When I was a kid, I was a great fan of Emma. She was a fierce, funny and adventure-loving friend… and she was a steam locomotive.

While Emma is just a fictional prop from a kids book, it’s also easy for adults to project a friendly personality onto inanimate objects. Have you never seen a face in a faucet? On the other hand, machines can sometimes be intimidating: Boston Dynamics’ robot Atlas can run, jump and do backflips. And he isn’t confined to some closed-off factory, but intended to mingle with us human beings!

According to Forbes Magazine, collaborate robots (or cobots for short) are “the latest generation of robotic systems, and they are intended to work alongside humans.” Between Emma and Atlas, there’s a great spectrum of possible futures for cobots, from utopian to dystopian. And it’s exactly this tension that makes cobots so interesting for designers.

Over the years, I’ve had the great fortune to work on a few robot personalities. And in the following, I’d like to share five principles that helped me design cobots better. There are still very few successful cobots out there. So, in order to illustrate these design principles, we need to cast our net a little wider (from drones to virtual assistants, and from games characters to companion animals).

So, here we go with principle #1!

Don’t just look cute, be cute!

Make the cobot feel gentle or weak to give users a sense of superiority.

First of all, your cobot shouldn’t look threatening. A car, for example, is a large, fast-moving bulk of metal. And if they can act on their own, they’re even scarier. (I guess this is why Stephen King wrote “Christine ” about a homicidal car.) It’s no surprise then that Google designed their early self-driving car Waymo to look so needy. In fact, it’s so ugly, simple and harmless it hardly looks like a car at all. By the way, isn’t there the term “ugly-cute” for pets?

But looks aren’t everything; it’s also about how your cobot behaves. When Anki released Cosmo, a quirky small robot toy, they made sure to paint it a very bright white. Presumably, it was only after Cosmo’s great success that they dared make its successor, Vector, a less friendly black. Also, gentle motions make a gentle character. Panda by Frank Emika, is a robotic arm that responds very softly to touch, protecting its co-workers… but also coming across as considerate and benevolent.

When it comes to behavior, the great question remains “who’s the boss”? Users can get a feeling of power by having physical control over the cobot. This is easy when the cobot is small enough to be carried. But it also extends to camera lids for Amazon’s Echo: when you can manually overwrite any software, you’re definitely in charge. (Eat this, “Surveillance Barbie” !)

But cute looks aside — your cobot better be mindful of its core offering … enter principle #2.

Stick to your home turf!

Make the cobot part of a service ecosystem with clear functions, to meet users’ basic needs.

I was very excited when Jibo came out. This was a cute and beautifully engineered robot for my home that could do virtually anything. But a few years in, its makers went out of business. Clearly, Jibo over-promised and under-delivered. And according to Gartner’s Hype Cycle, “smart robots” are still 5 to 10 years from productivity. But even today, there are great opportunities — if you pick your fights wisely.

Qoobo costs a mere fraction of Jibo; which isn’t surprising because it’s a pillow that wiggles a motorized tail. (You heard right.) Evidently, users find it relaxing to cuddle a half-slumbering “pet” on their lap, without ever having to walk it or clean up its mess. This is very niche, but it’s also very affordable and successful.

Your cobot might offer more hard functional value than wellbeing, like NIO’s Nomi. Nomi is a small robotic blob that sits on the car’s dashboard, helping and cheering up the driver. Living inside NIO’s car makes Nomi the undisputed digital master of it, with no competitors having access. Similarly, Amazon’s Alexa is the mere tip of an iceberg made up of shopping and entertainment services.

It’s easy to get carried away by so many functions. But your cobot should better stay humble … which brings us to principle #3.

Show it’s just a machine!

Keep the technology obvious so that users can easily step in and out of the illusion.

Have you heard of the Uncanny Valley? It is that unpleasant grey zone where you can’t tell if something is human or not, and over the years it’s been the home of many scary clowns or evil sci-fi robots. Your cobot should make an even wider bend around that place than KUKA’s iDo concept. While it’s obviously a machine, its proportions and physique are eerily human.

