R2-D2 or C-3PO? Robotics designers weigh in on which approach makes the most sense
Some human qualities can be useful, but there are limits
David Dymesich, Design Lead at Fetch Robotics:
"When a robot has too many human features, people will think that the robot is smarter than it actually is and that robot will most likely disappoint them. If they are not disappointed, they will most likely find the robot creepy. And with most anthropomorphic robots, people are both disappointed and creeped out.
"But some humanoid features can be helpful. For example if the robot has a head that is capable of panning (pivoting around), it could show intent by looking in the direction it wants to travel before it actually does. This could be considered a safety measure to protect surrounding humans.
"Additionally, the more humanistic the robot is, the more empathy people will have toward the robot. Computers or even desk-top robotic arms can be viewed as just tools which can be replaced. If a robot can invoke empathy, people will want it to succeed and may even treat it like a co-worker or a companion instead of a replaceable object if it fails. Even if a robot is not humanoid, but rather cute and a bad-ass at its primary function, then people will still feel empathy for it."
Don't forget the "Uncanny Valley"
Cory Kidd, CEO and Founder of Catalia Health:
"To me it depends on the application of the robot and the intended interactions that people are supposed to have with it. Is it conversational and trying to build some kind of relationship? Human-like is great! Is it a tool that's going to manufacture some piece of a product 2,500 times a day? No need for a human-like appearance.
"Having any sort of social features makes robots easier to relate to, so if that's what you want, then including those features (which may include a human-like appearance, but could also mean other types of social interaction, from conversation [think Amazon Echo] to gesture [think Pixar's Luxo Jr.]) can be important. You want someone to *not* develop a relationship with this? Well, many probably will anyway (how many people name their cars?), but leave out the social cues to make it more like a tool.
"You also need to consider the uncanny valley, which is a theory first given by Mori, a Japanese roboticist, a few decades ago. Translated from the German unheimlich (particularly from Freud's usage in his theories), the concept is of something familiar, yet unsettling or eerie. When it comes to robots, this was applied to those that are intended to look human-like at first glance, but then as soon as you approach closer or see them start to move, they're clearly not human, leading to that unsettling feeling.
"I strongly believe that unheimlich actually extends well beyond appearance to other aspects of a robot's interactions, including speech and language capabilities, motion, social cues, and more. Any of these can lead to a negative impression if they're not designed and managed well in a robot."
Humanoids work well in environments built for humans
Steffi Paepcke, Lead User Experience (UX) Designer and co-founder at Open Source Robotics Foundation:
"Technical capabilities aside, one benefit of the humanoid form is its ability to function in a world designed by and for humans. Stairs are perfectly sized for human feet. Drawer handles fit cleanly into human hands. Doorways accommodate (most) human heads. Based purely on form, a humanoid robot is more likely to function effectively in a human-centric environment.
"Non-human form factors require some level of customization in order to adapt to the workspace. This is especially relevant in the case of search and rescue tasks, where a robot is meant to enter a degraded, human-built environment and perform a task. This task may involve grasping tools or valves designed for human hands, or driving a vehicle. The fewer customizations required to make the robot field-ready, the better."
We have an amazing capacity to project human qualities on non-human things
Steffi Paepcke, Lead User Experience (UX) Designer and co-founder at Open Source Robotics Foundation:
"We as humans can't help but project human-like qualities onto beings that are objectively not human. We relate to the world around us through a filter of our own experience, and often assume our own experiences or capabilities are generalizable. Think about that dog you see outfitted in a raincoat and adorable little boots. That dog's owner himself doesn't want to get wet, and extends that desire to his dog, who likely (but what do I know?) isn't perturbed by a bit of rain.
