Dr. Maja Mataric is at the forefront of innovative therapies for those with Autism Spectrum Disorder. Combining her skills in engineering and her passions in helping people, she has worked for over 20 years to improve human quality of life. To improve human life, she has created human-centered technologies, including socially-assistive robots for those with ASD and anxiety. Making the world a better place one robot at a time, Dr. Mataric empowers. She not only empowers children with ASD, but she empowers her students and most importantly women. I am so excited to see the future of socially-assistive technology and Dr. Mataric is truly an inspiration for us all.
The Real Spectrum: I wanted to start by saying that when I was researching more about you, I saw the video titled, “When Dinosaurs Ruled the Earth.” I absolutely loved watching it. It was very powerful when conveying your mission and goals.
Dr. Maja Mataric: Yeah, that was actually in an Autism Film Festival. The full credit goes to the film crew at the Viterbi School of Engineering at USC but we made the robots for that video.
TRS: Well, it was very cool. That actually leads to my first question. When did you become passionate about engineering and building robots?
M: Honestly, I would say that I wasn’t really passionate about it, I was more just really interested in that field intellectually. Since college, I’ve always been interested in AI (Artificial Intelligence) really before the rest of the world was interested in AI. But I was interested in AI and robotics for behavior in the real world. It wasn’t really until almost 20 years ago that I became interested in and became passionate about making robots to help people in the real world, as soon as possible. I was aiming to specifically help individuals with special needs or neurological disorders. Those goals separated me from the vast majority of researchers in robotics back then, and even still today. We ended up starting a new field: socially-assistive robotics. It’s been incredibly rewarding to work with amazing students and collaborators to see it really grow into something that hopefully will become increasingly accessible to real people out in the real world. So, that’s when I started becoming passionate when we created this field.
TRS: I’ve never actually heard of using socially-assistive robots before. When I came across your work, I was very impressed. It’s such a modern idea but has so much promise. How did you apply this idea to individuals with autism?
M: So, this is now no longer a new idea. We came up with it, along with some of our collaborators at Yale University. That was 20 years ago. It’s not new. More people should know about it. The fact that you didn’t know about it, especially with your family being a part of the autism community, means that this field needs more support to grow. I will say that there is so much hype around AI these days and yet, you really don’t hear about the stuff that could actually help people. Setting that aside, we’ve been working on this for 20 years. The idea is simple: how can we create machines that actually help people as opposed to machines that you can create just because you can, or because they’re entertaining or would sell? The way that I started to work with children with autism is by talking to them and their families to try to understand their needs. There are many ways of helping. There’s so many because the need arises not just for individuals on the spectrum, but also for their families, parents, caregivers, school districts, and therapists. So, I think this really is a wonderful field, because now finally, we have people who are developing all kinds of different technologies. Referring to your question about where we see robots for autism – and I mean socially-assistive robots not to be confused with functional robots that act like an autonomous car – we’re talking about socially-assistive robots that help children with social interactions. So, how does that happen? Autism, in particular, often consists of disrupted social behavior and social skills. People may ask, “well, why don’t we just help with social skills with screens?” Obviously, it’s easier to create an app on a phone than it is with a physically present robot, even though the robots now are quite cheap and cute. But the reason that we need to do it with a physical robot is because that engages the user’s brain much more whether the user is on the spectrum or not. We’ve worked with many other populations as well. We really want something that’s going to be available to people, be affordable, be present, and actually be effective.
TRS: That’s very interesting. You mentioned earlier on in your response, that all these people are building robots simply to provide for consumers and for entertainment purposes, but you guys are genuinely trying to do something to help a community. That’s what matters.
M: I think people develop things that they can support. So, that’s one of the issues is that you see robots being developed for places where they can get resources to develop them and also because they will sell. I think what’s poorly understood is the scale of the autism community. I don’t think of it as a market, but I think of it as a need and how much can be done. That’s generally true for a large number of communities that are just not noticed, and they should be.
TRS: Right. I also really admire how you acknowledge that you are using socially-assistive robots, rather than autonomous placeholder kinds of things.
M: They actually are completely autonomous. But the question is whether they’re helping someone physically or socially. The idea that robots doing physical work is not the only way that you can help people. For some people, the social part is so much more needed. That’s the part that’s lost and we need to change that.
TRS: That makes a lot of sense. There’s different ways to help people and that needs to be understood. So, my next question is what do you think are the strongest benefits of robot therapy and maybe some challenges?
M: The main benefits definitely revolve around the fact that when it comes to behavioral changes of any kind, especially in autism, it takes a lot of time and practice. So, what’s important is creating opportunities for individuals on the spectrum to practice those skills in a pleasant, game-like way. The goal is to create this technology, which serves as a multifaceted role. It acts as a helper. It acts as a therapist. It’s a friend and a buddy. It’s something that’s available at any time. It’s role helps to unburden parents who are working hard and want a break but also know that during that break, their child is doing something that they know is actually good for them. The other benefit is that the robot really fills a role of social peers because unfortunately, children on the spectrum sometimes don’t have many human friends or other kids won’t play with them. So this is something that until that improves, the robots are something to fill the gap. That’s the idea. The idea is to increase opportunities for practice and learning and training, and to remove some of the loneliness.
