Cynthia Breazeal MS '93, ScD '00

INTERVIEWER: Cynthia Breazeal, you're both an alumni of MIT and an associate professor at the MIT Media Lab, where you founded and direct the Personal Robots Group. You've been acknowledged as a pioneer of social robotics. And your book, Designing Sociable Robots published in 2002 by MIT Press, is considered a seminal work in the field.

Today, among other topics, we'll talk about your research exploring expressive social exchange between humans and humanoid robots. You're also the founder and chief scientist of the company Jibo Incorporated, which is bringing to the mass consumer market the world's very first family robot, Jibo, which, or whom, you've described as the meta of your work at MIT in the field of social robotics. Welcome, Professor Breazeal.

BREAZEAL: Thank you.

INTERVIEWER: We'll start chronologically with your early years. We'd love to know what it was like growing up in New Mexico and California, your family, your inspirations. What did you do for fun?

BREAZEAL: Sure. Yeah, so I was born in New Mexico. My father was a computer scientist at Sandia National Labs. But we moved to Livermore, California when I was three. So he was basically transferred from the Sandia branch in New Mexico to Livermore, California.

And my mother was also a computer scientist working at Lawrence Livermore National Labs. So I grew up in a very technologically scientifically rich experienced family. I think growing up it's quite different now than it was back then. Now, Livermore, California is considered an extension of Silicon Valley. So it's much bigger, obviously much more technology. When I was growing up, it was a lot of cow pastures. We made a joke that it's a lot of cowboys and physicists.

But again, because of the National Labs there, a very unique environment in that a lot of the parents were scientists, engineers. And so there were a lot of opportunities for me. So growing up, not surprisingly, my parents, back in the day first they'd be bringing home these punch cards. That's how you programmed computers back then.

So I was exposed to the ideas of computers and technology very, very early. They'd take us to science museums and exploratorium and things like that. We would always get the earliest computers or gadgets. So I remember going to school being the first student to come in with my essays printed out by a computer. So technology was just part of growing up for me.

And my parents always really encouraged me to think about a career in science and technology. Because they understood it was tremendous opportunities, very interesting work, which of course, was different. Because a lot of my friends weren't necessarily getting that kind of encouragement. So I think I was fortunate in having both parents as really powerful role models in that.

I also spent a lot of time in competitive sports. I was a very competitive athlete in a number of sports. And in fact, every now and then when California does a Title IX celebration, they often invite me to come--

INTERVIEWER: What were your sports?

BREAZEAL: --and give a speech. So I did soccer. I did tennis. I did track. I did some swimming, but really was mostly-- tennis was my big sport in high school. Soccer and track, when I was younger.

But I also credit that experience as being really formative, because-- I guess in two ways. So the first was I believe through being a part of a team you learn about teamwork. You learn about the importance of playing your position to the best you can. You learn about how to work with other people even when things might be going really rough, as well as when things are going really great.

You learn dedication and training, and I think especially for me for track, with wind sprints, you definitely learned to endure the pain and discomfort. Because that's what it takes to really be great at something. So I learned a lot of, I think, tenacity, and persistence, and a sense of never giving up no matter how daunting or challenging something may be. And so I think that was really important.

I think also, in retrospect, because I was an athlete at a high level, I often competed with guys, as well as girls. So I was always comfortable competing with and against guys. Which I think when I went to college and going in engineering, of course, it's a very male dominated area, I always felt very comfortable. And I think that helped me in a lot of ways. And I think the sports helped create a common ground that has really served me not only through undergraduate and graduate, but even in business life now. So sports was really, I think, a really formative experience for me.

INTERVIEWER: Yeah, I get it.

BREAZEAL: So between the sports and between all the technology, I guess it's not that surprising how I ended up where I am. But that definitely got me started on this path. And then I often talk about why social robots.

I saw Star Wars when I was 10. And I just fell in love with R2D2 and C3PO. It became like my theme of that year in school. Everything was about those droids for me.

INTERVIEWER: Although I gather you wrote an essay. There's something when you were in third grade.

