The movie Robot and Frank tells the story of Frank, a grumpy older man who lives alone, until his son shows up at his house with a robot that has been programmed to take care of him. Initially, Frank wants nothing to do with the robot. But over time they start to get along and even become friends. And the robot turns out to be one of the funniest characters in the movie.
Do you ever wonder if robots can have personalities?
To find out more, we talk to Maja Matarić, a roboticist and a distinguished professor of computer science, neuroscience, and pediatrics at USC. Her lab makes “socially assistive robots,” which help people do things that are emotionally difficult. And robot personalities are a key part of her work.
Have a question you've been wondering about? Email the Ever Wonder? team to tell us what you'd like to hear in future episodes.
Perry Roth-Johnson: 00:06
Hello! This is Ever Wonder? from the California Science Center. I'm Perry Roth-Johnson.
Perry Roth-Johnson: 00:13
As I was getting ready for this episode, my guest today actually recommended that I watch a movie called Robot and Frank. It tells the story of Frank, this grumpy older man who lives alone, until his son shows up at his house with a robot that's been programmed to take care of him. Now, initially Frank wants nothing to do with the robot. But over time they actually start to get along and even become friends. And the robot turns out to be one of the funniest characters in the movie. So, this all got me thinking... Do you ever wonder if robots can have personalities? To find out more, I talked to Maja Matarić a roboticist and a distinguished professor of computer science, neuroscience, and pediatrics at USC. She makes "socially assistive robots," which help people do things that are emotionally difficult. And robot personalities are a key part of her work. Let's get into it.
Perry Roth-Johnson: 01:09
Maja Matarić, you are a roboticist and a distinguished professor of computer science, neuroscience, and pediatrics at USC. Uh, welcome to the show. Thank you. It's great to be here. Your work really centers around creating robots that help people do things that are hard to do emotionally, but not physically. And these are socially assistive robots, or as one New Yorker article put it, "robots that care." And you're really a pioneer in your field. This area didn't exist 15 years ago, really. And now, especially, it seems like the pandemic might be accelerating the timeline for when socially assistive robots could go mainstream. Uh, but before we start thinking about the future, I just want to start with, if we could, where we are today. And I've heard you say that we can think of different robots on a spectrum from automation to augmentation, and that right now you're arguing, most people are working on automation. Uh, what did you mean by that? Can you help unpack that?
Maja Matarić: 02:03
So if we look at the history of robotics, people used to think of robotics as being created to do the three D's—the dirty, dull, and dangerous jobs—and all of those jobs were physical jobs, right? So, you know, if you could avoid going into a dangerous nuclear plant, deep underwater, far into space, things like that, everything physical, everything unpleasant more recently, of course, with, you know, a great advancement in robotics and AI. We've seen robotic surgeon to all kinds of areas, not just practical and these days, not just physical. And so if we think about what robots are being created for, they're typically created to optimize some process, to make things faster and cheaper, and that often can mean taking the human out of the loop. So that is automation, but there's really a spectrum of the kinds of things robots can do. And I am really, really passionate about getting people to remember, recognize, realize that robots actually do a lot of things that augment human ability that make us better at doing our own work rather than doing our work for us. So that's a distinction between automation versus augmentation and I'm, I'm rooting for augmentation.
Perry Roth-Johnson: 03:14
Got it. Right. Can you give me some examples of each, maybe like one on, on polar opposites of the spectrum and some, you know, that might kind of go in the middle of it?
Maja Matarić: 03:22
Absolutely. So an example of automation, uh, is a factory assembly robot where you take, you know, the, the worker out of the factory and you make the robot do the work. And this is what we see today. We see robots behind cage doors doing work. Um, and then we also see factory workers doing their own work. And the two are separate. Now there's a move to kind of put them together and to split up the work. Um, and what we're interested in is on the other side of the spectrum, which is how can a robot let's say encourage someone who has differences or suffering from depression, um, or, you know, might just have to relearn how to use their limb because of an accident or an injury. How can it help them recover, learn new skills, regained skills, so they can go back into the workforce and do whatever it is that they want to or use to do, right? It may be physical, but he may not be, it may be cognitive work. Um, so another example I like to give is, uh, you know, people are familiar with revive vacuum cleaners, you know, like the Roomba. So I don't, I actually, I I'm fine with the Roomba. I have a Roomba, but what you really need is you need the will to get out of bed and, and vacuum that floor, whether you vacuum it or the Roomba vacuums it, you gotta want to get out of bed and do something. And I like to think of it as we make the robots that help you really want to get out of bed and do whatever it is that you want to do.
