AI’s next ethical challenge: how to treat animals

Not dealing with the relationship between robots and animals is a more urgent threat than a Terminator coming for us, say animal advocates.

AI’s next ethical challenge: how to treat animals

The aerial video of a bear and her cub struggling to climb a near-vertical snowy mountain was both gut-wrenching and inspirational. The clip, which went viral in November, shows the cub sliding down, again and again, as its mom looked on helpless. But, in the end the dogged furball made it, and social media the world over cheered. For some, the cub’s victory was a metaphor for try, try again. One person tweeted, “We could all learn a lesson from this baby bear: Look up & don’t give up.”


But then, just like that, the Internet exploded when scientists revealed that the bear family was likely spooked by the very device that captured the drama—an aerial drone. Bears, it turns out, do not make a habit of climbing iced-over cliffs.

“This stressed me out to the max. Poor mama and baby were terrified of the drone and the little one almost lost its life—just irresponsible and awful,” wrote one animal lover on Facebook. “I don’t care if you make money using drones . . . please be responsible while filming wildlife. We know what a drone is; they do not.”

And she is right. A research project putting heart monitors on bears found that drones flying overhead caused the animals’ heart rates to shoot up, with some taking off running, a potentially risky behavior if roads were nearby.


Who will think of the animals?

Ecologists have proposed rules for flying near animals. While a human pilot was responsible for the bear video shot in Russia, autonomous drones are on the horizon, which means more freaked out beasties. Advocates worry that the intersection between animals and tech is largely ignored.

“Animal ethicists and animal rights activists are often not interested in machines and systems,” says Oliver Bendel, an ethics professor at the University of Applied Sciences, Northwestern Switzerland. “Conversely, computer scientists and AI researchers are often not interested in animals. They do not appear in their world.”

There is plenty of research dedicated to figuring out how to make AI human-friendly, but what about making sure AI is built to keep animals safe? The closest ally creatures have come from animal-computer interaction (ACI), a discipline that officially launched seven years ago with a manifesto that laid out three goals: to enhance animals’ quality of life and general well-being; to support animals in the functions assigned to them by humans; and to foster the relationship between humans and beasts.


What this looks like for livestock, pets, and wildlife varies wildly, and AI-powered tools can both complicate and simplify things.

“Animal technology has been developed for nearly a century, usually within practices that aren’t animal-centered,” says Clara Mancini, the author of the manifesto and an ACI researcher at the U.K.’s Open University.

Animals and driverless cars

In Bendel’s paper,  “Towards animal-friendly machines,” he examines the challenge autonomous vehicles could pose for wildlife. He suggests “annotated decision trees” to help make ethical choices when a machine comes into contact with animals. The trees show flows and outcomes that don’t necessarily stop the robotic systems but provide alternatives.


“I believe that they must be given rules, so that they treat animals well,” he says. “They should not disturb, frighten, injure, or kill the animals.”

MIT researchers ran an online experiment to see what people thought driverless cars should do in a situation where someone, or something, had to be hit. They collected almost 40 million judgment calls from people in 233 different countries and territories, according to a paper published in Nature.

People had to make judgment calls for people crossing the street illegally, the elderly, and animals like cats and dogs. Trends show up from one population to another, but patterns do emerge, including that participants were much more likely to spare humans over animals. There was variance, with Nigeria the country most likely to save humans, while Bolivia is the least likely to prioritize them. There is probably no argument to be made that will convince most that a person should take the hit as opposed to a cat.


That doesn’t mean there aren’t lessons there, or even low-hanging fruit to go after. The National Highway Traffic Safety Administration estimates that there are about 1.5 million deer-related car crashes each year, killing about 150 people. And traffic kills an estimated million animals in the U.S. every year.

Supercharging sensors and data

Driverless cars will not only be more agile at stopping when a last-second leap into the road happens, but lidar and radar could expand the vehicle’s vision to the side of the streets and into the distance. For smaller animals, the algorithms running the car could be taught to slow down when certain species are most active or migrating. Already, there are tools for mapping roadkill, and it wouldn’t be a huge leap to add that type of data as a type of collision hot spot.

