DU Want to Build-A-Bot?

Robots are being employed in children’s hospitals to help ease kids’ anxiety. Students with special needs are reaching new levels of learning with robots in the classroom. Robots increasingly are being built to engage with people, but will people want to engage with them?

That depends on the design, according to Kerstin Haring, assistant professor in the Department of Computer Science. For social robots, like those that interact with humans, design is crucial.

“One of the big problems that we have in robotics is that we're not quite sure what a certain robot design elicits in a person,” says Haring. So, she and her colleague, Dan Pittman, formed an interdisciplinary research team to learn more about social robot design. They started working with Pilyoung Kim in the Department of Psychology because human reaction is anchored in psychology.

Haring explains it this way: Upon meeting a person, humans look at surface-level queues and form an image of how an interaction is going. It’s similar when we look at objects. We have expectations based on our experience of how things work in the real world. This is referred to as mental model.

“Mental model is what we form of others. We do it all the time, and we do it very rapidly,” says Haring. “The big question…we don't know what happens when we look at a robot.”

Research tells us that if a robot has eyes, we expect it to see. If it has ears, we expect it to understand acoustics or language queues, Haring adds.

“There is something about the appearance of robots that gives us a very quick idea of what the robot can and cannot do. The big problem is that currently the people who design robots … have this very limited robot design team.”

Thanks to a University of Denver Professional Research Opportunity for Faculty (PROF) grant, the team was able to set the wheels in motion. They built a research platform created as a web-accessible 3D game to collect large amounts of data of user-provided robot designs. The “Build-A-Bot” webpage enables designers to create robot characters the same way they would create a character in a computer game using a game engine called Unity.

“It’s like game design that we're employing because we want it to be fun, we want it to be relatively engaging and easy to build these robots,” Haring says.

Anyone interested in robot design—from kids to grandparents—can participate by going to the website. Instead of using existing robot designs, the researchers are asking people to design robots that can experience hunger, pain or love, for example. They think there will be a pattern that will repeat in those designs.

Designs then will be used to train machine learning models to better evaluate robot designs, predict how a given design will be perceived and create new robot designs.

Up to 25 students, mostly undergrads, participate in the interdisciplinary research project and contribute in myriad ways. Computer science students work as designers and project developers, a music student created background music for the game and a business student is working on a social media presence.

A grad student in psychology is working on neuroscience evaluation. The researchers are trying to show that robots can activate certain areas of the brain.

“The idea of hunger, pain, fear, those are relatively well researched in that area,” says Haring. “If you see someone else in pain, there would be a very distinct pattern that we can see with this kind of neuroscientific evaluation.

“And then, if we have enough examples of a robot that can experience pain, we can somewhat manipulate an image or a video that would suggest that our robot is in pain. If it would trigger the same reaction in our brains, we basically would have evidence that we treat robots social in that way. And it's not because we want to, but because our brain tells us to,” she adds.

This fall, the team will test the Build-A-Bot homepage to fix any bugs before the official launch. It’s important that they ask the right questions and use the right terminology so that the game is easy to navigate and understand.

“We want to make it right, as good as possible, so that it's as [accessible] as possible for that kind of broad audience that we anticipate,” Haring says.

The team hopes to have the first systematic evaluation of the page and a small data set of robot designs by the end of the year. Ultimately, they want to produce a robot design guideline.

“Whatever you want to design a robot for, we will have a very good idea of how this needs to be designed to evoke a certain expectation or reaction in humans,” Haring says.