At a bakery, a robot kneads bread dough alongside a human worker. After setting the dough aside, the robot moves on to its next task. It watches once as the worker demonstrates how to pipe frosting on sugar cookies, then expertly takes over the job.
At the day care across the street, a robot with a mechanical smile and friendly voice helps a child wipe up spilled milk at snack time, then turns to stacking blocks with toddlers in the play area.
These scenarios might not be far off, if social robot developers have their say. "Robots that engage with people are absolutely the future. There's no question that's where robotics is moving," says Brian Scassellati, PhD, a professor of computer science, cognitive science and mechanical engineering and director of the Social Robotics Lab at Yale University.
The first autonomous robots—those capable of operating independently from human operators—were designed to tackle potentially hazardous tasks, often far from people, says Cynthia Breazeal, PhD, director of the Personal Robots Group at the Massachusetts Institute of Technology Media Lab. They explored deep oceans, the interiors of volcanoes and the surface of Mars. The next generation of robots will have to do something much more complex than navigating the landscape of the Red Planet: fit into everyday human life.
Breazeal likes to joke that for roboticists, the home is the final frontier. But designing robots that can interact with people in homes, schools, hospitals and workplaces is a serious scientific challenge. Designers will need to draw from psychological research in areas such as communication, perception, social-emotional intelligence and theory of mind. "We want robots that can learn from us, that can teach us, that can be aware and anticipate what we're doing," says Deborah Forster, PhD, a cognitive scientist at the University of California, San Diego, who studied social systems in baboons before turning her attention to social robots.
Designing robots that understand people is only half the equation. Developers also have to consider what features robots will need for people to trust them and want to engage with them.
"We know people will be applying social-psychological constructs—such as beliefs, emotions and motivations—to try to understand what these robots are doing and why," Breazeal says. "If this is the way that people are intuitively trying to make sense of robots, we should be trying to design these systems that meet them halfway."
Someone, not something
In order to adapt and respond to human behavior in all its messy unpredictability, social robots will need to excel at reading our intentions, says Bertram Malle, PhD, a professor of cognitive, linguistic and psychological sciences at Brown University who studies social cognition and moral judgment. "Robots need to figure out what the human partner wants and intends and assumes, and then take into account these inferences in order to adapt its own behavior and respond correctly."
To do that well, robots will need to be able to learn from their human partners, Malle says. Bringing a robot into your home might prove to be less like unboxing a fully functional tech gadget and more like adopting a small child. "You need a robot that adapts, learns, gets better, responds and tailors its actions to a particular interaction partner," he says. "That's impossible without the ability to learn over time through feedback, reward and criticism."
To get that far, however, people have to want to socialize with the robot in the first place. "When a person is interacting with a social robot, it should feel much more like you're interacting with a someone rather than a something," Breazeal says.
In other words, a successful robot is one that we perceive as having some agency over its choices and behaviors. Researchers are still sorting out just what makes something seem more like a self-directed agent and less like a toaster, says Scassellati. "We know from hundreds of years of work in art and animation that there are some very simple visual tricks that we can use to make something appear to be lifelike"—such as eyes, the ability to move in a goal-directed manner and actions that vary in response to our behavior.
In a series of studies, Scassellati has identified another, more surprising trait that makes robots more attractive to humans: an ability to cheat (Proceedings of the 10th Annual ACM/IEEE International Conference on Human-Robot Interaction, 2015). In one experiment, participants played the game rock-paper-scissors with a robot. When the robot behaved in a predictable way, participants treated it like a mere machine. But when the robot cheated to win, everything changed. "As soon as the robot cheats, it immediately transitions from being an object to being an agent. All of a sudden, people start making eye contact with it, they'll start talking to it and showing interpersonal space relations with it, they'll start using personal pronouns when referring to it," Scassellati says.
Research by Adam Waytz, PhD, a psychologist and professor at Northwestern University's Kellogg School of Management, also supports the idea that people are more likely to anthropomorphize a robot when it behaves somewhat unpredictably (Journal of Personality and Social Psychology, Vol. 99, No. 3, 2010). "When technology is unpredictable, it motivates people to make sense of that technology, and that sense-making leads people to humanize the robot more," Waytz says. "You wouldn't want a robot to behave completely randomly, but some degree of unpredictability can be positive for engagement."
And once people start seeing the robot as a "someone," it tends to stick. "We can have them play another 30 minutes where the robot does nothing unpredictable, and those animacy effects will persist. Even though the robot is back to its normal, boring state, they still talk to it and treat it like it's an agent," Scassellati says.
In some ways, it might not be that much of a stretch to design a robot that people treat as an agent, says Malle. Slap on some eyes, give it self-directed motion and a voice and it's almost inevitable that we'll treat robots as social beings. Malle says it's where the interaction goes next that matters. "Very quickly our expectations become very high," he says. But if robots don't live up to those expectations, it won't take long for people to give up on them, he adds. "This is the challenge right now for robotics."
