person outside or inside, which might make a huge difference for infections?
- nobody knows the ‘correct’ distance or time spent near someone infected that should trigger an alarm. Sometimes infections do happen in brief encounters, sometimes they do not happen even if two people spend more than a couple of minutes next to each other.
What does that mean? We will either either be too cautious or not cautious enough, potentially both at the same time. This is bad and means an app can be potentially almost useless just for that reason. Think about it: What are you supposed to do, once you get an alert after you walked by an infected person on the street?
- if people self-isolate based on alerts, a large number of false positives means people will spend a lot of time pointlessly self-isolating. This is bad especially for people who don’t have a lot of social or financial support
- A large number of false positives means that people will trust the app less and less. Nobody is willing to self-isolate for the third time in a row.
False Negatives – you can never be certain
You could make an app that only notifies you if it is really really certain you are in danger. But there are two issues: Firstly, you can never be really certain, because we don’t know what a good threshold is and the app does not about the nature of your contacts. Any case that is really really clear, like a spouse, a coworker, a close friend or partner, is a case that would as well (and better) be found by normal contact tracing. If someone is infected, you call their contacts and tell them to isolate depending on the nature of your contact.
Projecting that kind of certainty through an app would be even worse: Telling people to rely on the app for their safety means that people can start to become careless, thinking they did not get an alert. But: the number of infections missed by the app, will probably be large.
- many, especially the elderly, don’t have smartphones
- if the app is not compulsory, it will be hard to get more than say 30% of the population to install it
- Many infections happen at times where you don’t have a smartphone with you. Think about you just quickly getting some groceries or the delivery person leaving their phone in the car for a second.
- Transmissions based on surfaces are missed
Having people rely on the result of the app could actually be even more devastating.
There are quite a lot of other arguments against an app like this, ranging from concerns about privacy to concerns about the voluntary nature of users’ consent. An app could become de facto mandatory for people who work or use public transport. Forcing them to comply and potentially self-isolate based on potentially spurious evidence. This again, probably hits the weakest in our society hardest.
What else to do?
A contact tracing app will likely cost dozens or hundreds of millions of Euros. An obvious alternative investment would be to ramp up traditional contact tracing: just have people call all the close contacts of anyone who is infected. These contacts seem to play a much larger role in driving infections anyway, compared to random encounters on the street.
There is only one use case, where I can imagine a contact tracing app to become handy: Public transport. But surely, there must be a better alternative to finding out who rode a bus at the same time than an overly complex and erratic contact tracing app.
Not arguing that those apps would end this pandemic right away, but I am not sure, if you are not underestimating the effect: you could test those “flagged” by the app a couple of times in a few sensible intervals and asking these people to avoid unnecessary contacts for a few days, e.g. a week or so, before they get the result. If direct tests are cheap and available this seems to be a plausible way to reduce unnecessary infections because people had no idea they could be infected.
In this scenario it wouldn’t be terrible to have some false positives, because there is a more specific test readily available and those ‘false negatives’ would have not known either way and you would notice the infection at least when they become (very) sick, just as it is today.
Per cost of the app, vs ‘traditional contact tracing’, I am not sure if a) auch an app could inform traditional contact tracing and b) if employing a bunch of people to call every contact is not more expensive if ~30% of potential contacts are not informed automatically.