You might also consider revealing some inner workings of your cobot. The charm or credibility of your robot won’t suffer, as illustrated by Raven R. It’s essentially Baidu’s reply to Amazon Echo with beautiful body language. And haven’t you connected with Disney’s WALL-E? Even though he’s just a pile of junk? Lastly, if you’re transparent about your cobot’s insides, no one can claim it’s just a dressed-up actor, like this supposed Russian feat of engineering.

Also think about how users will talk to your cobot. A lot of parents were concerned, for example, when they heard their kids boss Alexa around. This prompted Amazon to release a “polite mode” that only works if basic kindnesses are observed. Another way to avoid contaminating inter-human communication is to invent a rudimentary language. A traditional German woodworker, for example, might direct their work horse with fantasy words such as hott or wüst , instead of rechts or links.

So your cobots might be a machine, but this doesn’t mean it can’t have a credible personality… and that’s what principle #4 is all about.

Stay in-character!

Find metaphors for all situations so that users can have a coherent mental image of the cobot.

One way or another, your cobot won’t be perfect. So why pretend otherwise? Google Assistant uses humor to sidestep shortcomings: when you ask it to clean your room, it might reply “Let me try… did anything happen? Sorry, I guess I can’t.”
Another limitation might be your device’s energy level: Paro is a robotic seal that dementia patients pet and cuddle for relaxation. When its batteries are empty, is recharges through a pacifier in its mouth, adding to its cute appearance.

Another challenge to your illusion of life might be when your cobot is reprogrammed. (Remember that creepy “Turn out your cheek” command in Westworld?) One way to avoid this is by using physicality as an input. Sawyer is an industrial robot arm by the recently rebooted Rethink Robotics. When Sawyer is calibrated for motoric tasks, users simply move its arm to wherever it needs to be. Who needs a computer anyways?

Still, your cobot is made up of software and hardware alike, and can do what we can’t: leave its body behind. So ask yourself what your cobot’s “spirit” might look like when it migrates onto screen. Siri always looks like Siri — be it on your phone, watch or Home Pod. While no mechanical manifestation of her exists (although BMW’s wavy kinetic sculpture might be close), I can’t wait to see her floating in virtual space or AR as a sentient haze.

Once you have a recognizable character in place, it’s time to reach for the stars… so, last but not least, here comes principle #5.

Foster emotional bonds!

Let users develop a long-term relationship with the cobot, so they are more satisfied with it.

Robokind helps children diagnosed with autism improve their inter-human communication skills. Some might place this cobot’s facial design in the Uncanny Valley (see above) or even find similarities with Chucky the killer doll. But Robokind stays positive (and very successful) no matter what. And isn’t it this unconditional attention that people love in their dogs too?

E.T. is an alien and not a machine, but I simple love how his heart glows when he’s very emotional. That’s also what the makers of (now out-of-business) Kuri must have thought when they embedded a LED in his translucent chest. This creates an effect of vulnerability and intimacy that can be reserved for special moments. By contrast, when a user tells Siri, Alexa or Google Assistant “I love you”, all shy away from replying “I love you too.” It appears that machine love better not be put into words.

Still, our love for artificial characters can go very far. Sony stopped producing its first Aibo robot dog in 2006. So when it broke, its owners had to say goodbye forever. This gave rise to a dedicated robot cemetery near Tokyo. If you think this is an exclusive far-eastern phenomenon, read through this gamer’s feelings of loss when his virtual horse was killed.

But where there’s funerals, there’s also weddings: as early as 2009, a gamer “married” a character from a virtual dating game Loveplus, with more similar ceremonies to follow. Consider preparing your cobot for the entire lifecycle of a relationship.

So, these are my five principles. As I said in the beginning, they rely on diverse examples. But should you ever have a chance to design your own cobot, I encourage you to seek even more far-fetched inspiration. It will be these outlandish stimuli and metaphors that’ll help us all explore this exciting new domain.

In closing, there may be many ways that robots can turn out badly for humankind. But I doubt that we’re doomed if we treat cobots kindly. In fact, today we treat many parts of our environment like soulless objects — animals, eco-systems, even other human beings… instead, wouldn’t it be much healthier if we could find something relatable in all that surrounds us?

I think so. And I guess so does Emma, the friendly steam locomotive.