Socially Sensitive Help Giving from Robots
Method — Laboratory Experiment
The robotic behavior tested in this controlled experiment was created after careful qualitative observation. The combination of methods offers a more complete picture than either method on its own.
Brought baking supplies to a university dorm and recorded people’s conversations (with permission) as they tried to make cupcakes. Coded these “helpful” tips by their linguistic features and programmed a robot with different helping styles based on the observed conversations. Ran a controlled laboratory experiment to measure perceptions of these different helping styles.
Robots are often conceptualized as ideal assistants, smart and steadfast. But can help from robots (like help from some people) become overbearing and disruptive? What makes a robot a humane helper?
Robots using an indirect style (“so actually you might just want to…“) were perceived to be less controlling, but there was a drawback. Indirect styles were also perceived to be less intelligent. An intriguing design direction: A not-quite-sure robot whose help will never be confused with a command.