BREAZEAL: In third grade. Yes, no it wasn't enduring. Robots that were much more like helpful companions than autonomous task drudging automaton or tools. Robots that would relate to us in this interpersonal way, robots that were, again, this aspect of relatability, this human element that could connect to us beyond the utility, but the emotional side always fascinated me and always captured not only my mind, but my heart.

INTERVIEWER: Does it fill a void somehow? Why would you think of a robot doing that as opposed to a friend?

BREAZEAL: Oh, I had lots of friends. No, I think because I loved animals. I loved animation. I loved characters and stories. And I think for me robots just fit into that. It wasn't about a hole at all. It was more about this whole new way of experiencing connection. And I think we as human beings are driven as part as our nature to want to connect, not only to other people, but to other things, companion animals to nature. And so for me, robots was just a really other dimension for that, that was really compelling.

I think there's something very special about connecting to the not quite human other. And I think that really captured my imagination. So as I went through school, it wasn't as if I'm like I'm going to build robots for the rest of my life. It was-- I did sports. I wanted to be a doctor.

There's a time in college I wanted to be an astronaut. And so I thought in order to be an astronaut I need to get a doctorate in a related field. And so I chose space robotics. And I did my undergraduate at UC Santa Barbara. I was majoring in electric and computer engineering.

And at the time, there was an NSF Center of Excellence in Robotics. It was a very fortuitous time for me. So I was starting to learn about robotics in a big way as an undergraduate.

And as I was applying to graduate schools, I of course applied to MIT. And it turned out there was a brand new project that Professor Rod Brooks was starting around planetary microrovers. And at that time-- this is in the early '90s. Rod of course, is a pioneer founder of autonomous robots. But all the robots when I first came to the lab were very much inspired by insect models of intelligence.

So they were-- I remember the day I came in to lab, and I was in one of the admitted graduate students. You come, and you look at the lab. And you meet the students, and you talk to the professors. And I remember this moment coming into Rod's lab and seeing these little insect-like robots going around doing their various things. And I remember it's like that Star Wars moment just rushed over me.

I remember thinking if we are ever going to see robots like that in our lifetime, it's going to start in a lab like this. In fact, it may start in this very lab. And so I'm like, I have to be here.

And that was really the moment I have to say that sparked that Star Wars dream could actually happen in a place like that. So it was this latent thing. But it really-- it was that moment like, oh my God. You could actually try to do that.

INTERVIEWER: You were in the right place.

BREAZEAL: I was in the right place. So obviously that's why I came to MIT.

INTERVIEWER: What was it like? What was the atmosphere there besides the robots? What were the people like?

BREAZEAL: Yeah, I felt like I was finally among my people. So I did my undergraduate at UC Santa Barbara. It's a gorgeous campus. It's on the beach. It's amazing in so many ways.

But the intellectual vibrancy, and curiosity, and just that willingness to work yourself into a stupor about going after something that fascinated you just was not part of that culture. But it was very much a part of MIT culture. And I really felt finally I'm in a place where I can be like this. And it's like people are psyched. And they're doing it with you. And so it was really-- it almost felt like a homecoming, I think, coming to MIT.

So I came in, and our group, which was then called the Mobot Lab, really had a fabulous culture. And there was--

INTERVIEWER: Did you say mobot?

BREAZEAL: Mobot for mobile robot lab. Because at the time it was basically about autonomous robots. But a lot of them were mobile robots. And there was a woman, Anita Flynn, who was a research scientist at the time. She also became a PhD student.

But she really was the person who helped set a really amazing culture in that particular group. So she set policy. She's like everyone digs in and helps each other. If there's a big push to get something done, we're a family. Everybody helps everyone. Or she'd arrange these fun events.

She was the instigator of the AI Olympics, and she got the whole lab involved in doing these silly things, silly competitive events, from athletic to chair races down the hall to hacking a competition. But she spearheaded all that, became a big part of culture in the AI Lab. But she was just that kind of person. And coming into that lab with Rod, of course, who was so famous-- but then also Anita, who was always there, and was really creating this amazing culture among the group. Really made it a very special time and a very special place to be. And it was also interesting that at the time I joined the lab, I think it was 50% women. So Rod was always a huge supporter of women in AI and computer science.

