In order to “teaching” robots social skill, we have to put ourselves into robots’ shoes.
The fact is robots are not really good at understanding human communication, especially in terms of speeches. While humans have little trouble to understand each other through verbal communication, a robot would have a lot of more trouble to understand ambiguous terms or accents and require clarification for further action. That’s why initially when I was trying to modeling robot feedback behavior after the raw human-to-human communications, a large part interactions which would occured in human-robot context was left out. To address the this question, I have to put human participants in the shoes of robots.
Robo Play is research game simulates robot-human interactions with just human participants. During the Robo Play experiment, a human participant will play the role of the robot. This participant will wear a sensory filter, where her/his natural sensory will be filter through a computer. The intention of this experiments is to observe how human would communicate and solve problems with limited natural sensory similar to a robot. By looking at human interactions, we can identify what type communication behavior and cue could adapted robot design to make its actions more readable.
Robo Play I
Robo Play I simulates computer speech recognition over natural language. The participant will wear a headset, preventing s/he to receive any verbal communication from the rest of the “non-robot” participants. Instead, the “non-robot” participants will talk through a speech recognition program where their words will be interpreted by the computer and then displayed to the “robot” participant through texts. This requires the “robot” to participant to constantly signal the other human participants for clarifications and confirmations with very intuitive communication cues such as gestures, and expressions. These interactions provide rich materials and insights for further developing readable robotic behavior.