I was kind of fed up with the fire hose of Covid news items. Then the worship of Covid apps as the way to redemption came along, and, well, I had to write something. I’ve been developing software for over twenty years, and guess what, I fear Covid Apps are a pipe dream.
What are Covid Apps?
In another couple of months — sooner in the US, if Trump gets his way — you might be riding a bus again, or sitting behind a desk. Even if from a distance, you will be coming into contact with more people again. Unknowingly, you might be beside somebody contaminated with Covid-19. They will already be spreading the virus, even if they don’t have symptoms yet. Oh no, how are you going to find out that you might have been infected?
Luckily, there will be an app for that. It tells you that you’ve been into contact with somebody when they test positive. That way, you can quarantine yourself for two weeks, and everybody will be safe. The end.
Such apps are based on either GPS or Bluetooth technology. Using GPS means it keeps a log of where the phone has been. If somebody tests positive, their GPS history can be compared to others using the app, and that way you can see if you’ve been close to somebody contaminated.
The Bluetooth version works differently, using proximity connections. Each phone has a unique ID. The app records if your phone has been into close proximity of another phone for a set amount of consecutive time. When a phone owner tests positive, all the apps that have a record of contact with the corresponding phone ID notify their users.
The theory is nice, but there are some flaws. Let’s put the privacy issues aside for now (we shouldn’t). Oh, and let’s make the app mandatory (1984 style). Even then there are still a lot of problems.
First of all, not everybody owns a smart phone. The Netherlands is high up in the charts, but even here ‘only’ 90% of people use a smartphone. The other 10% of people have no smartphone, hence are not measured, even if the app is mandatory. Oh, and let’s not forget empty batteries, a problem that will increase significantly if you start turning on Bluetooth or GPS all the time.
GPS is probably pretty useless in any case. It has an accuracy of a few meters, which is the difference between being safely far away and on top of somebody. You could be on the top level of a bus while the contaminated person is a floor below you. And it works for crap in buildings.
For Bluetooth, there’s the connection issues. Everybody who owns a Bluetooth mouse will have experienced connection issues at one point or another. Crowded areas — like trains and supermarkets — have more interference, even as they are the most important for this to work.
And of course, not all smartphones are equal. I have issues connecting my iPhone to my car, which I didn’t have with my old Android phone. A colleague of mine has an app using Bluetooth beacons, which occasionally reports him next to the beacon when he’s, in fact, miles away. This could easily mean a lot of false negatives and a lot of false positives.
Don’t beleive me
What’s the success ratio then?
The success ratio? Short answer: nobody has a freaking clue. You see, this type of pandemic hasn’t happened for over a century. We still know very little about the disease and how exactly it spreads, and smartphones were not around last century.
We’re testing, but not nearly enough for an app. Not in the US, and especially not in the Netherlands. There are a lot of patients with mild or no symptoms who could be contagious, who could account for three-quarters of the contaminations. That’s a high number.
The next problem is that contagion through close proximity is one vector of contagion. However, doorknobs, elevator buttons, and who knows what else are also vectors. Sick person touches nose, then doorknob, you touch doorknob, then face, and bang: you’re infected. So, John from HR on the other side of the building could walk onto your floor to drop of a form, never even register on your app, and still infect you through the doorknob. How high a percentage of contaminations will that be? Fifty percent? Eighty? Nobody has a clue, especially given we’ve already implemented social distancing measures.
Oh, and you know, six feet (1.5 meters) distance is nice in theory, but an air conditioner cycling air could increase that. Yep, that hip restaurant with bad bluetooth signal just missed twenty infections in one night. And walking into a cold dry room after an infected person leaves could spell trouble as well.
If you take all these numbers, and combine them, you get a new number. That number is the chance you get a message on your app when you have in fact been exposed to Covid. It could be anywhere between 0% and 100%, but my guess is somewhere in the lower digits. Keep in mind: two of those numbers of 50% combine to 25%, and here we’re combining three or four that might well be in that range or worse.
And false positives?
Then there’s the chance you get a message when there’s actually no chance you were contaminated. I think that chance is higher than we want it. Your neighbor could trigger it, while separated from you by a wall. That person on another level of the bus or train. You live in an apartment building? You’re screwed, mate. Oh, and do cars isolate enough that I won’t be linked to somebody one car over in a traffic jam?
I won’t hazard to guess about this number either, but my gut says this will be pretty abysmal as well.
Even so, if it helps, even a little…
What if this app could help! Only a little. It doesn’t harm, right? Well, we’ve already brushed aside the privacy issues. When will the police want access? I wonder. But okay, let’s still assume that’s worth it. There’s more.
The combination of high false positives and low success ratio means there’s a good chance the app will have roughly the same success as random noise. So it would be just randomly selecting people for quarantining.
Keep in mind, we have this app specifically so we can isolate people. We’re going to ask people to stay home when they might have been contaminated. So, there’s a good chance a Covid app doesn’t work at all and is basically the same as a quarantine lottery. Congratulations, sir, you’ve won a no-expenses-paid mandatory two-week vacation… to your five-by-five apartment.
And, like in many cases, the app will hit people harder lower on the social ladder. Those stuck in cramped apartment buildings with thin walls are more likely to get digitally exposed than the rich living on their large estates.
Of course, in the Netherlands, it will probably all be voluntary. Aside from that further undermining its effectiveness, I doubt that will actually work that way. Day one that an app launches I fear my daughter’s school might send a message ‘we know it’s voluntary, but we ask you stringently to install the app and keep your child home if you get a message that you have been exposed’. My employer will possibly do the same.
So, employers, schools, and who knows who else will pressure people to follow an app for which the effectiveness might be equal to random noise. The repercussions are very real, of course. Being quarantined is no picnic, and if you have no income, you’re pretty screwed, or it could put your child even further behind in their education. Oh, and let’s not forget about domestic violence.
Covid apps are dangerous. They have very real consequences for all of us, while nobody knows how effective they are.
Don’t get me wrong, if it helps, I’ll use it, and it could be a good tool, but for now, there is little evidence to support that notion. And as long as that is the case we may not want to open this Pandora’s box.
Yes, we need to get our society up and running again, but maybe not by randomly sacrificing people on the quarantine altar.