And coming into that environment with Anita in such an important role, also extremely technically competent. You could always go to Anita to ask her questions. But the students doing that with each other, it's a culture that I've definitely tried to replicate and foster in every other organization I've done since then. So with that sense of family, that sense of play, that sense of passion and love for what you're doing.

Never squashing each other's ideas. Being able to give thoughtful critique, but in a supportive way. It was amazing. It was amazing.

INTERVIEWER: That sounds pretty fabulous. Were there are a lot of interdisciplinary-- were people coming in from different areas? How did you, as a group, or you personally get into cognitive science? All the other things one learns. And besides electrical engineering, computer science, what were you studying? What did people bring to it?

BREAZEAL: It was an interesting path. So I can tell you the day I really started thinking seriously about social robots-- and I didn't know it would be called that at the time-- but it was the day that NASA landed Sojourner on Mars. And I was a graduate student at the time. Of course, Rod was one of the people who really was an early proponent of the idea of small planetary microrovers for planetary exploration.

He wrote this famous paper, "Fast, Cheap and Out of Control: A Robot Invasion of the Solar System." So I was fortunate to be in the lab in a time when it was really pioneering that whole idea. And of course the culmination of that, in so many ways, was NASA landing these small rover on Mars. And I remember thinking huge success for, not only not NASA, for all of robotic, and thinking OK, now we sent robots into the oceans, and we send them into volcanoes, and we even send them to Mars but they're not in our homes yet. Why is that?

And I remember thinking, when you think about the complexity of the human environment, especially in the home, it is so radically different than the context and environments we had thought about when you think about exploring someplace far from people. So in the home you have people, and people's behaviors not governed by laws of physics. They're governed by-- people have minds. It is the minds that give rise to all of these behaviors.

INTERVIEWER: Personalities.

BREAZEAL: And beliefs, and thoughts, and emotions, and no one was really seriously tackling that question of, how do you design a robot that could interact with anyone, regardless of education, or life experience, or age, or whatever, and have that be successful interaction that people can get value out of? So in many ways I liken it to that moment when computers were these expensive things that only experts use. And then finally someone thought, what would it mean to actually have anyone be able to use a computer? So it was that moment for me. And we already knew that people tended to anthropomorphize these autonomous robots because it just triggers these very low-level cues of animacy.

And so we thought, well so maybe the social interface is kind of the universal interface. That if people are already inclined to try to understand robots in these terms, we should design them to support that mental model. And I knew that in order to dig into that problem, you couldn't start by trying to essentially simulate an adult. That's too complicated. People have amazing social cognitive capacities, amazing emotional richness.

Charles Darwin said human beings are the most social, and the most emotional of all species. So I thought, well so what is the simplest creature, in some sense, that you could try to model that could still get into this dynamic with a person, with an adult. Understanding that through that interaction from those experiences the robot could learn and could develop-- almost much like how we socially develop-- could be pulled along this developmental trajectory, become more sophisticated, until eventually someday you could reach the adult state.

So that was always the underlying motivation of a lot of that early work. Very inspired by developmental psychology. And I had no background in developmental psychology. So I just started reading as much as I could, attending classes.

It was really a crash course. And not just for human development, but even for other species, other animals. Looking at different theories and models. And it was a very fortuitous time as well because the field of in psychology of understanding emotion was kind of going through this renaissance. So there are more theories down from the behavioral to neuroscience on emotion and interaction.

So I was able to pull from all these different disciplines. And it was fascinating because it was almost a process of taking a theory from here, a theory from there. Some of it combining stuff from comparative psychology to animal behavior, some of it coming from developmental psychology to create this artificial creature that became Kismet.

INTERVIEWER: I was going to say, how do you start? There's so much that you're talking about tackling. Where did you actually start?

BREAZEAL: I started basically getting that inspiration from what am I trying to model? What's kind of that simplest thing I can start with? Because nobody had done anything like this before. The simplest thing that I feel I would allow me to explore that phenomenon I'm trying to get at. And the phenomenon was this interpersonal interaction between people and this non-human thing, but where those interactions were rich enough and of a sort that the robot could actually benefit from them, in the sense of being able to learn from those interactions.

