MIT researchers have managed to create what could end up being the first social interaction framework for robots. The framework enables machines to consider not only their task at hand but also how their actions would affect others. In a simulated environment, a robot watches its companion, guesses what task it wants to accomplish and then helps or hinders this other robot based on its own goals.
The researchers also showed that their model creates realistic and predictable social interactions. When they showed videos of these simulated robots interacting with one another to humans, the human viewers mostly agreed with the model about what type of social behaviour was occurring.
Allowing robots to demonstrate their social skills can lead to smoother, more positive human-robot interactions. For example, robots in Assisted Living facilities can use these features to create a more compassionate environment for the elderly. The new model will also allow scientists to quantitatively measure social interactions, which may help psychologists study autism and analyse the effects of antidepressants.
Robots will live in our world soon enough, and they really need to learn how to communicate with us on human terms. They need to understand when it is time for them to help and when it is time for them to see what they can do to prevent something from happening. This is very early work and we are barely scratching the surface, but I feel like this is the first very serious attempt for understanding what it means for humans and machines to interact socially.
– Boris Katz, Principal Research Scientist & Head of InfoLab Group, MIT’s Computer Science and Artificial Intelligence Laboratory
To study social interactions, researchers have created a simulated environment that pursues physical and social goals as a robot moves through a two-dimensional grid. Physical goals are related to the environment. For example, the physical goal of a robot is to move to a tree at a specific point on the grid. Social goals include guessing what another robot is trying to do and acting on that guess. For example, it’s like helping another robot water a tree.
Researchers use models to specify what a robot’s physical goals are, what their social goals are, and how much the robots should focus on each other. The robot is rewarded for the actions it takes to reach its goal. If the robot is trying to help a companion, adjust the reward to match the rewards of other robots. If it is trying to interfere, it adjusts its reward to the opposite. A planner is an algorithm that determines the action a robot takes and uses this continuously updated reward to guide the robot to perform a combination of physical and social goals.
Researchers have used a new mathematical framework to define three types of robots. Level 0 robots have only physical goals and cannot be socially inferred. Level 1 robots have physical and social goals, but all other robots assume that they have only physical goals.
Level 1 robots can perform actions such as assisting and sabotaging based on the physical goals of other robots. Level 2 robots assume that other robots have social and physical goals. These robots can perform more sophisticated actions, such as participating in help together.
The team is now working on developing a system with 3D agents in an environment that allows more types of interactions. They also want to modify the model to include environments where actions can fail, and they plan on incorporating a neural network-based robot planner into the model. Lastly, they’ll look to run an experiment to collect data about the features humans use to determine if two robots are engaging in social interaction.