Perry Roth-Johnson: 04:46
Hmm. Why are robots good at doing those kinds of tasks as well as the automation tasks?
Maja Matarić: 04:50
So it's a good question. Why we need robots to do augmentation? Um, we are social creatures, right? So we are very, very driven by social factors. We care if someone else cares about us, right? So we're very driven by knowing that what we do impacts someone else, that someone else cares, that someone else is paying attention. So it's that social bit. That's very important. And of course, conversely, when there is no one around, you know, we get very lonely, we get very isolated, right now is a great example of all the negative effects of, you know, lack of socialized social interactions. And so, um, why are robots good at helping with that? Because if you're trying to help a person help themselves, they need social support. And unfortunately social support could be lacking. If, for example, it's lacking very much for the elderly because their peer group is, you know, often no longer around, um, their offspring, children, grandchildren have lives of their own. So they're very lonely, they're isolated, there's a demographic. Uh, but similarly there are giant demographics, um, on the younger age of that spectrum. So for example, consider individuals on the autism spectrum because they have social differences. They're often socially isolated or even socially shunned. And yet they need social interaction in order to practice, to interact with people. So that's another place where robots make a difference, but really if you think about it, we are all social animals and we need social support and we really need social support 24/7, if you will. And so, you know, some of that support will come from people, but sometimes that's not enough. And that's why robots are the important supplements.
Perry Roth-Johnson: 06:28
So why don't we pivot into some examples of your research work specifically? 'Cause I do want to ask you about a green bird named Kiwi, which can help children with autism. Tell me about your green bird robot Kiwi.
Maja Matarić: 06:39
So we have been actually working with, uh, children with autism and, and in, in general individuals on the spectrum for almost 15 years. It's one of the first things that we started off when we developed the field of socially assisted robotics. 'Cause it was, it was a very large population and a growing population that was looking for a social skill training. So when we wanted to develop a robot, 'cause unfortunately we don't live in a time where you can just go to a store like the robot store and you could pick out a robot to do research with, and I want to go into a store and say, okay, I want that one and I'll have 20, right? We are not there yet. So until we get there, we're often in research, um stuck with the need to create our own, which is, you know, an interesting challenge because it takes electrical engineering, mechanical engineering, computing, it takes a lot of skills. Um, and it's, it's a wonderful, interesting thing to do. But when you create something, of course, you first and foremost, think about your, your user, who is the user, what did they need, what will work for them? And so when we wanted to work with children with autism, we talked to children with autism, uh, and we talked to their parents and their families to find out what might work and that's how we always designed robots. So we always talk about the user first and talk with the user, um, because you'd be surprised, even though that's so obvious, people don't do it, um, engineers in particular. Um, and I'll say we I'm an engineer, but engineers in particular, we tend to think that we know all the answers. Um, and so, you know, we don't really have to check, let's see, we'll build it 75% then we'll check. Well, that just doesn't work. You have to go from day one. And so we did, we worked with a lot of participants. We also wanted the robot to work, not just for children with autism, but for families, for people across the age spectrum because people, again, don't exist in a bubble. And so you really, when you're creating something, when you're building it, you're building it for that whole ecosystem. And so that's why we chose to make it not a human, not a humanoid robot because often individuals on the spectrum have some challenges in interacting with other humans. So making a robot that is very much like a human doesn't help with that, you want something that they will find fun and interesting and less intimidating and flooding to interact with. And yet it to have enough human features so that they're learning social skills, which is why our robot is not a human, but it has a humanlike face. It has the eyes and it has the facial expressions that are humanlike, but very, very much simpler. So again, they're not too much that they're just enough for helping to train and learn social skills, learn to read facial expressions, eye gaze, turn-taking, etc. And then there were also other key things about the design. For example, there are a lot of things you could put on a robot that are simply fun, but they might be distracting for the user. And in particular for children on the spectrum, if we had added wings on the robot, they would have been really fun, but then they would have played with the wings and that's both a distraction for them and not really socially appropriate. Right. Cause you wouldn't go and touch other people's arms if you didn't know them well. And so we're always developing these robots in a way that they would promote appropriate social behavior training so that the kids would actually be better equipped to play with other kids. 'Cause that's the goal.