And what are those core initial behaviors that you would have to design that system with? And so a lot of that was inspired by developmental psychology. What are their earliest behaviors that get the infant and the mother or father in that dynamic? And then starting to abstract, in some sense, what is the right level of abstraction for that model? So it was very much combining in weaving the science in with the technical and in the engineering.

But it was also a real fascinating design process. Because as you interacted-- and I knew as you would interact with this robot-- you have to want to interact with it. You would have to want to invest your time. You would have to want to, in some sense, nurture it. It's this process of wanting to teach and play the same way that we would with children. So how the robot appeared was also really important.

INTERVIEWER: To make it engaging.

BREAZEAL: To make it really engaging and to pull on-- [INAUDIBLE] talks about Kindchenschema. So this baby scheme that you see in dolls and toys and Disney characters. But the design of the robot would have to naturally put the person in this mode of, oh, it's a young creature. It's adorable. I want to play and teach.

It's just pulling on what's already I think innate within us. And so the design of Kismet was also very inspired, both by animation, as well again as development psychology area. So there was a reason why Kismet had these big eyes.

INTERVIEWER: Stephen Jay Gould has written about that. Something about the big eyes.

BREAZEAL: And the proportion of the head to the body. So all of those things were designed to really get people naturally into the right mode of interaction that the robot could then modulate and dynamically interact to make this dance happen. One of the paradigms that was so fascinating to me, reading a lot of the literature was talking about mother-infant or parent-infant interaction was a dance. It's this constant adjustment simultaneously. This kind of attuned reciprocity.

Which is so unlike when we think about how we interact with Siri or other systems. Like you say something, and it responds. It's like moves of chess. They're discrete. Whereas our interpersonal interaction, it is this dance.

It's this rhythm. It's these micro adjustments. That's what I wanted to capture because in some sense, that serves as the vehicle for a lot of other stuff. It sets the rapport. And so a lot of the early work with Kismet was really trying to understand that.

And so, not only the social cues, but the emotive cues and the motivational systems. And not just kind of the decision making, but these underlying drives and motives that would shape and bias all the other behaviors the robot would engage in to try to sustain, in some sense, this communicative dance with you. And through that dance if it wanted to try to get interactions or experience about a certain kind, it would, in some sense, nudge the person to get them to engage in those behaviors with it through these emotive social cues. Much like how babies are so good at having us do that. So it was that game for Kismet. So it was exploring--

INTERVIEWER: Kismet was the first?

BREAZEAL: It was the first. It's recognized as the first social robot. And it was really the first to explore this human robot dynamic interplay on the social, emotional, physical, expressive dimension. And again, all of that was in place because it became the substrate for learning and development. That was the whole point.

INTERVIEWER: Now you had an evolution of various-- there was Kismet and there were others. And were you trying to answer specific questions with each type? With each robot?

BREAZEAL: Absolutely. I think every project, not surprisingly, inspires a whole host of follow-on questions and ideas. And one of the really truly amazing things about being at a place like MIT, is that you can then build the next robot-- in some sense it becomes your experimental apparatus, to explore the next question. Whereas other colleagues of mine tend to buy robots because they can't build them, it almost becomes not the greatest match between the scientific question they're trying to answer and what they can actually access. Whereas at MIT, you can design the actual robot that you think is going to be best suited to get at the underlying science that you're trying to do.

So I think that's been a real asset of being able to do that work here. So we would design a different robot in many cases to explore a different sort of question, about either human robot interaction, or understanding something about social intelligence about the machine, or emotional intelligence about the machine, or some kind of social learning. Learning by imitation, learning by social referencing, learning by tutelage. All these different questions and models around really thinking about human robot interaction as this tightly coupled partnership that robots of the future were going to interact with us, not as tools that were teleoperated, but as partners, autonomous partners, that would be able to understand us, be able to learn from us, be able to communicate in ways that are natural for us, that treat us like human beings, as opposed to other technologies or rocks to be navigated around.

So it was all about this intimate partnership that really was informed deeply by human behavior, human psychology, our minds, the way we experience things. And trying to come up with the best match where it was never about trying to build a robot that was identical to people. It was a very different tact of saying the thing that is so fascinating, and the value proposition of this work is that robots are not human. Because of that they have different capabilities that can really complement ours.