The technology that’s created specifically for animals has different concerns. Farms are a place already getting a dose of AI including Ida, for “Intelligent Dairy farmer’s Assistant,” which tracks cows movements and actions through collars and feeds it into a system that can then make assessments about diet and predict health issues. Ida, which has been called a Fitbit for cows, currently tracks seven types of behavior—walking, standing, lying down, eating, chewing, drinking, and idling, with more on the way.


Sensors are nothing new on dairy farms. But AI supercharges the data collection and expands what we know about cows, even, as the Washington Post points out, it’s ripe for abuse, like relying on algorithms to decide what individual gets sent to the slaughterhouse.

Factory farming, which was only possible with modern automation practices and technology, could be even worse, with AI squeezing more efficiency out of an already bad system.

“That’s the worry, that the systems will just refine that further,” says Peter Singer, the moral philosopher and animal rights activist and author of Animal Liberation.


The problem of empathizing with other creatures is the cause of much of the disconnect.

“The way that humans sense their environment are the ways that matter,” says Lucas Joppa, Microsoft’s chief environmental scientist who oversees the company’s AI for Earth grant program.

Robots and the animals

The rights that some argue that we should give to robots use animal rights as a jumping-off point. When it comes to designing systems that won’t end up killing us, much of the language around the fear of how robots will treat us is couched in the terms used for human and animal relations, with Steve Wozniak even saying humans will become pets to robots. But missing the connection between how these robots will treat animals is a gap and a more urgent threat than a Terminator coming for us.


There are more reasons for even the most nature-adverse meat eaters among us to care about animal-friendly AI. The idea of One Health supposes that humans, animals, and the environment thrive and suffer for the same reasons. Technology that keeps all life-forms happy will probably be good for humans in the long run.

“Even if you aren’t someone who naturally finds themselves affiliated with nonhumans, you might still see there is a huge advantage of what role all these other species play in the survival of the planet,” says Steve North, a Computational Anthrozoology research fellow in University of Exeter’s Symbiotic Ethics (EASE) working group.

Technology can even be put to use to save animals from us. AI is being used to stop poaching, track populations better, and get more insight into nonhuman communication. PETA is even onboard and has used drones to detect illegal hunts (of course, hunters now use drones, too). For situations like sheep sheering or laboratories, where PETA has documented abuse, they think machines could also be more humane.


“AI technology can build empathy for animals and even prevent them from being sent to laboratories and slaughterhouses,” says PETA President Ingrid Newkirk. “So while we’re cautious about some exploitative uses like ‘cow Fitbits,’ PETA does support positive moves toward a more respectful future and away from the commodification of other living beings.”

AI animal testing

Introduced in 2002, Roombas were the first home robot system for many Americans. Stories followed from owners detailing the reactions from their pets, including barking and attacking the bots. IRobot, the company behind the Roombas, claim that the tech was tested with animals but pet owners ended up posting in dozens of forums to share best practices, including treating the Roomba like another pet and scolding it.

That thoughtless way technology is introduced extends beyond Roombas when being a little humane can go a long way. New York dog trainer Annie Grossman points out that people approach dogs differently if they have an iPhone in their hand, and some dogs get scared of iPhones because they are always shoved in their faces in an uncomfortable way.


“We aren’t tuned into how it’s changing our behavior,” she says.

For home robots, Grossman suggests to turn it on for a brief moment, introduce the sound and equipment separately and at a distance, and look for signs of stress with the dog. The last step would be putting treats on new, potentially scary, things. It lets the pet interact with it on their own time, in their own way.

“‘Then they can be more like ‘I know all about this, I eat treats off of it all the time,'” Grossman says.


About the author

Jackie Snow is a multimedia journalist published in or on National Geographic, Vanity Fair, The Atlantic, Quartz, New York Observer and more.


Call for Most Innovative Companies entries! Apply now.

500+ winners will be featured on Final deadline: 9/23.