Perry Roth-Johnson: 10:01
And the face of Kiwi, it's a screen, right? And then the rest of the body is made of fabric and it kind of looks like a puppet, is that right?
Maja Matarić: 10:09
Right. We actually had a puppeteer help us design the robot, but the robot is autonomous. So it's important to know that it's not a puppet that's controlled by a human. Um, but yes, there is a face, which is on a screen. And that's a really important thing because we really want our robots to have enough facial expressions so that the human brain of everyone who interacts with them, reads them as if they are human. Like, um, and so all the social cues are there, right? Eye gaze, when I look at you versus when I look away. That's a really important cue that children with autism need to practice, but it's very hard to build physical place with faces. So if you, if you have a robot that has physical eyes and physical eyebrows and a physical mouth, that's a lot of motors. Um, it's much more expensive. It's much heavier. So the robot itself has now heavier, bigger, potentially more dangerous. Um, and also if you don't do it right, it can be quite creepy. So there, there are some robots out there that have physical heads that are kind of creepy. Um, we call that falling into the Uncanny Valley. And so the other reason that it's really nice to have a screen is that you can try different things, right? So for example, some children might find eyes to be a little off-putting so we can make smaller eyes or bigger eyes. We don't have that luxury on a physical face. But having said that, we also don't want the whole thing to be just on the screen, because if it's just on the screen, then it doesn't tap into our social interaction and the way that our brain communicates and interacts with other physical creatures around us. So that the physicality, what we call physical embodiment is very important.
Perry Roth-Johnson: 11:49
You stole the question right out of my mouth. I was going to ask why not make the whole thing a screen?
Maja Matarić: 11:54
Everybody asks that. But yeah, they're really important things that can be done through a human computer interaction. No question. It's very powerful and very important, but we are social creatures and our brains work in a particular way that when we interact with other physical creatures around us, whether they're pets or other people or robots and more of our brain is engaged, we learned better. We're more motivated, we're more affected. And that's what we want when we're helping people.
Perry Roth-Johnson: 12:21
On that note, you know, your work obviously focuses on helping people with these special needs: autism, stroke, Alzheimer's. How did you get interested in helping these kinds of patients? And like, why did you pick this area of research? Because it wasn't the very first thing you did when you first became a researcher and a professor, right?
Maja Matarić: 12:39
That's right. So I, um, I'm an immigrant. And when I first moved to the US, um, I was 16 and my uncle said to me, you should go into computers. Computers are gonna be big. Um, and he was right, and this was a while ago, but I won't say how long. Um, and so I, you know, and then I was interested in robotics in general because it's computing in the physical world. You know, you create something, but it's right here next to you, as opposed to, you know, in this machine. So when I first got into robotics, um, first of all, human robot interaction didn't exist as a field. Um, and I sort of went with teams of robots. I was interested in teams cause you know, I like social interaction, but I didn't really realize it. And so, um, so that's why the teams were robots. And I had them herding and flocking and doing fun things.
Perry Roth-Johnson: 13:22
Like robot swarms?
Maja Matarić: 13:23
Yeah, it was, oh, it was a lot of fun and it was really hard and it was new. And so I'm not, I don't regret doing it. I'm really pleased. I did that, but it was only after I had my first child who is now 22, I have three. Um, but it was only after I had heard. And I started to answer questions about not just what does Mom do, but why does Mom do it? I mean, it's cool that Mom builds robots. What are these robots for Mom? What did they do? And by then they weren't herding. They were doing other things. And I just realized, I really, I want to be able to say that I'm doing something that, that is just so that everyone can say, wow, that is great, that you're doing that. That's really helping someone. And so that's what I did, what we now call a pivot. So I went away from what I was doing before, and I really looked around to see where robots can make a difference, given that they're not perfect, that they can't, you can't just throw one in the home, like in that movie robot and praying where you say, okay, empty the, the, you know, the trunk of the car and go make lunch. Well, we can't do that yet. And I'm not going to work on that because that's just, it's too hard. Okay. I want to work on something where the robots can help people today and in three years and in five years, um, and in my lifetime, and that's where we discovered this and I say, we, my lab, um, we discovered this niche where for example, people recovering from surgery or from stroke.