So we can work as a team that can up the capability of the team. But you need to get that engagement right. You need to get that interface right.

INTERVIEWER: So the design was another thing you played with.

BREAZEAL: It's always been at the intersection of design, of psychology, of engineering. I'd even say of art, and expression and animation. It's always sat at that critical intersection. Because your fundamentally dealing with people. And people have all of these dimensions.

And I think one of the big epiphanies of the work over time has been, as we've created robots to explore different context domains, like robotic health coaches, or learning companions for children, or thinking about social robots as a new kind of telecommunications medium beyond Skype or something like that. Where you have a physical embodiment in place, or elder care. All these domains. What we're coming to realize is the better a technology can really support richly human experience, not just around the cognitive, which is where a lot of technology is focused today, but the cognitive, the social, the emotional, and the physical.

It deepens our engagement. We're able to bring more of ourselves into that interaction and experience. And as a result, we as people, are more successful. So with the physical, social weight robot health coach, people were more successful with the robot then they were with a computer that had the same dialogue model, gave the same advice.

In some sense, the quality of the information was identical, but that relational other dimension of support was missing. And so people had greater engagement. They had an emotional bond with the robot but they didn't form with the computer, that led to again, this sense of-- this robot is really-- the computer-- I do my taxes on the computer, I do email, I do all these things on the computer. But this robot is here to help me with something that's really important.

INTERVIEWER: What did that one look like?

BREAZEAL: It was a robot that combined a flat screen on a body. It had a head. It was very simple, but it did a couple really critical social cues. So it was able to make eye contact with the person. It was able to share attention on the screen. So it would look at the person or it would look at the screen.

But it was like an appliance, in some sense, in the home. As you can imagine, you'd put it on the counter, it'd have a persistent presence. It could do face detection, so it knew you were there. It would talk to you. So it gave you that social engagement, and that persistent presence, versus a computer that, when it's off, it's off.

So I think that sense of persistent presence was also a fascinating different element that the computer didn't offer. But the bottom line is when you talk about something like weight management, the question isn't your ability to lose weight, because people lose weight using all kinds of crazy diets. It's really about the long term engagement and keeping it off. And we know from the human social psychological literature, this social support is known to be very effective for people.

And so we basically are applying that to say let's create a technology that's capable of also providing social support. And then we compared that form of social support with just a screen versus the robot. And again, the social presence of the robot seemed to make just a quantitative and qualitative difference in people's experience with the system.

INTERVIEWER: And the art and animatronics part of what you've been doing also, how much were the choices about being cute? How important was that in engaging people? And did you find that that was an important ingredient? The big eyes, all the way through the various robots you did? Or was there something childlike that really works? What were the design elements that you played with?

BREAZEAL: Again, the design of the robot is always informed by the context. What's the scientific question you're trying to understand? What is the nature of the relationship even, between the person and the robot you're trying to facilitate? So the way that you would design a learning companion robot for a preschool aged child, not surprisingly looks different than maybe one that you would design for an adult who's trying to lose weight.

So maybe then it's more streamlined. It looks more like a high-end consumer electronic, whereas for a young preschool child, it may be more toy-like, more fanciful. It always has to be appealing. So the design has to always feel good for a person.

Otherwise they won't engage with it. So the aesthetics always matter. But again, depending on the question and the population, you have to consider all these things carefully.

INTERVIEWER: And then what you've talked about with the weight loss robot, is sounds like the one that you're working on, is it Jibo? Jibo Incorporated. Could you describe that? Because that's something you've left MIT to launch. And I'd love to know how that's working. And what it is.

BREAZEAL: I think a lot of the insights that I've been able to gain from the research at MIT over the years, and exploring these different kinds of social robots in a multitude of different contexts-- when we think about robots today, largely, and I mean kind of the general mindset, is that the value proposition of a robot is physical labor. It can do physical work of some sort. So whether that's manipulating something. Or whether that's navigating from point A to point B. What can you do with that physical affordance?

And so we see many robots today that are really based around the task they perform around that physical capability. And social robotics is really taking a very different mindset and saying, well actually, the core value enabler is engagement. The ability to engage people in a much richer deeper way. And so social robots, as we've designed these again, robots with these different context, I started to think of robots as a new medium. And new medium for content.

