We're social creatures, us humans. And often, social connections are a driving factor behind our healthy or not-so-healthy behaviors, as you realize when a friend invites you along for a jog or, less helpfully, urges you to finish that last cookie.
Tapping the power of these connections is the basic premise behind socially assistive robotics, a field that develops bots to aid human health, explained Maja Matarić, PhD, director of the University of Southern California's Robotics and Autonomous Systems Center. Matarić plans to speak at this year's Big Data in Precision Health conference.
I spoke with her recently about how these robots can elevate human health.
When we think about robots, we tend to think about them doing physical work. But that's not the only way for robots to help people. That's where socially assistive robotics comes in -- by helping people in social, not physical ways.
My work was inspired by the need for accessible and effective care for large populations. When people are engaging in difficult health-changing behaviors they need companionship and encouragement, and that support is most effective when it comes from a physical agent, such as another person, pet or a robot.
In general, the robot needs to have the right expressive abilities to be engaging and helpful, but what those are depends on the robot's task and context.
We've used the humanoid robot Bandit, which has arms, to show stroke patients how to do rehabilitation exercises, elderly users how to do chair aerobics and children with autism how to imitate movements. In other cases, we needed the robot to be encouraging and fun, but smaller and inherently safer, so we used the owl-like robot, Kiwi, in the homes of children with autism.
We're in the booming era of big data, but the data available are actually quite specific, so machines can only learn a narrow range of things, such as how to look for anomalies in medical images or recognize a specific person in photos.
Human behavior is much richer and more complicated -- it's not just what you say, it's your body language, the volume of your speech, inflections, how close you stand to the other person, and much more -- plus every action you make influences the response from the person you're interacting with. We as humans pick up on and employ all of these subtle ways of interacting and communicating. But we have almost no data sets of such rich interactions, especially in really important, sensitive health-related settings.
It's even more difficult to collect or analyze the data of people with depression, anxiety or autism, because it's sensitive and protected. So most researchers just use convenient data sets available online, which are not sufficient.
I challenge my lab to go out and collect real data, in real settings, with real people interacting in natural ways. For example, we just finished a study where we left robots in the homes of families with kids with autism for a month at a time. It's an unprecedented dataset, yet it's barely enough. We could only afford to collect data from 17 children and it took a long time. There needs to be more support for and engagement in that type of research.
In one case, we left socially assistive robots in homes of elderly people. The robots were programmed to monitor how much the participants sat in their favorite chair. If they sat for too long, the robot encouraged them to get up and walk around. If they got up, the robot rewarded them with a little dance.
We wanted to see if users would listen to the robot or get tired of it, and whether they would continue to minimize sitting after the robot was removed from the home.
People adhered to the robot's instructions very well, even over multiple weeks. But once we took the robot away, the behavioral change ended. So the question is: how long would a robot have to be in the home to change the user's behavior permanently? It will take more studies, with many more participants, to better understand human behavior through human-machine interaction.
Matarić is slated to speak on May 23, day two of the two-day event. Registration for the conference, which will be held on campus, is currently open.
Photo by Elizabeth Bayne