CECS Associate Professor Samir Rawashdeh creates interactive robots for his kids to play with. It’s taught him something about what it might take to make human-robot interactions meaningful.
Associate Professor of Electrical and Computer Engineering Samir Rawashdeh builds robots for his kids, many inspired by Star Wars characters. Photo courtesy Samir Rawashdeh
For the past few years, Associate Professor of Electrical and Computer Engineering Samir Rawashdeh has been building robots for his 7-year-old daughter, Maia, and his 5-year-old son, Adam. One of his first creations was a modified Roomba, the popular autonomous floor-cleaning robot, which he equipped with a vision system and topped with a googly-eyed Lego head and multicolor mini pom poms for hair. The souped-up Roomba would primarily look for and follow the kids around and then stop at a safe distance when it caught up to them. Even though its programming was simple, it was a huge hit with the kids. Maia would squeal as the Roomba “chased” her around the room. She’d ask it to accompany her on play missions around the house. She also figured out pretty quickly that she could trick it by hiding in her play tent, outside the reach of the Roomba’s camera.
Rawashdeh's second robot had a little more polish. Using 3-D printing plans he got from a Facebook group devoted to such things, he built a full-size replica of D-O, a scrappy Star Wars robot with a ducklike, pivoting conical head mounted to a thick off-road wheel. He then used his electrical engineering expertise to make it move. While the Roomba was autonomous, Rawashdeh opted for remote controls for D-O, thinking it would give the robot a wider range of intentional movements. With a radio controller in hand, Rawashdeh could make D-O not only chase the kids but run away from them. It could move its head in all directions, including shaking up and down for yes or sideways for no. When the kids would hug it, D-O could give them a catlike head love tap right back. With Rawashdeh’s latest version, D-O can talk.
Rawashdeh's D-O in action.
Rawashdeh says he initially started making the robots as a fun pandemic engineering activity that he and the kids could do together. But as he watched the kids play, he became fascinated by how they were interacting with the machines, especially D-O. The kids figured out pretty quickly that their dad was the one actually animating D-O, but in many ways, they didn’t care: They still reacted to D-O as though he was a “free agent.” The kids would talk with D-O and ask him to do things with them. They would give him affection. “One time, when I took D-O into one of my classes, Maia texted me and said, ‘Tell D-O that me and Adam love him,’” Rawashdeh remembers. Even without his own agency, D-O had become “somebody” to them — at least enough to miss him when he wasn’t around.
Associate Professor Samir Rawashdeh
So what’s going on here? Or as Rawashdeh frames the question: “When does a bunch of plastic and wires and batteries and motors become more than that?” It’s not a question he can claim a definitive answer to, nor likely can anyone. The modern discipline of human-robot interaction is still very new, and you’ll find a wide range of ideas about what ingredients are essential to humans having meaningful social interactions with machines. Rawashdeh’s own take is that there is still a lot of exploratory trial and error going on among the designers, engineers and philosophers who think about these things. Most robots intended for social human interactions seem to contain at least some suggestions of biological anatomy. For example, Miko, an AI-powered robot that's being marketed as a companion for kids and who can even teach them social behaviors, has a friendly face. The engineers behind Paro went with a full biomimicry approach when they built an interactive baby seal to serve as a therapeutic companion animal. But there are also examples that completely buck this trend. The engineers behind Yolo, for example, which is advertised as a “creativity boosting robot” for children, actually let kids design it. The result was something that looks like an artsy bedside table lamp. It has no face and no eyes. Other than the wiry plastic strands shooting out of the top of it, which sort of look like hair, it’s not very creaturelike.
In his own informal trials with his kids, Rawashdeh noticed a couple qualities that seemed to be important for them to connect with the robots. First, movement helps, and specifically, an ability to move in response to something the kids are doing. Both the Roomba and D-O can do this. In the Roomba’s case, its movement is fully autonomous. But interestingly, the kids seem to be more engaged by the dad-controlled D-O, who, with its nods, sounds and affectionate head bumps, has a much greater range of responses. Rawashdeh’s observation that the kids know he’s animating D-O yet don’t seem to care is actually supported by some recent research suggesting that children’s tendency to anthropomorphize robots can be a powerful bonding force — even when they know the robot is not an independent being with its own inner psychological life. Rawashdeh also comes down in the camp that thinks looks matter, but it doesn’t take much. Even D-O’s hint of ducklike anatomy is enough to give it something that can be interpreted as a face. Interestingly, Maia has more or less the same take on all this as her dad. When Rawashdeh asked her if she could ever be friends with a robot, she quickly answered in the affirmative, then clarified — as long as it could follow her and was “very cute.”
Surprisingly, one thing Rawashdeh thinks is probably not that essential to forming a bond is speech. For example, they’ve had an Alexa in their house for a long time, and the kids will occasionally ask it questions, but they don’t think of it as a “playmate.” He also points out that R2-D2 from Star Wars is beloved but does not speak, and we form very strong bonds with our pet animals, who at least don’t speak our language. He thinks this peculiarity of speech, or conversational written language, might also explain why, despite some initial fears, technologies like ChatGPT seemingly haven’t unleashed a plague of socially isolating human-chatbot friendships. “I think part of what is going on here is nobody cares what a machine thinks, which is a key part of what makes our human-to-human relationships meaningful,” Rawashdeh says. “I think this is even the case with social media, which sort of feels like a lonesome experience between humans and technology. But it’s really more of a human-to-human interaction that’s mediated by technology. Why do we post our vacation photos on Instagram? It’s to show off where we’ve been to other people. If you had a robot liking all your posts, that wouldn’t be satisfying.”
This line of inquiry also, of course, begs some important questions about whether such human-robot relationships are good for us. Could they, for example, further transform or replace the human-to-human social interactions that have already taken a beating from technologies like television, the internet and smartphones? And should we be worried, in particular, about kids forming bonds with machines? Research in this area is still very new, but it is producing some interesting findings. On the positive side, there’s work suggesting that children with autism may actually learn better from robots, due to their more predictable responses. Paro, the therapeutic robotic seal, was shown to produce many of the same effects of traditional therapy animals. At least for now, a headline-grabbing study showing that human society could be a risk should we choose to befriend robots doesn’t exist. Rawashdeh says his kids, at least, don’t seem to be overly attached to D-O. Playing with their “very cute” droid is not an everyday activity at their house. Maia and Adam still like to play outside.
For Rawashdeh, this experience building robots for his kids has, however, inspired him to do some more formal work in this area. For one of his next projects, he’s experimenting with building a therapeutic robot that would essentially be a good listener, with a goal of potentially addressing certain gaps in the availability of mental health services. Powered by a modified large language model, it could engage people who, say, were having a bad day, with supportive, helpful humanlike dialogue. For the body, he’s thinking something suggestive of an owl — an animal associated with wisdom. It likely wouldn’t move around the room, but he would give it the ability to move its head and large, empathetic eyes, so it could appropriately make and break eye contact and “wouldn’t just stare at you.” “I’m thinking about the shortage in mental health services and the current loneliness epidemic, and how there are many people out there who could benefit from a helpful companion,” he says. On the other hand, he wonders if technology can be a salve for social isolation, which, in many ways, has been caused by technology. “That’s something I guess we’re still figuring out,” he says. “I think we still lack the language for describing what is going on,” he says. “Like, with music or art, we have established traditions and genres. We know, more or less, the key ingredients that go into making a Cubist painting or a certain type of song. But what makes a robot that humans will actually connect with?” On the front, Rawashdeh thinks we may still be shooting in the dark.