And that became this kind of epiphany because right now when you think about robots, they're kind of limited around a certain set of tasks they can do. Because it is around this physical confidence. So you have high-end robots that are for manufacturing, or surgery, or things like that. You have robots in the home that are very niche.

Vacuum cleaning robots, pool cleaning robots, and things like that. Maybe pure toy, entertainment robots. But there's not a robot yet that can do many things for you. But once you start thinking about a social robot as a platform for content, now suddenly you have a robot that can do many, many, many different things for you. Because of its ability to give you deeper engagement.

So it's like content on screens that now is liberated to come alive in this whole new way that's much more holistically supporting all of our attributes. So again, the physical, the social, the emotional, the cognitive as opposed to just a screen that displays with information. It's a social presence. So Jibo, in many ways is the meta. It's a platform.

We think about it as a platform for content. Much like how we think about tablets and iPhones today, it's a platform for all kinds of content where you're leveraging the engagement and personalization and the relational attributes of this medium to be able to make that content and those services that much more engaging for you, and therefore helping you be that much more successful with that content. And so Jibo is about really anything that any developer anywhere would want to create for that robot. So we call them skills versus apps. But we envision Jibo helping people in all kinds of ways.

People of all different ages, from kids around playful learning, or people wanting to age in place in their homes, or family caregivers trying to help manage their remote family care-giving tasks. Or even just getting content like entertainment, sports, weather but in this whole new enlivened way, where the robot, because it is in your home and can interact with you over time, can learn about you and personalize itself for you to give this highly personalized experience of the information and content that you're able to get.

INTERVIEWER: So for example, aging in place. Just give me some specifics. How would we be interacting with Jibo?

BREAZEAL: We did a lot of research interviewing both caregivers and people who are aging in place, as well as institutions who serve those communities. And there it's a fascinating intersection of help me be independent-- I want to do things myself for as long as possible, but I wouldn't mind a little bit of help and support. So whether it's just make it easier for me to get groceries, get a car, keep connected to the world around me, help me stay connected to my children and my grandchildren in a really fun engaging way. So addressing the social as well as the more logistical sorts of things.

For the family caregivers it was also help me make that easier, more frictionless-- help me be able to-- through telepresence be able to engage with my aging mom or dad, to be able to make sure they're OK. But also just to have them be able to engage with my family, with my kids. So a lot of it is around helpful support, but also reinforcing the family unit, the family bond, and the family connection. And this is a really important point to make, because there's a lot of concern or pushback assuming that whenever you talk about a robot, you're trying to replace someone. You're trying to take a person out and put a robot in.

And I think the more lightened view of robots today-- and I think it really started with social robots-- is it's about teamwork. It's about supporting the human network. It's not about replacing anyone. It's about the fact that it adds this other element that helps the whole team be more effective.

So whether that's effective in a tasks sense, or whether that's more effective in a social sense, it's really about creating a technology that supports what people already want and need to do. So we see robots as basically being again, this sort of helpful companion. For the first time bringing together this helpfulness with this companionship, which we're finding is really important to people. They don't necessarily want cold sterile technologies in the home.

Home is special. Home is where the heart is. Home is about family. The family's most important thing in your life.

You want a technology that contributes to that feeling of home. You want a warm feeling technology. So that's really what's so different about Jibo.

INTERVIEWER: Sounds fabulous. Actually in general, all the things you've talked about, you talked about the robot learning too. And just because I'm just naively-- How do robots learn? Just tell me about that. Because I can understand how they're giving to us, but how does that process work?

BREAZEAL: I guess very much inspired and formed by how people learn. Robots learn in a lot of different ways too. There's learning from examples, where the robot may be told, this is an instance of something correct, here's an instance of something incorrect. They get enough of these instances, and they can start forming a model for like what's the idea or the concept of the task that I'm trying to learn. Based on being able to recognize things that belong to the correct category, versus the things that aren't part of that category.

So things like classifiers for recognizing objects falls into the supervised learning paradigm, for instance. You can learn through reinforcement learning, through feedback. So I might want to learn what we call a policy. So given that I'm in a certain context, what's the best action? What's the best thing might I choose to do at this point to help get me the best possible way to the goal I want?