Maja Matarić: 14:46
Um, they needed support to do physical exercise and they had to it, you couldn't do it for them. And yet there was no one there to keep supporting them, to keep telling them all, you're doing great, lift up your arm a little bit and to do this for hours a day. And then same with autism. Who's going to be there to help the child learn what eye contact to make and how to keep trying when they're demoralized and isolated? And then you could just expand that. And then there's so many populations, right? So, you know, look at anxiety today. It's huge. It's pervasive. So, so we just really got motivated. It's saying, how can we create social partners for people when there aren't human social partners to support them. And then also intersected with that is, well, why robotics? Well, first of all, I have a roboticist. So that's why. But also...
Perry Roth-Johnson: 15:35
Yeah. Got to work with the toolbox he has.
Maja Matarić: 15:37
Exactly! And I think that's really important. I think people need to realize that. I often talk to young people and they worry about like, Oh my God, I don't know what my path should be. It's okay. Build your toolbox. You will build your path. And so that's what we did here. We kind of had the robotics toolbox. We looked at the research and it said, Hey, people really are influenced by physically co-present physical buddies, right? Again, pets, animals, other people and robots. And so we thought, Oh, okay, great. We want to help people. Being around social creatures helps people, let's create these social creatures that revise, that will assist. And that's why the field is socially assisted robotics. So we just, we made it up as we went along and, and because we kept looking for a place where we could make a difference. That's I think why it stuck, um, because a lot of ideas get, get born and thrown out in the world. And I think a lot of them die off. I think the way to make an idea that'll stick is to look at human needs when you find human need, that you can help even a little bit that will stick, hopefully.
Perry Roth-Johnson: 16:43
Yeah. Yeah. Um, and it seems like a key part of meeting that need is the personality that, that a robot might have. Kiwi doesn't move around? Right? Kiwi is just kind of on a tabletop, doesn't have any wheels or legs. So it's the personality of these robots. That that's the main point. So, so how do you even do that? How do you give a robot a personality in the first place?
Maja Matarić: 17:05
Well, so personality is incredibly important, right? We all have personalities, and I don't mean, you know, Myers-Briggs like the basic, you know, for whatever. I mean, it's a very complex set of properties that we have, our features and they change depending on what situation we're in. And there's no such thing as being just an extrovert or just introvert, all kinds of interesting stuff. And so even, even before I was doing human robot interaction, I just could never be happy about the general notion of a robot, what things is are robotic. And it does. It's like, why, why would it be like that? And so I could never, I could never understand that. Right. And so we realized very early on that we could create machines that people found interesting and engaging, even if they weren't doing anything useful. But if they were kind of like curious, like you were curious, like, what is it going to do next?
Maja Matarić: 17:54
Something that was non-obvious and that's really related to personality to character. Um, and so, so now you think, well, okay, so how would you create a personality for a robot and the place where you look of course is, well, what is personality? Personality's pretty well studied, if not well understood, um, in social science. And so we looked at, um, different kinds of personalities. We honed in on extroversion/introversion because that's a pretty well understood spectrum where any person at any point in time could be behaving with an extroverted or an introverted personality, although they themselves are not necessarily one or the other, right? It's situational. Well, so what defines extroversion/introversion? How fast you talk, how close you get to someone, how much you move your arms, um, how high pitched your voice is. So think about it. All these things can be measured. All these things can be programmed into a robot. So that's what we did. Even before worrying about what an extrovert or an introvert might say, we focused on what an extrovert or an introvert behavior would be. And so we basically created this robot that could either move its arms around a lot or not so much, that could talk a lot or not so much, they could get kind of quite close to you. I mean, not creepy close, but just reasonably close, like within social distance.