TRS: That’s amazing. One thing can be so beneficial across so many areas. Do you think that the use of robots can also be as beneficial to adults with autism?
M: Yeah, we’ve already started working with adults. We originally started working with adults as a whole, not specifically in autism but adults in an age group. We’ve done a lot of work with stroke patients and elderly patients who may also have Alzheimers or Dementia. We’re now working with kids who have anxiety or depression. Robotics is a very large and broad field and in general, focuses on helping people across the age span and across the ability span. So that’s the first thing to realize also because people tend to focus on children with autism because that’s more compelling. However, of course, those children grow up to be adults and are largely forgotten and neglected. I think there’s a large intersection between what is done with children with autism and what can be done for adults with autism. You can still use socially-assistive robots to create opportunities for practicing skills and providing companionship and support. That aspect is not any different but what is different is the content of skills being taught. We have discussed with people about how we may be able to help young adults with autism learn how to do job interviews. So, whether they’re learning how to do job interviews, learning how to make eye contact, or to take turns, it’s really not that different. My point is that this field really does cut across all ages, especially because everyone can use a helpful companion.
TRS: I love that idea. No matter how old you are or no matter how neurodiverse you are, you can rely on the use of robots for so many different things. Relating to that idea, since autism is a spectrum, how can you adapt each robot to accommodate each child? I don’t know the specifics of AI so could you elaborate more on that?
M: Well, there really is no such thing as the specifics of AI. It’s a massive field. It would be like saying, “I don’t know the specifics of math.” The underlying mechanism is learning. It uses machine learning to identify what specifically describes the needs of a particular child and then giving them more support in that area. There are many different approaches for that. Some of these are being used, for example, in intelligent tutoring systems where the system can figure out how hard of a problem it should give you. But that’s just scratching the surface. We’re more interested in things like how we can give the child support and reward them to keep them motivated, even if the child isn’t doing well or is getting bored. We’re interested in how we can teach the child social skills, in addition to doing things, because in a typical learning environment you learn both. So, when the child is answering a question from a teacher, they’re not only learning the subject material, but the child is learning how to listen, wait for their turn, make eye contact, etc. All of that also has to be personalized. If the users of the robot have a problem with a certain skill, then the robot will notice that and work on it, rather than working on something that the child is good at. There’s a lot of promise.
TRS: That’s fascinating. Do you have undergraduate students help you in your lab?
M: So if you had looked at my website, you would have seen that I have about 35 undergraduate students and nine PhD students. We always have a lot of undergraduates, because they’re very passionate and there’s so many of them. My lab will always have a lot of undergraduates, and they’re just absolutely wonderful. Some of the projects that PhD students are leading are so impactful and important. But anyway, I want to make a point about the earlier question that I forgot to mention that I think is very important. The idea that the robot should be able to figure out what each individual user needs is very important, because a lot of machine learning today is still about learning from a lot of different people, but not personalizing to each individual. And with autism, we really need to personalize to each individual because each individual is unique along many dimensions and with their challenges and strengths.
TRS: Yes, definitely. I’m also sure that this idea of machine learning having to adapt has to apply to so many other various areas. For example, with stroke patients, severities may range among different people.
M: With stroke, for example, if you have medium severity, like loss of tactile sensations in your right arm, or another person with medium severity, will have a lot in common, whereas the same can’t be said for autism. It’s a very large number of symptoms and a large scope of severity within those individual symptoms. So, you can’t generalize exactly. People always say that once you know one person with autism, you know exactly one person with autism.
TRS: Yes, of course. The machine learning being used has to adapt to each individual person. Do you think you can talk about the technicalities of machine learning?
M: It’s really hard to explain this in a non-technical kind of way. But what I would say is that there are many different approaches in machine learning. There are many different methods as I’ve mentioned. So we use like four different methods of machine learning in order to figure out different things like, you know, is this person paying attention? Well, what does it mean to be paying attention? For some kids, it’s eye contact or for other kids, they’re looking out of the corner of their eye. We know that attention is a complex thing. So just determining attention means that the machine needs to learn a model of the child, then figuring out how much feedback to give them. You also need a personalized model of feedback. So there’s just a lot of things for the machine to learn and adapt to the individual child. There also exists reinforcement learning and it’s important to understand that there is no one size fits all in machine learning and therapy. Sometimes people in machine learning like to choose their favorite tool and think that it will solve everything. And autism is a great example when there isn’t a one size fits all for anything at all.
TRS: Yes, it’s a perfect example. It’s also very interesting because your lab is characterized more by engineering, but you incorporate psychological concepts that come from different schools and perspectives. That’s really cool. You obviously have a lot of collaboration from different fields. This is kind of irrelevant to this topic, but would you consider applying for a patent on these robots so that anyone can have access to purchasing them?