So you're learning kind of this routine, so to speak, of the steps I need to take, and each step of the way, no matter where I might end up, in order that I can successfully achieve my goal. And so the robot maybe getting positive, negative feedback as it goes along and explores, and it's trying things out. And then learns that that was bad, and this was good. And maybe I didn't find anything out here, but later on I discovered something. So I need to figure out how that good thing that happened, or that bad thing can be credited back to the actions that I took. So it's learning again, this little routine, based on positive, negative feedback.

And then the third kind is called unsupervised. So that's kind of the area that's maybe the least developed of the supervised and reinforcement learning. But that's when the system has mechanism, but which is getting a lot of data, and is trying to find structure through the data on its own. So that may be something along using patterns of activity of flow over time of a family, in terms of when are the sweet spots for engagement and activity? So they tend to wake up at this time of day. They're quiet at this time, so the robot shouldn't be making noise at that time of the day.

To kind of fit into the flow of the natural family life. That might be more of an unsupervised learning problem. But again, these are analogs to how we learn. And then we also learn through things like interactional things, like tutelage or training. So you can take these underlying algorithms and you can change how the robot gets that experience.

Whether it's through its own exploration, or whether there's someone explicitly teaching it. What's the nature of that interaction? Is it just positive, negative? Or is it actually much more rich, like the way you might teach someone else? Might there be emotive cues that give that robot a sense of, am I doing the right thing or the wrong thing?

Are they giving me informational cues that really help me build the knowledge base? So you can imagine these learning systems can get much more rich as time goes on. But some of it can be through just straight interaction. The robot may just ask you a question. And that might be a way that it can learn that piece of knowledge. So many, many, many different kinds and ways that a robot can learn and get to know people, get to know about how to adapt and fit into that family better over time.

INTERVIEWER: And then about the science of learning that you are exploring through your robots, can you talk about your recent research, or research in general about working with children and children's learning.

BREAZEAL: When I first started doing the work on social robotics, a lot of the emphasis was on how do you design robots that can learn from people? And really to learn from anyone who's not an expert in machine learning and understands the machinations of the underlying algorithm, but truly anyone. And that's when we looked at all these forms of social learning, like learning from imitation and demonstration. As time went on, basically I think the big moment for me was I became a mom. And I would see my kids grow up.

And again, so much of my work was already so heavily inspired and informed by developmental psychology, but seeing my own kids growing and learning just kind of added a whole other layer of reinterpretation on everything I was doing. And I was becoming much more aware. I mean, I was already aware of it before, but just much more cognizant of the importance and criticality of early childhood learning. And that even if children start kindergarten not ready to learn, they're compromised in their success in school for a potentially a very long time. It's very hard for them to catch up.

And there was a lot of innovation happening with older kids or adults, but not a lot of really groundbreaking innovation happening at this really important critical preschool, kindergarten age. And so my kids were right around that age too. They were three, four years old. And I started developing a whole host of technologies where now the robots facilitated children's development and learning. So from people teaching robots, it was really about robots helping children to learn and grow.

And we explored a whole range of topics from kindergarten readiness, second language learning, oral language development, those critical readiness skills. We looked at ways of thinking about-- my kids were playing video games and it was so frustrating for me because they're sitting there, and they're pushing a buttons and stuff. And I'm like how can we make digital play feel more like playground play? So we started mixing virtual reality, augmented reality environments with projection to physical spaces with robot characters that could literally go "into" the screen into a virtual world and out, to make kids run around and play and move like a playground, but in this world. They're in the virtual world now. Where the virtual world was physical around them because it was all digital projection. To looking at--

INTERVIEWER: No more flat screen.

BREAZEAL: No more flat screen. It's like it's all around you. And then thinking about well, so computational thinking. So can children learn about these concepts at a much earlier age by "programming" these robots, but through teaching them through rules.

So we literally had these vinyl stickers, mediums that children are already familiar with at this preschool age, but being able to almost socially "teach a robot" by literally putting these stickers, that were a syntax, on a vinyl sticker board, and showing it to the robot. The robot would look at the "rule," and then learn it. And that it would do that behavior.