Perry Roth-Johnson: 19:15
Maja Matarić: 19:15
But closer versus kind of keeping his distance a little bit. Um, and when you combine these features, what you get is a robot that looks shy versus a robot that looks, you know, really like socially interactive. And so it looks extroverted versus introverted, but it's very measurable and it's in ways that people also appear extroverted and introvert. So I would always say, you know, when you think about how are you going to create personalities in machines, you really first have to look at the science of personality, which is really interesting. Um, and blanket lessons from there because otherwise what might happen is you might, if you throw the kitchen sink at it, you again might get something creepy and inconsistent and not so great. And there's quite a lot of interesting social science, um, that can inform us. And then the thing that's really interesting about personality, um, is that when you endow a robot with personality, people will actually prefer its behavior with one personality or another, depending on their own personality.
Maja Matarić: 20:16
And they may not even be aware of. So let me give you an example. We did a study where we had, uh, an introverting behaving robot and an extroverted behaving robot, and we had people interact with it and people who were introverts because we also gave them a personality test. We give them a lot of tests, so they didn't know which one mattered. And so, uh, when we, people who, uh, scored as being more on the introverted side of the spectrum, um, tended to prefer the introverted robot, but they didn't say so if you ask them, they said, Oh, like them all, there's no difference in what they said, but in how they interacted and how long they interacted, they always preferred the robot that was more like them. So this is really important because what it tells us is that our personalities drive what we like. And also it tells us that we're not necessarily conscious of our preferences, because if you ask us, we may not know. And yet our behavior gives it away.
Perry Roth-Johnson: 21:13
Yeah. So that's actually a nice segue into this next question. Are there different cultural approaches to making the socially assistive robots work better in some communities than others? There are different cultural attitudes towards robots in Asia, in Japan and Korea, for example, compared to here in the US. How does that play out when, when, when you're designing the personality for the robot?
Maja Matarić: 21:34
So cultural influences are really, really critical. And in fact, if we look at, um, the general public's willingness to embrace robotics to begin with, um, in Asia, there's much more general interest to kind of like, Oh, I'd buy a robot. And, you know, even if it's useless, Oh, I don't want to buy it and I want to have it in my house. Um, so that's, that's interesting, whereas in the US in particular, there's more skepticism. Um, there's more concern. Um, but there's also more skepticism in terms of like, ah, is this useful? What is this going to do for me? Um, so that probably speaks to the fact that we have to work harder on personality, um, in the US than we might in Asia, where they're more willing to kind of try it out and give it the benefit of the doubt. Um, having said that, whatever personality you build in of course has to be culturally informed. And so it is not the fact that we can actually easily build robots here and have them be just as well accepted in Japan or Korea, um, or for that matter, even in Europe. So that's going to become more subtle. Um, that's a challenge for companies, but what I think is perhaps more interesting again, is if you think about what problem you're trying to solve. So if you're thinking about solving the problem of, let's say isolation, um, and you know, depression related to isolation in the elderly and many people these days, that is really that, that cuts across all cultures, because now you're looking at how can you, what can you say to this person? And what can you offer them to make them feel less isolated? And those elements tend to cut across cultures. So in fact, telling jokes across cultures is quite complex, whereas helping someone feel less lonely, um, is not as complex because, you know, we know when makes people feel less lonely and it's the same thing. It's, you know, asking them how they're feeling, hearing them out, offering something related, emotional support that cuts across cultures. Um, so that's, that's where I think for the kind of work we do, we find grounding and relevance across cultures by looking at the actual human problem.
Perry Roth-Johnson: 23:37
Right. Um, I want to go, uh, where we, where we started in hinted at as like, looking ahead towards the future, like, how are we to these socially assistive robots going mainstream? Like, has the pandemic accelerated this timeline in, in your view?
Maja Matarić: 23:54
The pandemic really has accelerated the timeline because it's, well, it's exposed so many needs that were there at a smaller level, and then just got blown up. Right? So we talked about autism. So think about, um, just autism as an example, about one in 45 to 65 children in the U.S. is diagnosed, and those numbers are going up, they continue to go up. So, so all those kids are getting a lot of services, if they can afford it and their family can afford it. And if they're properly informed where they can get them. And so it's, it's very, very hard, right? And so people are always looking at something in the home. Now, imagine what's happening through the pandemic. They can't get some of those services because it's risky and difficult. And then look at all of K-12 education. Now it's largely in the home. So it's, it's a tremendous opportunity. And I don't think it's lost on the companies. So there is an incredible niche and we have the right technology already today, and then we have the silent need, which is why I think it's going to happen, like, right, right about now in the next, you know, two, three years, things are going to start to come out more. Um, but again, I'm always an optimist.
