Professor Cynthia Breazeal sits at the helm of the Personal Robots group she founded at the MIT Media Lab. She is a pioneer of social robotics and human-robot interaction – robots socially capable of creating an emotional connection and interacting with people in natural ways. Her robots have been applied to various fields, be it education, chronic disease management, pediatrics, aging, or emotional wellness. NTT DATA is working with this research group to develop a care service that utilizes social robots for the elderly(*). Professor Breazeal offers thought-provoking insight into what the relationship between technology and human society could look like.
--You coined the term, “social robots,” but could you tell us about its concept?
Social robots are designed to be helpful companions. Not only can they perform practical (physical or information-based) and entertainment tasks like current robotics technology, they can also interact with people and provide the support they need over long-term encounters. Going forward, we hope they will be able to autonomously engage people socially and emotionally to help people to learn and flourish. We design robots to act as teammates and collaborate with people toward a common goal. Humanistic experience coupled with meaningful value will pave the way for social robots to enter the home and help families.
--Are technologies such as social robots already being implemented?
Since the emergence of Amazon Alexa back in 2014, we have been living with AI as an ambient service. Today’s children are growing up with smart speakers and as a result, are talking to AI on a daily basis. Yet we know very little about the long-term impact or consequences of this. We’ve already heard concerns around social media and the decline of empathy, so people are becoming increasingly anxious about what our future with AI is going to be like. When robots step out of the research lab and into people’s homes, what tasks should they be performing, how does this complement what people need to do, and how will all this promote people’s well-being? These are things we want to look at in great depth.
--What sets social robots apart from other AI or robotics technologies?
Currently, we have many services that sense and analyze various data, provide information, and connect us. Although such services can be useful, they are highly transactional, they do not make people happy. Getting the weather, calling an Uber, ordering from Amazon are certainly things people want to do with current digital assistants. But I’m far more concerned with how personified AI systems, like social robots, will help us address the big challenges we face as a society, both now and into the future. Education. Healthcare. Aging.
I want to create systems that help people in a meaningful way, that are human-centered. How can we help people be more creative, be life-long learners, and live healthier lives? In terms of health management, no matter how many articles say exercise is beneficial, for instance, it’s really hard to change people’s behavior to exercise regularly. We also know that social support is one of the most effective ways to persuade and support people to change. Social robots can help create an environment for positive change by offering social and emotional support that helps people to make beneficial decisions.
--Each year NTT DATA publishes “NTT DATA Technology Foresight,” which highlights advanced technologies and societal trends we believe will impact businesses in the coming three to ten years. And under the section, “Natural Interaction,” NTT DATA paints a future in which conversing with machines is quite commonplace and they even help people be more creative and learn new things. In this respect, social robots are beginning to have an impact on education, are they not?
Before entering school, young children learn through interacting with people and through play. Social robots, likewise, learn by playing with children rather than lecturing them with curricula. We are particularly interested in peer-to-peer learning where you learn with your friends and you teach each other things you know.
Moreover, you may feel embarrassed if you make a mistake in front of your friends or your teacher. With robots, you may not worry as much. We see that children think of our social robots more like a really smart pet who can actually talk and play digital games like a friend. Children can explore, take risks in trying new things, make mistakes. The robot does, too, and models positive learning attitudes like curiosity and growth mindset. Over time, we see children also behaving with greater curiosity and having a growth mindset.
--Social robots may help change the current educational environment.
At the moment, in the United States, roughly only 40% of preschool-age children are enrolled in quality preschools. As a result, many children enter Kindergarten not ready to learn. We know that it is very hard and very expensive for these children to catch up. Quality education is the key to future success. All children deserve a quality education, and I believe we can use social robots to help address this social injustice. Imagine a scalable and affordable AI-based service, delivered via a social robot, that provides personalized education optimized to each child. Every child could enter Kindergarten on the right foot.
Children grow the most through effort, experience, and practice. When they have a growth mindset and persevere through challenges, they learn and grow. People with a fixed mindset think that hard work doesn’t help you improve, either you are naturally good at something or not. So they tend to give up once things get hard. We are discovering that social robots can help children develop this growth mindset. In other words, education and educational technologies shouldn’t be just about curriculum, rather, we should create a supportive learning environment in which children can develop positive learning attitudes such as curiosity, growth mindset, and creativity.
Social robots born in Professor Breazeal’s lab
--That’s very interesting. Do you expect these social robots to be used in homes as well?
Yes. Social robots are not only great for 1:1 interaction, they can also help promote family connection. For example, when the robot is reading a story aloud to children, it could naturally encourage parents to take part as well. The robot could model dialogic storytelling; it’s the best way to read and ask questions during storytelling to help children learn. Our conjecture is that if a robot models dialogic storytelling, and parents see how engaged their children are, it could persuade parents to do the same. As a result, parents learn how to best support their children’s learning at home while also feeling more emotionally connected to each other.
We plan to analyze the effect social robots have on the people in the home environment through carefully randomized controlled studies. One of the characteristics that sets our lab apart is that we co-design with stakeholders so they have a voice in the technologies we develop for them, and we study the use of the technologies in real-world contexts, like the home, so we can iterate and improve the design.