So exposing children to the idea of rules and sequencing, and early concepts of computational thinking. Kids loved it. They just love it, because it's in the context again, of social interaction. It would almost teach the robot how to interact with them.

And so much of preschool is about socialization. So it's a topic and it's a context that's very developmentally appropriate for them. And they just love interacting with a social robot. So we looked at all of these things.

We've been looking at children in the context of pediatrics. Around coping with emotion and stress through social robots. So having a pediatric companion robot that would interact with a child to help alleviate the sense of stress. We know that's so key to, not only a positive experience in hospitals, but to healing and recovery. Even potentially compliance.

So all of these areas we've been really thinking about the context of social robots and children. And I think the sweet spot there was just because social robots are physically present, it's very different than these individual devices of these screens where we all see it. Even my kids, they pull out their tablet, their nose goes in it and they shut out everything around them.

With a social robot, suddenly all that information is co-present in an interpersonal group dynamic way. So we sell parents they could play with their child with a social robot, where it was a group dynamic. Where the robot was almost facilitating the ability for a parent to be able to guide, praise, scaffold basically their own child's learning was like supporting that social dynamic. So for me, it's a very exciting medium for children because it brings the social world together with the digital information. That's much more are in line with what children do and need as they're learning.

So being physically present, being socially present with others around them, but facilitated by the flexibility of digital media and the ability to capture data about what they're doing to be able to form models about understanding what they're learning, where their areas of strength are, where you may want to guide them next to have these personalized learning experiences. So a lot of the themes that we've been exploring is technology has the opportunity to bring the power of that one on one personalized interaction to any child to kind of optimize their learning trajectory. And to do that again, in the social context that's so critical.

And that's very different again, for this lecturing paradigm that we're all familiar with. It's like through AI we can bring that personal experience of personal tutoring, which we know is much more effective than just broad lecturing, if you can do it. AI can bring that to any child. So looking at the personalization aspect of it-- and then interestingly what we've also been discovering is because these robots are modeled not as tutors, but as learning companions, they're actually modeling a peer to peer interaction, you see these really intriguing dynamics where children learn a lot of things from each other.

So there's the content. There's the curriculum, like vocabulary. But there's also things like learning attitudes. So we did a study where we created a curious robot. So a robot that, as a learning companion, as a peer, interacting with a child, would exhibit pro-curious behavior.

If something was wrong it would say things, oh, that's okay. That's how I learn. Or, oh, that was cool! I love discovering that. So communicating this positive attitudes about learning.

And we had a control condition where a robot would do the task and say positive things, but it wouldn't be these pro-curiosity behaviors. And lo and behold, what we found is that when children interacted-- and we did a randomized control trial-- children who got the curious robot themselves exhibited more pro-curious behaviors than those who didn't. So again, modeling the robot.

And we've seen this even through storytelling. As a robot engages the child in a storytelling task, first the robot may tell a story, and then it invites the child to tell a story. And you could imagine over time, you could start to customize the stories a robot tells to the child, and vice versa, to optimize their vocabulary learning.

What we were also starting to see is children were starting to model the robot stories. So it's like just because of this is peer-to-peer dynamic, that's almost with this sort of like furry robot companion that is kind of like your dog, but got much smarter in some sense. It sets up this really fun, not afraid of being embarrassed, not afraid of losing face by making mistakes in front of this other creature, but being able to explore new things together where they actually learn again, not just the curricular aspects, but these other attitudinal and these other kind of behavioral things.

So it's really fascinating to think about how to again, design these systems to really support children's learning and development. We've been working with a number of schools now in the Boston area around storytelling, around vocabulary, around second language learning. Tremendous interest by the teachers in doing so. So that's been work that's been extremely rewarding. And again, I see these robots adding a really fascinating important new dimension that you don't see when it's just a flat screen.

INTERVIEWER: And I want to explore that more, but I'm just curious just to flip it back a second, and how this work-- you said the big epiphany was when you became a parent. How has your parenting been affected by all the things you've been thinking about all these years? Did you see the children in ways that you-- well, it's hard to know what you would've otherwise-- but your experience with them as infants? And did you have things you wanted to make sure to do with them because of all the things that you'd been doing at that--