Perry Roth-Johnson: 25:01
If socially assistive robots were widely successful and widely distributed, we have human caregivers. Uh, would those jobs be replaced in some sense, or is there just such a need, like we don't have enough human caregivers that it's sort of a moot point and a false premise?
Maja Matarić: 25:17
Well, I mean, you'll say I'm biased, but you know, I have actually done a lot of looking and studying and had my own family and like, you know, Alzheimer's care. And the things that are needed are not going to be replaced by robots anytime soon. Do you know when you have an individual suffering from Alzheimer's, they basically need something 24/7, they have a spectrum of needs. You know, they, they might be having psychotic episodes. They might become violent. Um, they need to be gotten out of bed. They need to be toileted. Their beds need to be changed. Their nails need to be painted. None of these things can be handled by machines. We can't, we can't be around machines and be confident that we could be that safe with individuals who are frail. So there's plenty of room for people here. It's, you know, when I talk to caregivers for people in nursing homes, which is a kind of job that has the long, the shortest staying power, people just get out as quickly as they can, because it's such hard work. And it's so heartbreaking to see people losing themselves. And so that's such a hard job. People only do it if they're really have a mission of helping others. Um, and you know, when I talk to people who do that, they say things like, you know, I really mostly enjoy actually painting, you know, Becky's nails, then she's calm. And she seems to smile and it makes her happy, but I can't paint her nails because I have to go change all the beds. Um, so, you know, we need robots that can the beds, so people can paint the nails, but right now we need everything because what's happening now, you can't put a bunch of people into a nursing home and keep him safe because of the pandemic. So, so it's an even bigger problem. People are isolated, no one is taking care of them. So yes, people should be taking care of people. We have to change the entire structure of society. So that taking care of people is valued and trained and paid. And even then there will still be room for machines to help us because we're complicated creatures that need a lot of support. And that's what that augmentation is about. There is room for really helpful social support for both people and machines.
Perry Roth-Johnson: 27:26
As we're wrapping up. Do you have any final thoughts? Any other stories you'd like to share? Well, what else do you want people to know? Leave them with?
Maja Matarić: 27:33
Um, yeah, I mean, I, I just think there's, it is an interesting time. It's a very difficult time right now, but it's a, it's an interesting time of opportunity. Um, and the other thing that I keep trying to tell people is, uh, you know, you know, what came after the dark ages, the Renaissance. So this is, we have to see this as an opportunity. It is an opportunity. It's an opportunity to do a lot of good. There are a lot of really hard problems out there that we can be working on. And we have so many incredibly smart people hold potential. So let's work on those problems and we'll make a difference.
Perry Roth-Johnson: 28:06
Well, Maja, uh, that's it for our time. It was wonderful to talk with a pioneer in your field and, and telling us about the robots that you're working on that are, they're making an impact in our community. Where can people follow you online and find your work?
Maja Matarić: 28:19
So I bet if you Google me, you'll find me very easily. Uh, I run the interaction lab at USC, not hard to find, and we have lots of videos in our website and, you know, we take volunteers and high school students and we just, they just love to get people, get excited about doing something good.
Perry Roth-Johnson: 28:37
Awesome. Well, Maya, thanks for spending some time with us. Thanks for coming on the show.
Maja Matarić: 28:41
Thank you. This is fantastic.
Perry Roth-Johnson: 28:43
Well, that's our show and thanks for listening. Until next time, keep wondering. Ever Wonder? from the California Science Center is produced by me, Perry Roth-Johnson, along with Jennifer Castillo. Liz Roth-Johnson is our editor. Theme music provided by Michael Nikolas and Pond5. We'll drop new episodes every other Wednesday. If you're a fan of the show, be sure to subscribe and leave us a rating or review, or tell a friend about us. Now, our doors may be closed, but our mission to inspire science learning in everyone continues. We're working hard to provide free educational resources online while maintaining essential operations like onsite animal care and preparing for our reopening to the public. Join our mission by making a gift at californiasciencecenter.org/support.