Researchers have successfully trained mice to detect landmines, sniff out tuberculosis and even drive cars, but their next challenge – finding survivors in a collapsed building – could be their bravest yet. We talked with Dr. Donna Kean, a behavioral researcher at APOPO, an organization at the forefront of training rats to save lives.
They can contribute something that other technologies cannot, at least in the areas we train them in. Their sense of smell and their ability to train are on a par with dogs. But it’s the mice’s small size that really makes the difference.
At APOPO, we work with the giant African mad rat [Cricetomys ansorgei]. We teach rats to detect landmines, because they’re too small, too light, to detonate a landmine. We are teaching them to detect the scent of wild animals illegally smuggled into ports of shipment, because they can reach containers stacked on top of each other.
For my search and rescue research training rats, the main reason is that they can get into the small, tight areas of a wreckage site. Search and rescue dogs normally roam around rubble sites, while we wait for the rats to actually get in, through all the rubble, because they are so small.
Any application has to be in response to a humanitarian challenge and needs the unique capabilities of our mice to help. if it already exists [other] technology available and accessible, we are not going to train our mice just for fun.
How do you train a mouse to do these things?
We use positive reinforcement to train them in a basic sequence of behavior. So here’s to looking for a human, indicating to us that they found him, and then returning to where they were released.
[Training] starts in a really basic setting: a small, empty room. So we gradually expand and increase the complexity to make it more like real life. We can start adding debris and make the training area look more like a real collapsed construction site.
What happens when they find someone?
They have to flip a switch, which makes a noise. Currently, we put them in a vest that has a ball on the collar, containing a microswitch. Rats are trained to pull the ball when they encounter someone, which triggers the microswitch and beeps.
Pulling the ball is not a natural behavior for them, but they can be trained in a process we call modeling. We start by putting the vest with the ball in them. Rats are naturally quite curious, so when they have the ball hanging there, you can see that they’re like, ‘What is this?’
Read more rodent research:
In the beginning, we are just reinforcing them to touch the ball. So, as is standard for modeling, you would stop reinforcing by just touching it so they would realize, ‘Oh, I’m not being reinforced anymore.’ So they’ll try harder [to get the reward]. This usually leads to them pulling the ball, and we have to reward them quickly so they know that this is the target behavior. So similarly, we can shape up until they’re pulling for two to three seconds, so it’s a very strong signal for us.
Of course, in the field we won’t be able to see the mice, or hear them. So we’re working with a group of engineers to develop an all-in-one backpack that will plug into our computer so we can be notified when the mice pull the ball. We will be able to know exactly where they are, because the backpacks must also have a location transmitter.
How can they tell the difference between those who are alive and those who are not?
We’ve talked a lot about this already, as we’re only going to do the training with live people and we don’t know how they’ll react to dead bodies until they’re in a real setting. However, dog trainers have told us that the odor profile of a living person, compared to a dead person, is very different. Dogs can tell the difference between a living and a deceased person about three to four hours after death.
We thought that we might have to train the rats using some sort of scent that we could get our hands on – it’s hard to know what to call it, but we’d need the scent of death, basically.
But dog trainers told us we don’t need to do that, because the change in odor between living and dead people is so different that it’s not a problem.
When will these mice be in the field?
We started training in August 2021 and we still need to conduct training tests outside of the research environment. We are working with a search and rescue group called GEA, based in Turkey, which is prone to earthquakes. Hopefully next year we can take the rats to Turkey for testing there, but in terms of going to real disaster sites, real collapsed buildings… it’s very, very hard to say.
APOPO’s landmine survey began in 1998, and its first operational tests took place in 2003 and 2004. Our tuberculosis detection survey began in 2003 and the rats began operating in 2007. For these projects, it costs, on average, about of € 6,000 [roughly £5,175] to fully train a mouse so that they are ready for operations.
We are currently training seven for search and rescue, although they have to take turns wearing a backpack!
Read more ways resourceful animals are helping humans: