A menagerie of robot animals

When people meet a new robot, they instinctively relate it to a familiar experience – a character or object that becomes a reference point for how they expect a robot to behave and what it can do. Research has shown that people perceive animal forms as the most warm and friendly type of robot. But do robot animals have more potential than just as talking toys? Petra Stock talks to their researchers about the tricky combination of function and form, and how we might use it to our future advantage.

The baby harp seal is propped on the table in front of me. The pup’s eyes are closed. Its dormant, rotund, furry body looks soft and inviting. On hearing its name – “Paro” – the seal lifts and tilts its head. It opens wide, deep-lagoon eyes, and blinks.

Unable to resist, I reach out and gently scratch Paro’s neck. I stroke its back and whiskers. The seal pup leans into my hand, and its body rumbles like a purring kitten.

“Oh,” I hear myself say, as I exhale and allow the creature to nuzzle its warm, soft body against mine.

Pleased to meet you

Like people, first impressions matter for robots. Human-robot interaction researcher Nathan Dennler says when people meet a new robot, they instinctively relate it to a familiar experience – a character or object it reminds them of. Those metaphors act as an anchor point for how they expect a robot to behave, and an indicator of its capabilities.

Dennler works in the Interactive and Collaborative Autonomous Robotics (ICAROS) lab at the University of Southern California, USA. He and colleagues recently assembled a collection of 165 interactive robots, to better understand the ways people respond to a robot’s physical form.

They asked nearly 2,000 participants across three surveys to describe and rank a range of humanoid, animal-like and mechanical robots. Participants were asked to choose metaphors to describe each one and their expectations for its social interactions, including warmth, competence, gender, social role, likeability and function.

“We wanted to figure out how people conceptualise those robots,” he says.

Science fiction has perpetuated a fascination with humanoid or droid-like forms, but Dennler’s research suggests zoomorphic designs offer certain advantages. People perceive animal forms generally as the most warm and friendly type of robot, and least discomforting, he says.

For instance, participants in Dennler’s study rated Paro highly in terms of warmth, alongside Leonardo – another small, tactile robot which looks like Gizmo from the movie Gremlins.

The machine’s physical form is about more than looking cute, or cool, or futuristic. Dennler’s research shows design breeds expectations, which boosts the human experience when those presumptions align to the robot’s capabilities and intended purpose. 

When Paro’s creator, Dr Takanori Shibata, first thought of designing his “artificial emotional creature”, he pictured a big egg shape covered in fur: something nice to touch. Now chief senior research scientist at Japan’s National Institute of Advanced Industrial Science and Technology, Shibata was among pioneers developing personal robots in the early 1990s. At the time, his conception of robots was in stark contrast to the prevailing fixation on fast factory limbs for welding, assembly and packaging.

In a 1996 paper, Shibata introduced the concept of a personal – or pet – robot. 

“When we think of machines, we design them as convenient tools and evaluate them in terms of objective measures, such as time, accuracy and energy,” he wrote. “We may search for the best solution. This is a rational way and quite reasonable in engineering.”

Instead of precise industrial automatons enslaved in factories, Shibata pictured personal robots living at home alongside people. “They might hold it or put it on their lap, while stroking, watching TV or talking with others,” he explained.

Shibata’s egg idea hatched into an animal. In designing Paro, Shibata considered three possibilities: a baby seal, or the more familiar form of a dog or cat. To test the concept, he made robots modelled on all three options and asked people to evaluate them. Dogs and cats were well known. But this established certain expectations – to purr, lick, play fetch or feed – which the robot couldn’t always live up to.

A social robot engages with people’s emotions. Human-robot interactions might simulate positive feelings like affection, happiness and love, or negative responses like anger, sadness and fear and will be evaluated accordingly, Shibata prophesised.

“Human beings are not machines. We have ‘mind’. The emotions have been considered to be important.”

Electric dreams

When you picture a robot, what comes to mind?

The hard and shiny almost-human kind like C3PO from Star Wars or evil Megatron from Transformers? Or, the squat, droid designs of Daleks?

The term “robot” is loaded with stereotypes and expectations. Science fiction has primed us to overestimate the technology, and often to expect the worst. We hear robot and think of the Terminator, or an uprising of killer machines just over the technology horizon; maybe also artificial intelligence surveillance systems tracking our every move, prompting fear and suspicion.

But does that say more about the robots, or about us?

Paro was developed during the first wave of social robotics in the late ’90s and early 2000s. It was a golden age for zoomorphic robots – mechanical objects designed to look like animals and programmed to interact with people in a social way. The strangely furry and owlish Furby, sleek ­plastic puppy Aibo, bopping yellow duck Keepon and baby dinosaur Pleo all made their way off shelves and into the hearts of a generation of wide-eyed kids and kidults, including me.

Since then, more than 340 different social robots have since been created. Most are prototypes, made for research purposes only. Fewer than half ever became commercially available, and about a third remain in production.

In the second period of social robotics – between 2006 and 2012 – creative critters gradually gave way to a population boom of human-like models, as roboticists chased the technical challenge of replicating themselves. Some took things literally, creating machines with eerily hominoid epidermis, hair and facial expressions. Others came skinned in hard, white plastic.

But as the anthropomorphic wave subsides, a more beastly brood is making a comeback: like the soft-bodied Tega, a red and blue blob with a tuft of troll hair; or the blue-eyed Photon, ­somewhere between a bat and a tricycle; or the hard-shelled Hexa that creepy-crawls on six legs. Increasingly, form is now following function. Insectile Hexa can navigate various terrains. Friendly Tega charms children to assist with early literacy. Animal forms bring other benefits too.

“People were expecting them to be fulfilling the role of a pet. Something like a companion, that maybe doesn’t do the dishes for you”

In general, people don’t ascribe these animate objects a gender, says Dennler.

Zoomorphic robots avoid concerns about bias and stereotyping arising when designers or users attribute a specific gender to a robot. Problems can arise, for example, if a feminine-looking robot is assigned a stereotypically female role, or assigned to servant-like tasks.

But he advises that not all robots can or should be animals. It’s important to align a robot’s functionality with people’s expectations based on its physical form.

“For the robots that look like animals, people were expecting them to be fulfilling the role of a pet. Something like a companion, that maybe doesn’t do the dishes for you,” Dennler says.

Dinosaur designs

Kate Darling’s home has become a gathering place for baby robot dinosaurs.

Darling is a research scientist at MIT Media Lab, and an expert in human-robot interaction and ethics. Her home is a refuge for lost Pleos, the robot designed by Innvo Labs to emulate a week-old baby sauropod. Darling houses seven of these robot dinosaurs, though only the one she calls ‘Mr Spaghetti’ remains functional.

‘Yochai’, ‘Peter’, ‘Bones McPleo’ (an early – and slightly disturbing – skinless, skeletal prototype the company gave her) are heaped next to three other nameless and broken shapes, sent by owners hopeful she might give them a second life.

In her book The New Breed, Darling argues people should stop comparing robots to humans and think of them more like animals instead.

“It’s always struck me in our conversations around robots and artificial intelligence that we like to subconsciously compare this technology to humans. So, robots to people and artificial intelligence to human intelligence,” she tells me.

“I think that’s the wrong metaphor. Artificial intelligence doesn’t think like human intelligence. It perceives the world differently, understands the world differently.”

Darling’s not suggesting animals and robots are the same either. But, she says, thinking about them that way allows the chance to consider new possibilities for what robots could be.

One practical reason for designing robots like animals is biomimicry – borrowing ideas from nature. It’s ­useful when thinking about locomotion and ways that robots could navigate different parts of the planet – for example, a fish-like robot designed for moving underwater, a featherweight robot fairy for travelling on the breeze, or a snake form for navigating narrow spaces.

Another is our own preconceptions.

“We have so many expectations of robots. By stripping this away and playing the expectations down, then people could form a better bond.”

Darling’s Pleos are based on a half-size baby Camarasaurus, and benefits from the lack of judgement about its capabilities or behaviour. After all, no human has lived alongside a real one. Like Paro the seal, this helps the dino-bot bond with humans on an emotional level, without the expectations of a more familiar fur-child.

According to its maker’s website, the dino-bot progresses through four life stages as humans interact with it. From opening its eyes and learning to stand, it begins to recognise its name and explore its home environment, ­making goofy “haw” noises and demanding human attention. As its personality evolves, Pleo even develops likes (sugar cane snacks and pats) and dislikes (it cries if punished).

Animal robots can be delightful, but Darling warns that this means care needs to be taken with how they are deployed.

Out of context, they can freak people out and send things “completely off the rails,” she says.

“We know from research and human-robot interaction that people have a very visceral response to certain designs of robots, because we subconsciously treat robots like they’re living things.”

By way of example, she says, public responses and reactions to the Boston Dynamics robot, Spot, tend to vary wildly depending on its use.

Spot is a crayon-yellow quadrupedal robot that many people consider to be a dog.

“Depending on the setting, people have either enjoyed having the robot around, or they have found it very upsetting,” she says.

When police in New York city trialled using Spot for surveillance and CCTV in hazardous environments, including hostage situations, it repelled many city residents.

“It’s creepy, alienating and sends the wrong message to New Yorkers,” says a spokesman for NY Mayor Bill de Blasio, capturing the public response.

In Australia, a young woman in Brisbane’s Fortitude Valley had a similar response when coming across a prototype military-grade robotic dog out walking with its engineer at 2:30am According to media reports, her immediate reaction was to give ‘Stampy’ a sturdy kick to its front sensors, causing thousands of dollars in damage.

It seems we want our animal robots to be with us rather than against us.

Bringing bots closer to home

Little robot Blossom is hardly more than a ball of wool, wooden ears, small motors, rubber bands and a series of 3D-printed gears and platforms waiting to be freed from a rectangular mould.

“When you build it yourself, you value it more,” says Blossom’s creator Dr Michael Suguitan, who developed the robot as a doctoral student at Cornell University, USA, with collaborator Dr Guy Hoffman.

Blossom is a handmade, DIY robot. Its internal skeleton can be 3D printed, its movements and behaviours programmed by open-source software. The robot’s “skin” is crocheted or knitted; ­depending on your crafting abilities and choice of ­appendages, the robot could be a rabbit, a bear or even a squid.

“The whole ethos of the project was ­questioning many things about how robots are designed now,” Suguitan says. “If you Google search robots, you’ll see a bunch of robots that look like the robot Eve from WALL-E. They often have hard, white, plastic exteriors, a bit like ­consumer electronics.”

He wanted Blossom to be the antithesis of this idea.

“We have so many expectations of robots that look like this […] Eve from WALL-E, who’s a super futuristic, super smart, intelligent agent,” he says. “And robots, […] they’re really not to that level yet. I think by stripping this away and kind of playing the expectations down, then people could form a better bond.”

“If they’re more animal-like then it opens up this whole new idea space of possibilities and different roles that they can play.”

Beyond roles as toys or substitute pets, aged care is one area where robot companions are being used relatively extensively, to help older people who are lonely and disconnected, or who miss having animals in their lives.

Dr Simon Coghlan, a philosopher with a background in veterinary science who now researches the ethics of digital technologies and robotics, has a few concerns; including one he defines as “deception”.

“The robots are designed to mimic or replicate the kinds of behaviours and appearances of living animals. The closer to the real thing they are, the more likely deception will become an issue,” he says.

Another issue he flags is “consent”: whether the person being given a robot has understood the technology and agreed to its use, especially if a robot is collecting data or being used for surveillance.

But Coghlan can see that, a bit like living ones, robot animals might offer some comfort to people.

“Studies show some people – not all – do get a lift in spirits from interacting with these animal-like things, these robots,” he says. “And so  [they produce] those kinds of emotional effects on people that maybe reduces their loneliness.”

Nonetheless, he cautions, the relationship can’t be mutual.

A living animal, “a creature with sentience or an inner life”, can recognise its owner, miss them, try to comfort them, Coghlan says.

Whereas a relationship with a robot animal? “It’s completely one-sided.”

Role models

Paro is accepted in the US, Europe and Australia as a therapeutic tool or “biofeedback medical device” in aged care, particularly with dementia patients.

After the first-generation Paro in 1998, the seal robot was commercialised in Japan in 2005. There are now more than 7,000 Paros in use across 30 countries, including Australia. The robot’s design is robust and durable, and Shibata is aware of some owners who have been living with Paro as a pet for more than a decade.

Under its white pelt exterior, Paro’s hand-manufactured robot skeleton contains an internal artificial intelligence system connected to touch, light, audio, temperature and posture sensors. Speakers replicate a seal voice sampled from real baby harp seals, and an internal heating system makes its body feels warm to touch.

The baby harp seal robot requires only electricity to function, with power supplied to the pup’s mouth via a plug which looks like a baby’s dummy.

In one sense the robot seal is advanced technology – it can learn a new name, and adapts its behaviour to be better liked by human companions. And, unlike many modern technologies, Paro does so without software updates, WiFi connection or the usual amassing of personal data to be shared with the cloud or third parties. Shibata says that’s because he wanted Paro to be robust and durable over the long term, and owners “do not want to be bothered by technical, privacy or security issues”. 

Meeting Paro takes me back to my ’90s childhood, a time when I considered the baby harp seal to be the pinnacle of cuteness. As a child growing up in suburban Adelaide – about 20,000 kilometres away from the real animal’s North Atlantic home – a large, framed photograph of an alabaster seal pup hung on the wall of our suburban study, the computer room.

It was also a time when new technology inspired awe and delight.

Today, it’s near impossible to separate the march of tech progress – new platforms, apps, devices, AIs, robots – from the accumulated wounds of pervasive creepiness, data gluttony and algorithmic deceptions, constantly surveilling and intruding unwelcome into our private lives. I know this, but as this animatronic harp seal mews and snuggles, I feel joy and warmth.

Unlike humanoids or droids, zoomorphic robots like Paro can be comforting. People – even me – are inclined to accept and warm to them, so long as they’re on our side.

The robots we see in popular media – the Terminators, battle droids – reflect our fears for the future, but it’s only one possible path that robotic design and engineering might go down. It’s as easy to create a vision of loyal C3P0 and R2D2 helpers – and equally possible to consider a world in which dinosaur-inspired harvest helpers and spider-engineered search and rescue bots are the norm. Just don’t expect them to do the dishes.