The first generation of social home robots (like the first generation of many other new applications of technology) was not particularly successful. A series of high valuations followed by mediocre sales and reviews leading to several company shutdowns has made it much more challenging to develop in this space. And that’s not necessarily a bad thing, if we’re honest—it’s become clear that social home robots are difficult to get right, and any company that wants to make one at this point needs to have both the technical capability and (more importantly) a long-term business case that’s more comprehensive than just making a robot into a “part of the family” or a “best friend.”
Today, a social robotics startup called Embodied is launching a new robot called Moxie (no relation to this or this), a social companion for children aged 6ish to 9ish. According to Embodied, Moxie is “designed to help promote social, emotional, and cognitive development through everyday play-based learning and captivating content.” In some ways, it’s like all the other social robots we’ve seen in the past, but in others, it’s different enough that it could find success—especially right now, when both kids and parents are in need of some extra help.
Embodied was founded in 2016 by USC robotics professor Maja Matarić along with Paolo Pirjanian. We first met Paolo back in 2010 when we checked out the Mint floor cleaning robot that he developed as the CEO of Evolution Robotics. Evolution was acquired by iRobot in 2012, turning Mint into the iRobot Braava and Pirjanian into iRobot’s CTO. He left iRobot in 2015 and founded Embodied the following year. Pirjanian tells us that Embodied is about combining his desire to use robots to help people more directly, and his passion for lifelike animated characters. While the obvious application for a robot in that space might be elder care, Pirjanian felt that helping kids with social-emotional development offered more immediate impact.
“Kids are quick adopters of technology,” says Pirjanian, “and the pain points for the parents are super high. Especially if you have a child that’s on the spectrum—we started talking to some families and it was very clear that they needed a solution to help them. But at the same time, they told us that they would want to use the same product for their other child who is neurotypical.”
Moxie is designed to help all kids improve their social-emotional and communication skills through relatively short daily interactions, and while this video is obviously not a live demo and depicts absolute best case scenarios for Moxie, we did get a (remote) demo of an alpha unit running early software, and it performed reasonably well and with more nuance than you’ll see in the video. As with other social robots, it’s important to keep in mind that performance varies in different Wi-Fi environments, different noise environments, different lighting, and so on.
With Moxie, children can engage in meaningful play, every day, with content informed by the best practices in child development and early childhood education. Every week is a different theme such as kindness, friendship, empathy or respect, and children are tasked to help Moxie with missions that explore human experiences, ideas, and life skills. These missions are activities that include creative unstructured play like drawing, mindfulness practice through breathing exercises and meditation, reading with Moxie, and exploring ways to be kind to others. Moxie encourages curiosity so children discover the world and people around them. All these activities help children learn and safely practice essential life skills such as turn taking, eye contact, active listening, emotion regulation, empathy, relationship management, and problem solving.
Unlike other social robots (or robots for children) that we’ve seen before, Moxie specializes in one specific thing and more or less keeps its interactions focused there, rather than trying to be some kind of generalized (and inevitably disappointing) robot friend. The general hope is that kids will interact with Moxie about once a day for 15 to 20 minutes initially, and then over six months or so, Embodied expects interactions to happen three to five times a week, which the company says is what you need in order to have a measurable impact on the child. It’s worth noting that the robot does enforce limits by getting tired and going to sleep if the child tries to do too much with it at once.
The key to sustained interest is content—making sure that there’s always new stuff to keep kids engaged. “Of course kids are going to love the robot, but then they can also quickly lose interest, and we’ve seen this happen over and over again,” says Pirjanian. “So it was something that was top of mind for me from the very early days, even before the inception of the company.” Embodied has 20 people working on developing content right now, including its own experts in neuroscience, psychology, and child development, and they work with an external group of advisers representing organizations like Autism Speaks and the Rogers Foundation. Twelve weeks’ worth of content should be available at launch, with a full 52 weeks of content eventually available. Pirjanian notes that this represents unique content, and doesn’t factor in that (in their experience) kids often want to redo the same activities over and over.
“We’re focusing on some particular techniques from childhood development,” he explains, “like structuring conversations as social stories that allow the child to start reflecting on what does kindness mean, what does family mean, and those kinds of things.” Embodied says it has trained the robot on conversations that it has gotten from kids by working with about a hundred families for more than a year. The testing allowed the company to identify certain common themes and topics that the robot can talk about like school, friends, bullying, doctors, and so on.
It’s unlikely that there will be any kind of user-generated content, since Pirjanian knows how risky that could be. “There is a big responsibility we are taking on in terms of helping children with social-emotional development, so we don’t want people just haphazardly putting content on the robot.” Embodied also takes data protection very seriously, with 90 percent of the processing taking place on the robot itself, and camera images in particular staying local. Anything that does go into the cloud is anonymized using a key that a parent chooses, so even Embodied itself can’t tell where the data comes from.
Unlike other social robots, Moxie specializes in helping kids with social-emotional learning, and keeps its interactions focused there rather than trying to be some kind of generalized, and inevitably disappointing, robot friend. Photo: Embodied
So what does Moxie do with the data that it collects? The idea is that it’ll be continually measuring how the child it’s working with is progressing and communicating results with a parent through the app. Embodied is using an established framework known as the Vineland Adaptive Behavior Scale for this, and in addition to keeping a parent up to date on how the kid is doing, Moxie also uses these metrics internally to help it make decisions about which activities to encourage the child to do and when, adjusting its pace as necessary.
In general, Moxie is fairly resilient to kids’ whims, and can gracefully handle kids saying things like “I don’t want to help you” or “I don’t want to talk about it” or “I don’t want to share anymore” with a combination of gentle insistence followed by moving on and coming back to important topics later on. Pirjanian says that there are certain milestones that the robot tries to hit over defined periods of time, and to help encourage kids to participate, there’s a reward loop that combines best practices from both child development therapy and the video gaming industry. “We do not want to make addictive games,” says Pirjanian, “but we’re using similar psychology to get the child to work through these milestones that are going to be beneficial to them.”
We asked Pirjanian what happens if, say, the child tells the robot that they’re being bullied at school. “We’re looking for both specific trigger words and also intent, and if there are risky things that the child is having a conversation with the robot about, we’ll flag those and notify the parent through the parent’s app.” Moxie can provide general advice to parents through its app, and connect them to experts to address specific questions or concerns.
It’s worth noting that the robot does enforce limits by getting tired and going to sleep if the child tries to do too much with it at once. Photo: Embodied
The real challenge, I think, is how successful Moxie will ultimately be with children who may not have the interest or patience to stay within those very specific activity-driven guardrails that Moxie is designed around. It’s certainly possible that Moxie’s deliberate focus will make it more effective, although a few videos that we saw of nonscripted kids interacting with the prototype occasionally made it seem like Moxie’s (alpha version) AI was having a hard time hanging on, especially as kids have gotten used to speaking to Alexa and Siri. What makes me more optimistic is how aware Pirjanian is of these challenges, and the fact that Embodied has spent over three years building Moxie to fit effectively into this specific niche, which has tangible value attached to it for families.
Moxie will first be available (for free) through a beta program so that Embodied can get feedback from a wider group of families, and the goal is to be shipping robots to everyone by October. You can reserve a Moxie for a (fully refundable) deposit of US $50, which goes towards the $1,500 total price of the robot. That does include a year’s worth of content, after which you’ll need to pay an additional $60 per month. But remember, this isn’t a toy, it’s a tool, and if it’s something your kid needs, or could benefit from, it may very well be more than worth the cost.
[ Embodied ]