--I’d like to ask you about your technology. What kind of technologies make it possible for your social robots to provide personalized content?
With respect to data sensing, our social robots adapt the curriculum by tracking how children play educational games with the robot — we record what children say, detecting children’s facial expressions, biometric signals of the skin, and behavioral data such as gestures. AI and machine-learning programs are constantly running in the background, so the robots continuously learn from the input data to optimize their behavior to promote children’s learning and engagement. The data that the social robots capture are not only about how well they are acquiring new skills and knowledge, but also subtle data that reflects the children’s psychological and emotional engagement. We also assess the quality of the interpersonal relationship that develops. We are learning that the better the relationship between social robots and children becomes, the more children socially emulate and learn with them. For instance, children will repeat the target vocabulary and phrases that the robots use even more, and this can help to pull children along a learning trajectory as the robot introduces more sophisticated words and syntax.
--Social robots can be applied to healthcare as well, can they not?
Today, there is a wide range of healthcare services available, from medication adherence to analysis of physical data. But most of these services merely deliver information unilaterally. We hope to create an environment for healthcare where people and robots can have a dialogue. We see the robot acting more as a personal health coach.
For example, people may be skeptical of their medication; they may be wondering if it’s really effective, so they don’t take it as much as they should. They don’t want to confess to their doctor, though, because they want to be seen as a good patient. We are finding that people tend to be more open and honest with robots because they don’t feel the robot is judging them. More accurate monitoring of patient behavior helps doctors understand what patients are really doing and feeling, and then the doctor can adapt the treatment to the patient.
--But what about privacy?
That’s an important ethical issue that pertains not only to healthcare but to all aspects of life with AI. The transparency of how the data will be used, and in what context privacy will be respected – these are things that need to be shared with each individual and they should be given a choice. In this respect, we need to design systems even more carefully. On the other hand, robots that can learn personalized models could potentially pick up changes that are out of the norm. For instance, people may not realize that they actually have symptoms of depression. A robot who really knows you could pick up on these changes and alert you to them so the person can take appropriate action.
--That kind of feedback is important. Will these robots be used in facilities?
We did a three-week study of social robots in an assisted living community. And what we discovered there was that robots acted as social catalysts, encouraging people to communicate and have face-to-face interactions. It’s ironic that people who live in assisted living facilities can be lonely even though they are living with other people. They may not have spoken to each other much before, but after the robots were introduced, they began to congregate more and interact with each other more. It was a very interesting phenomenon. It was also telling in terms of the social psychology of engagement and robots as social catalysts.
In the lab, not only do they program robots, but they assemble them as well.
--Why do you think robots can become social catalysts?
The key is for robots to be physically co-present in the same room with people. Most of today’s AI technology is invisible; they work behind the scenes for text analysis, voice recognition, etc. But when the robot is actually present, it’s more engaging and persuasive. When you look at social media, you can communicate with hundreds of friends, and you’re connected to so many people. Having said that, we live in a time of hyper-connected loneliness. We have never been more connected yet we have never felt more alone. What are we not getting from a Facebook post or 140 characters? There are many people nowadays who may have hundreds, even thousands of followers, but feel profoundly lonely. Perhaps this is indicative of the importance of face to face, direct sharing. In this day and age, social robots can help people feel less lonely and deepen interactions.
Furthermore, we’re now in the time of lifelong learning. It’s not only children who are learning something new. Learning is important for people of all ages. And the opportunities robots provide to learn will be relevant for young children and our oldest citizens alike.
--It’s all based on a very attractive concept, but until recently media had fanned the flames by talking about how AI could take away people’s jobs. How do you think the perception of robots has changed?
As I already mentioned, what we aspire to create are not robots that would do work in place of people, rather our robots are human-centered and are designed to help people communicate and expand their capabilities. I started this work 20 years ago. I’m a big Star Wars fan, and I grew up watching R2D2 and C3PO. They had rich personalities and deep relationships with their partners. Like Japanese anime and manga culture, my fascination with robots started with science fiction. But 20 years ago, when I started talking about social robots helping humans, robots with social and emotional intelligence, and how they could be friends, people in the Western culture didn’t get it.
From the left: “Leonardo,” “Tega,” “JIBO,” and “Cyberflora.” Leonardo was the second social robot Professor Breazeal unveiled after Kismet.
Of course, values can be defined by culture, but good or bad, media has significant influence. We always want to be mindful of the context presented by the media. Some jobs, rightfully, may be supplemented by technology, and the nature of the job may change. But hopefully these technologies will only change the nature of what we do, and help amplify us, and allow us to concentrate on things that are more creative, and ultimately help more people to prosper doing more fulfilling work.
--That is a wonderful vision. Thank you for your time today.
(*) NTT DATA is conducting joint research to improve medication adherence (which refers to whether patients agree with the treatment policy and take the appropriate medication) by utilizing social robots. The combination of Media Lab Signal Kinetics Group’s technology, which detects human movement using RFID and wireless signal analysis, with the social robotic technology of the Personal Robots Group, patients’ medication adherence can be easily detected, and based on the information gathered, social robots can proactively remind and encourage patients to take their medication. There are hopes to apply the findings to services that safeguard the elderly.
RFind: Extreme localization for billions of items