Even as researchers work to figure out what triggers human expectations and to build those capacities, they're amassing evidence that robots can provide value to human lives.
Several studies have explored interactions between robots and young children. In one study, UCSD's Forster, with Mohsen Malmir, PhD, and colleagues studied interactions between toddlers and a socal robot called RUBI, the brainchild of UCSD research scientist Javier Movellan, PhD. For 28 days, toddlers at an early childhood education center interacted with RUBI while the robot collected data about the children's facial expressions and activities. The researchers found that RUBI learned to use information about the children's facial expressions to accurately predict their preferences for different activities. On average, the robot's predictions agreed with those of human judges as much as the human judges agreed with one another (Proceedings of the IEEE International Conference on Computer Vision Workshops, 2013).
Other research suggests that robots can influence children's behaviors and attitudes in positive ways. In one study, Breazeal and her colleagues used robots to model a growth mindset, the belief that success comes through effort and perseverance. Children ages 5 to 9 played a puzzle game either with a robot that exhibited a growth mindset, or one that displayed a neutral mindset. Afterward, kids in the growth mindset group tried harder to complete a challenging task (Proceedings of the ACM/IEEE International Conference on Human-Robot Interaction, 2017).
In another study, Breazeal and her colleagues showed that school-age children exhibited more curiosity after playing games with an autonomous robot that showed curiosity-driven behaviors (Proceedings of the Tenth Annual ACM/IEEE International Conference on Human-Robot Interaction, 2015).
Robots as therapists and companions
Socially assistive robots also show promise in providing therapy to children with autism. Research has found that engaging with therapy robots increases engagement, attention and novel behaviors (such as spontaneously imitating the robot) among children with autism, as Scassellati and colleagues described in a review (Annual Review of Biomedical Engineering, Vol. 14, No. 1, 2012).
"Human behavior can be overwhelming for children with autism, yet a social robot can provide a sort of manageable social experience," Breazeal says. "You can start to see children exhibit these social skills and capabilities that you may not have seen them perform before with another human being, and then the clinician can start to build on that."
Scassellati is heading a large project funded by the National Science Foundation to further explore robotic therapy for children with autism. The robots spend a month in the child's home. They analyze and adapt to each child's behavior, tailoring their interactions to suit the child's abilities, preferences and behavioral goals.
So far, eight families have completed the protocol and half a dozen more have welcomed the social machines into their homes. Though the data are still being collected, Scassellati is hopeful. "Anecdotally, the families love them. These children are doing hours of therapy a day, and the robot makes therapy fun," he says. "One mother told us that she's learned new ways of doing things thanks to the robot, and she plans to keep doing them that way. That was fantastic for us."
At the other end of the lifespan, older adults are using robots for comfort and companionship. The robotic harp seal Paro wiggles his flippers, displays a variety of emotions and responds to a user's touch and voice. A pilot study by Wendy Moyle, PhD, RN, and colleagues found that older adults with dementia who regularly spent time with Paro had higher levels of pleasure and greater quality of life than peers who participated in a reading intervention (Journal of Gerontological Nursing, Vol. 39, No. 5, 2013).
Seals may be just the start. Breazeal predicts that more humanlike robots might one day provide companionship to older adults. "Elder care is a huge area of opportunity, and chronic loneliness is an epidemic in our society. I think technology will be one part of the solution."
Last year, a company Breazeal heads, called Jibo, released a tabletop robot of the same name that can tell jokes and stories, play games and otherwise interact with the people around it. The robot uses facial and vocal recognition to get to know each member of the family, adapting its interactions to each individual. "Jibo is designed to interact and express its own internal states, such as its likes and dislikes," she says. "It's the very beginning of bringing these technologies out into the world, but it's opening the avenue to socially intelligent robots."
APA is hosting Technology, Mind and Society, an interdisciplinary conference exploring interactions between humans and technology, on April 5–7, in Washington, D.C. For more information or to register, visit https://pages.apa.org/tms.
What Kind of Mind Do I Want in My Robot? Developing a Measure of Desired Mental Capacities in Social Robots
Malle, B.F., & Magar, S.T. Proceedings of the Companion of the 2017 ACM/IEEE International Conference on Human-Robot Interaction, 2017
Moral Judgments of Human vs. Robot Agents
Voiklis, J., Boyoung, K., Cusimano, C., & Malle, B. Proceedings of the IEEE International Symposium on Robot and Human Interactive Communication, 2016
Integrating Socially Assistive Robotics Into Mental Healthcare Interventions: Applications and Recommendations for Expanded Use
Rabbitt, S.M., Kazdin, A.E., & Scassellati, B. Clinical Psychology Review, 2015
Letters to the Editor
- Send us a letter