Dogs are being trained to sniff out COVID-19, and they're better than the tests

Originally published at:


The problem is there are not enough dogs in the world to meet the overall need; this is a great application of this ability. Many years ago a family friend who has no use of her legs due to a congenital cause had a baby; when she brought the newborn over for a visit our basset hound gave the baby a thorough sniffing inspection, focusing on the legs. After the inspection, he started to celebrate, clearly signaling his conclusions…dog noses can be pretty remarkable .


I’m seriously worried that those will be used the in same way drug sniffing dogs are used - to harm people disliked by authorities. For example, your skin pigmentation is not light enough, or you don’t conform to gender stereotypes, so they cue a dog to “detect” coronavirus, and you’re treated as “heavily infected” now.


That claim of “better than tests” is highly misleading, as it only relates to sensitivity. One of the links says:

and correctly identified Covid-negative samples 96% of the time.

With the current low prevalence of Covid-19 you would need another quick test to rule out a substantial number of false positives, though. If we replaced tests with dogs in Germany, with that kind of specificity we’d currently have around 4 times more false positives than actually infected people. PCR tests can have a specificity higher than 99%.

Translate that to a plane with 100 people. If on average only one passenger has Covid-19, they’d sniff out 4 more who actually don’t.


But because COVID isn’t a crime, if a lot of false COVID arises, and start to be correlated of melanine content of the skin a PhD that need to publish something will start to investigate the correlation making the test with dogs less reliable and dismissed.

Be extremely suspicious of all such claims of the magical abilities of dogs to smell all our problems away. Yes their noses are amazing, but any time that bomb or drug sniffing dogs are tested in properly controlled and double-blinded conditions, they do very poorly. It’s been shown over and over again to be Clever Hans effect when these dogs are deployed in the real world.


Not if dogs will be used by TSA. And people selected by the current “leader of free world” will make sure that data is not avaliable, for example using the excuse of natonal security.
That COVID is not a crime doesn’t mean that it can’t be used as another way to victimize people. It was already used in anti-immigrant hate speech by you know who.

1 Like

The problem is that various kinds of sniffing dogs, from drug dogs and bomb dogs, to those corpse sniffing dogs that can supposedly locate traces of bodies long after they’ve been removed. Have been proven to be less reliable than chance. Most of it is a figment created by the dogs responding to unconscious cues from the handler. There’s a very limited number of things dogs can actually be used to track and I highly doubt that infection with a virus is one of them.

It’s in large part not a deliberate thing. Drug and bomb sniffing dogs are pretty much only responding to such cues from the handler. All of the time. As these cues are unconscious on the part of the handler (like ideomotor response with pendulums and Ouija boards) they’re ruled and dictated by the biases and expectations of the handler. Leading to documented racial skews in who they “catch” and where.

Though of course it’s easy enough to just make the dog signal or claim they did if you have an agenda.


If the handler is strongly prejudiced against some ethnicities, he may give such cue completely unconsciously. There is probably less hate rhetoric in Finland, but how the handlers in USA may not be prejudiced when their president says such things:
I can’t find the quotes now but he said way more things like this.
It doesn’t really matter if the prejudice is conscious or unconscious at this point (it’s probably both IMO).

I think that dogs and cats can sense if a person is somewhat ill, so maybe it’s like using an IR camera to find people with fever. I had a cat that when I was ill and in bed came in the bedroom, jumped on a chair and started watching me, a thing that she didn’t normally do.

What lizard-of-oz said. This is a Bayes issue. With a relatively small percentage of the population infected, the screening method needs to be on par with the real incidence rate–or else you get swamped with so many false positives, you might as well not run the screen at all.

Peer owlishly at the nomogram! Gar Ming Chan - Bayes’ theorem, COVID19, and screening tests.


You don’t have to be strongly biased to send such cues. Everyone internalizes societal bias to some extent and the cues in question can be as subtle as the handler being slightly more tense in one situation than another. Or a slightly more enthusiastic reward conditioning the dog to signal in response to certain people or objects. If this was a factor of conscious effort, why would it turn up in controlled testing? With and without trained handlers? Whether we’re searching for lead paint in apartment buildings, or drugs in an airport?

I’m also pretty sure that this has been found universally. A lot of the research showing the bias and how ineffective such dogs are is European. And it appears as often in non police contexts.

I dunno that I’d go with less hate rhetoric in Finland either. While the context is different the European right is similarly ascendant, built on a baseline of anti-immigrant and anti-muslim populism. And Northern Europe, including the Nordic countries, have huge societal biases against immigrants, Africans and Muslims.

It’s a bit like the assumption that racism in the US is relegated to or the fault of the South or red states. Despite Long Island and New Jersey being our most segregated regions. The Midwest inventing sundown towns, and Oregon being founded as a white supremacist homeland that excluded all blacks.

Dogs read people. Either the handler or in the case of something like a service dog their owner.

To the extent that they notice some one is slightly ill, it’s the same way we do. The person or their behavior is slightly off. I’m not really up on whether seizure dogs and the like are valid. But this is the theory behind how it works. The dog is noticing physical and behavioral cues that signal an oncoming seizure, and is conditioned to give an obvious response so that those signals can not be brushed off or go unnoticed.

That is a very, very different situation than a detector dog though. Such dogs can’t detect a seizure in any random person, the entire concept is that it is that persons dog and the dog is intimately familiar with the patient. There’s no way a dog can be intimately familiar with the healthy baseline of every person it encounters. Without that we’re basically positing that COVID makes you smell distinctive enough that a dog can smell it, and can be conditioned to sniff it out. You’re gonna need to prove that before you can train a dog to distinguish it from anything else. Otherwise dogs are magic.

When there are consequences like missing your flight and being quarantined, they are worse than ineffective, they are perfect tools of harassment.

What about people who are somewhat ill all the time due to failing immune system, not to the point of having fever, but enough to trigger such dogs? Being falsely accused of being infected all the time would suck.


The dogs in Helsinki airport are kept in a separate space from the traveller. They only sniff swabs taken from the person, who must wait until the dog has given it’s verdict.
If you wanted to create a system to target certain groups with false positives there are easier ways to do it.


Not just dogs apparently

1 Like

Unless they’re double blinding it there’s still opportunity for the dog to take cues, conscious or unconscious, from staff.

There’s also still the problem of what are they smelling, if anything. Is it unique to covid, or in anyway directly tied to covid. Absent anything to smell the dogs are only responding to behavior of staff, and all sorts of weird creeps in that way.

1 Like

OK, but as soon as the process is sold to LabCorp and Quest, it will STILL cost $119 to be tested by a dog. And then they’ll withhold the results for 48 hours and make you download an app to see them.

Capitalism in America always defeats the public good.

1 Like

Then you do a traditional test on those five, which is still a lot better than having to test all of the passengers.

"Would you like to be tested using traditional methods, or with floof?

“Floof, please.”


But that doesn’t help you with all the people who are in fact SARS-CoV-2 positive that the dog missed, and who will merrily go about spreading the virus once they’ve exited the airport.