Researchers at the University of Birmingham, Aarhus University and Linnaeus University have put a name to something nobody has measured yet. In a paper in AI & Society they call it robotoid humanness: the drift by which a customer, after enough exchanges with a service robot, comes to feel most fluent and most correct when speaking in the way the machine reads best. Their framework covers the setting, what the customer expects, how the robot looks and sounds, and what each side then does.
The mechanism they propose is mirroring. Service robots are built to imitate human gesture, speech and emotional cue, and people imitate back, mostly without noticing. Feedback from other people is unpredictable; feedback from the machine is affirming and statistical, so the edges get filed off to keep the exchange moving. The authors are plain that this is theory. Nobody has yet been followed over time to see whether it happens.
So…:
Where have you started phrasing things the way the system prefers, not the way you mean?
Source:
Picture: