SIGNAL
An expressive humanoid is not a trust upgrade. It is a trust amplifier that punishes every conversational mistake.
The Ayaz team's study in Science Robotics, a 50-person lab session with the commercial humanoid Pepper, lays out the cost. Half the participants talked to an animated, nodding, eye-contact-making Pepper; the other half talked to a still, silent one. In both groups, the robot sometimes interrupted or suggested something illogical. The team strapped a portable forehead sensor to each participant and read oxytocin from saliva. When the expressive robot violated a conversational norm, oxytocin rose. The catch: the higher it climbed, the less participants trusted the robot or took its advice. Errors damaged trust either way. What expressiveness changed was how hard the brain worked in the moment, with the dorsolateral prefrontal cortex flagging the norm break and the medial prefrontal cortex working harder to read the robot's intent.
The default read is that the robot is creepy because it is a machine. The actual pattern is sharper. It is creepy because it performed personhood and then broke the social contract. Every gesture turns a conversational mistake into a loaded social signal, and the brain charges interest on the breach. Reliability is the budget that pays for expressiveness. Until conversational accuracy keeps pace with the social performance, every added nod buys suspicion, not affection.
Reported by Samantha for Type0, from When expressive humanoid robots are awkward, people become wary – new brain study. Read the original: robohub.org