"My point is that we often can't help but assume that an entity is human-like in some capacity. This appears to be the case with robots as well. Robots such as the Paro and Pleo, seal and dinosaur simulacrums, respectively, elicit strong emotional reactions and care-giving tendencies despite our rational understanding that they are mere machines. If the goal is to evoke a more emotional response such as delight or affection, with few expectations of cognitive ability, a non-humanoid form is likely to be more successful than a humanoid. If your robot can carry out a small subset of tasks well (or poorly, for that matter), I think it best to steer clear of the human form factor, lest the real human in the interaction project more complex expectations onto the robot. Of course, if your robot is in fact more capable than its form lets on, you could be underselling your bot which will either pleasantly surprise your users, or deter them altogether.
"While an anthropomorphized miniature dinosaur and hopelessly defenseless seal may "require" our affection, they do not trigger within us the notion that they might equal us intellectually or physically, as a humanoid robot might. Once a robot takes on physically human traits, our expectations of its capabilities enter the realm of human as well. This can have powerful consequences.
"The most common consequence these days is disappointment. If we see a massive, bipedal machine attempting to open a door, we expect it to open the door. If instead it topples pathetically on its side, our expectations are not met and we may feel let down. We didn't expect sweet little Paro The Seal to open a door with its precious little fin, but this behemoth? It should have opened the damn door! When our eyes suggest that a robot might be human-like in its ability, our brain quickly follows suit. If the robot is indeed as capable as it appears, this is no problem. If, however, the robot topples at the door, our esteem for the robot dwindles, perhaps irrecoverably so."
Pet-like robots are more approachable
Adrian Canoso, co-founder and Design Lead at Savioke:
"With a pet-like form, there's the advantage of making the robot really approachable by everyday people, something that's incredibly important for our robots that are delivering items, from staff to people who may have never seen a robot before. The cool thing about robots is that they are typically persistent and infinitely patient. When that's coupled with a pet-like form something magical happens, it's a robot that seems to be unconditional and always ready to help it's human counterpart.
"We worked hard to make Relay feel "right" by designing in just the right amount of humanity into the robot while carefully balancing ergonomics of the interface and getting access to the robots payload. This didn't come easy though, with our earlier prototypes were too minimalistic and in user testing, people had the desire for the form of the robot to have more humanity in it.
"During the design phase of Relay, I constantly thought about those first 5 seconds when a hotel guest puts their hand on the door, opens it and what do they see? Are we setting the right expectations? Does the guest know what to do next? Are we making sure they are happy with what's going on? After multiple prototypes and over 10,000 deliveries we think we got the form and interactions right, something we're really proud of."
Non-humanoids have more potential than humanoids
Jeff Gee, Co founder of Simbe Robotics:
"A robot that looks like a copier may be treated like one. Humanoid robots in a similar way may be expected to have emotions and empathy as well as human-like kinematics, and when they do not meet the expectations they are rejected by people.
"On the other hand, when you need a robot to operate machines designed for a human to use, the robot may require the same dexterity as a human. In this case, humanoids may be appropriate.
"But non-humanoid robots have the potential advantage to exceed humanoid robots. Non-humanoids can solve problems in completely new ways, without human constraints. Of course there are risks. In a situation like a senior retirement community, a non-humanoid robot may have a negative effect if it isn't comfortable to around. A robot with four arms may perform well, but may also frighten your grandmother. This is where design and empathy for the user and their environment are critical in the success and acceptance of robots."
Too human or not too human ... that is the question
What's clear is this: As robots become more dexterous, intelligent, and versatile, roboticists are facing an unusual choice. Should their creations look and act human, like C-3PO, or should designers avoid anthropomorphizing machines, taking the R2-D2 route instead? (Sorry, I can't help it; my Twitter feed is full of Star Wars-themed Arduino builds.)
It's a more nuanced question than it appears, as these designers point out. Non-humanoids can have human characteristics, such as eyes and natural sounding voices, and some humanoids try to be more human-like than others. What is clear is that the decisions made now, as robotics technology begins to proliferate in daily life, will have lasting consequences on how we think about robots for years to come. GE's Unimate served as the template for industrial robots for decades after it was introduced. In much the same way, the most successful of today's bots will likely cast molds and spawn lineages of their own.
Shout out to Tim Smith, for the gallery idea and for helping arrange the responses.