M: That is a lovely question. I’ve already started a company that was heavily backed by investors, however, they didn’t want to make it about autism because they think it’s a niche market. They want to make products that will sell to everybody so that they can make more money. The idea of doing a patent is not really so simple. Investors are very conservative, and they’re only interested in making a huge profit quickly. That’s why you don’t see a lot of companies for things that help sub-communities like autism, anxiety, etc unless it’s software, like apps. Everyone can make an app. However, a lot of apps don’t work, right? Some do, but most don’t. Apps are cheap to make, relatively speaking. So the idea of starting a company has not only occured to me, but I’ve done it. It’s probably easier to do a nonprofit, or something like that. A community has to get very vocal in order to get investors to actually back the research in this area.
TRS: That’s very unfortunate. It’s sad.
M: Unfortunately, yeah. That’s how capitalism works. I think we have to think of an incentive structure that would incentivize people to do good. When you can match profit and good, it works really well. Sometimes, people don’t understand that. It can be very unfair. I’m a researcher in academia so I have the honor and the pleasure to work on all kinds of problems. I don’t have to worry about making other people lots of money. That’s a privilege. The challenge is how do we get what we know would work for real people. The challenge isn’t making it work. We can make it work. It already works. But, someone has to put some serious money towards it.
TRS: Right. How much money does it cost to make each robot, on average?
M: A robot itself could cost literally $20 to make. But that’s not what it’s about. My lab is actually going to start open sourcing how you can create robots for about $200.
TRS: That’s amazing.
M: But the point is, you still have to know how to program them, etc. So you know, we’re going to open source as much as we can. But in the end, for a company, we have to raise money and then whoever gave the money has to get it back which affects your profit margins. So, there’s a whole lot of complicated business stuff and business models. Business models operate on profits, revenue, and things like that. It’s tricky. But, I do believe that as younger generations learn more and more about computing, and learn more about robotics, we’re at an excellent path forward because we can open source and make a lot of things free. The main thing is who’s going to program the robot. It takes a combination of understanding autism and having therapists’ input and it’s coming. It’s coming. I’m very, very optimistic.
TRS: That actually leads to my next question, what are you most excited for when thinking about the future of socially-assistive robots?
M: I am just always, always really excited by the ideas that my students bring to the table. Young people have the best sense of what’s going on because they’re living through the new realities and the new challenges. I think that’s just really infinitely refreshing. So if you had asked me 20 years ago, or maybe 25 years ago, what was my life going to be like in five years, I know that we were going to do this. But once we discovered the need, we have continued in this field. And so I’m never going to pivot out of it. Right now we’re very interested in trying to do some good around anxiety and depression because it cuts across many things, including autism. It’s affecting so many people, all people but particularly young people, and we’re very motivated. That’s a new direction that I didn’t have maybe five years ago, but I have now, and I am excited to see what my students and I will do with that. I think the most exciting thing is always young people. I have the luxury of working with lots of wonderful young people. Yeah, that is probably the best part. Practically speaking, I think that the need becomes ever more obvious which will create a market where everything around robotics could get cheaper. Right now materials and processors are expensive because of all the unrest in the world, which is really unfortunate. I’d like to think we’ll get over it, or one hopes, or somebody will invent something cheaper. It’s just so close to doing a lot of good that I feel that for one way or another, they’ll get socially-assistive robots out there. We have to.
TRS: I agree exactly. If there is a greater understanding, it’s bound to get out there. Final question… what do you have to say to girls wanting to pursue a career in STEM?
M: I have a lot to say about that. And I have a mentoring page on my website which started out as being for girls, but then I realized that we can’t only focus on one gender. We need to include everybody. And so there’s an entire webpage off my homepage or mentoring that says it all. I really encourage everybody to go over there. But what I would say is my main piece of advice is, don’t let anyone tell you you can’t do it. You have to believe in yourself because you can do it. But the way you will believe in yourself is to pick someone else to help. So autism is a great example. I remember having a student I think more than 10 years ago. She was a high schooler. She waited for a long time to get to talk to me after some talk that I had done. Everybody else who waited to talk to me was like, “I’ve done this and this and this and I want to get in your lab.” And she just waited and waited and waited. She was the last one. Then she just said, I have autism and I will do anything to get into your lab. You think she got into my lab? Oh, she did. But the point was, it was that commitment. And it was about helping others. There’s so many people who will come to me and say, “I have these amazing skills in machine learning and I will come into your lab and I will do amazing things.” I’m not interested in that. What will you do for someone else? If you come in and tell me that you will do something for someone else that resonates with you and I think that when you’re fighting for someone else, that’s when you have all the confidence. So that’s my advice. My advice is to find someone to help. And that will drive you and you will do great.
TRS: That’s amazing. I love that so much.