How Dr. Wendy Ju Designs Robots That Won't Freak You Out

If nosotros're going to coexist with our silicon cousins, a deep agreement of how humans interact with machines needs to exist explored.

This is Dr. Wendy Ju's expanse of expertise. As Executive Director of Interaction Design Research at Stanford's Heart for Design Enquiry, she studies how design affects our interaction with robots, which she has applied in the practical realm every bit a enquiry collaborator with Ford, Toyota, and Bosch. PCMag was at Stanford recently and saturday downwards with Dr. Ju to learn more than nigh her piece of work there, and her upcoming move to New York'due south Cornell Tech.

Hither are edited and condensed extracts from our chat.

Your PhD dissertation at Stanford was on The Design of Implicit Interactions. Tin can you explain what you mean by that?
It's a subject area which has always fascinated me. There used to exist a lot of discussion about what was called "invisible computing"—frictionless and seamless—and some people took that literally, trying to brand the interaction itself disappear, but they still needed to brand sure they were communicating with the user. It'south not just text and talking; there are many other ways in which we communicate implicitly which nosotros need to bring into man-to-auto interaction. It's very hard for engineers to know that design is an integral part of creating a production that people tin can use.

What was your start exposure to computing?
Well, both my parents worked in Silicon Valley and then I grew up hither. I tin't remember my starting time contact with a calculator—nosotros had an IBM PC, [merely] I learned to plan at a very early historic period. In some weird style, that's made me more open-minded, because I'chiliad not falling back on my introduction to computers. For example, I've been educational activity students using a Raspberry Pi and they said: "Er, is at that place a web interface?"

Right! if someone's cognition of computing starts with getting the 2007 iPhone, they're going to have a very dissimilar concept from those of us who remember a command line pulsing waiting for a response.
So true.

So how about robots, were y'all a C-3PO or R2-D2 fan?
The kind of robots I'chiliad into are everyday objects that suddenly come to life. Similar a well-made chair that'southward really a robot, or a trash tin, or the automated cars that we've been working on.

Ah. Less humanoids and more automata. Have you always attributed subconscious behaviors to household objects?
When I was very small I'd sometimes stub my toe on the door and my parents would go up to the door and whisper "How could you?" Then they'd turn to me and say: "Hit the door dorsum!" So ridiculous [Laughs]. Fifty-fifty now I meet the anthropomorphic robots and remember, hmm. I actually prefer a well-made chair which we can then add motion to, because its design makes information technology await as if it's well-nigh to motility anyway, so let's aid it. I'chiliad a big nerd, I have an applied science degree, and I've worked effectually robots my whole life. But I'm coming to the field from a deep interest in industrial blueprint and bringing a dynamic aspect to that.

Which brings us to your experiments with moving furniture and objects.
We did a Wizard of Oz field experiment using a remotely operated human being-in-the-loop trash butt to workshop how the interaction would go in the real globe. 1 of the things nearly interaction is information technology takes two. As a designer you tin can't design the interaction itself—that dynamism requires the homo aspect.

We were shocked to observe out that people expected the trash barrel to say "thank y'all" when they put trash in it—as information technology'due south a service robot, not the other manner round. Nosotros also saw a little boy baiting the trash barrel with a piece of garbage, to get it to follow him and motion somewhere. I realized he had a model in his head where he idea the trash barrel wanted to "eat" the garbage, equally if he was feeding an animate being. 1 of my team said: "Oh no, we thought we were making a service robot, but they think it'south a beggar!"

Not the status they expected?
No. For example, when I evidence the video to roboticists I say: "See how the robot is low status" and they frown and say: "Well, we can make it talk, then information technology'll be high status." But, no, that's not going to work, that's non what people wait. It's only a low status trash barrel robot.

Bet the engineers were depressed. They simply want to work on the latest cool shiny high condition object.
[Laughs] It was so bizarre. But it does explain the strange blueprint behind some of the robots that are around now.

Like the moving robots with big format touch screens which are essentially big Manga style eyes on top of a giant iPad.
Right.

In contrast, your moving ottoman experiments showed how people might respond when other household objects suddenly spring to life. Tin you talk about that?
In some ways the mechanical ottoman projection had another calendar. It revealed all the different unanticipated use cases of how people responded to information technology. Some people put their anxiety on it, others opened it up to look within, some were very shy, others scooped their legs underneath and shrunk back on the seat to avoid it when it started to motility. People do different things to signal when they're comfortable or uncomfortable, and we need to notice them in lodge to sympathise how to design interactions.

Talk nigh your autonomous car experiments. Those were fascinating.
People oftentimes get move sickness in automatic cars because they're usually doing something else, and then their torso tells them i matter—due to the machine's motility—merely their eyes tell them something else. We were testing embedding haptic signals to give you a sense of what'southward coming up alee to show what the car is doing. What we plant was in many ways more than interesting, about what people tend to pay attention to, or not, in autonomous cars.

Did people respond to the car as if information technology was a robot?
Nosotros establish something chosen the "Hello event." We weren't thinking virtually the car as a character at all. But we decided to get the car to rev its engine once the driver approached, and people interpreted that as "Oh, the motorcar is proverb hi!" And so they'd say, "I know the machine isn't going to hurt me because it said hullo." And I'd remember "Oh god, that's a terrible assumption to make."

Shades of Herbie the Dear Bug sentient automotive. Funny how Disney is conditioning our response to The Future.
Correct. Considering when engineers talk about trust in automation terms they mean the car is roadworthy. Merely when people talk about trust they mean it in the social sense—"Are yous on my side? Will you defend me?" It'south so of import for engineers to see the results from our experiments, otherwise this field won't take off in the mode nosotros expect. In fact, if you look at the Disney movies they literally showtime from the inside—projecting the character's intention into each activity. When we bargain with objects and machines, we use the model we already take from observing other humans. And so having a machine tilt its head, or make a fiddling "hmm?" dissonance to illustrate its metacognition is helpful.

Outside of Stanford, y'all're besides working with the big motorcar companies, correct?
Yes. I'k working with the Toyota Research Found, Ford, Bosch, and several others. We are so close to having democratic cars on the route at present, from a technical perspective, that people have realized interaction is going to be the key to making this work. Or they'll exist a failure if people don't know how to apply them. Oft the car companies don't take the research processes to work in this space, so they come to us.

Near of the human factors work comes at the end of the design cycle—they really don't practise much in the traditional car blueprint process to bring people into the middle. Which is why car companies are currently having their lunch eaten by the tech companies, [which] have shorter product design cycles and are more used to dealing directly with consumers. Some of the work we've been funded to do past Ford recently is on household robots—they're working out what business they want to be in beyond cars.

Exterior of cars, what other attribute practice you want to work on side by side?
We've been talking to Steelcase, and I'd love to piece of work on more furniture that starts to move.

Finally, what made you lot decide to become dorsum Due east to teach at Cornell Tech [starting in January]?
Cornell Tech is an educational startup and in that location have only been so many successful education experiments in the past one-half century so I felt it was a real opportunity to establish a stronger pattern inquiry background. I don't experience the intellectual underpinnings of blueprint take been established as well as they could be. I like the East Coast; I loved my time in Boston but I barely left the MIT Media Lab during my time there [laughs] so it'll be absurd to go dorsum. I experience New York is such a global city that what happens there oft gets picked upward throughout the rest of the globe really rapidly. I'm looking forward to it.