We've seen robots that mimic human expressions, and not all of them actually had something that could truly be called a face.
We're not entirely sure MIT's Nexi can be said to have an actual face either, but it comes close enough to actually exhibiting certain human cues.
During an experiment, researchers from Northeastern University asked people to talk to Nexi.
Whenever Nexi detected so-called “signs” that people were lying, it imitated them: touched its face with its robotic arms, leaned back etc.
Thus, the stiller the robot stood, the more trustworthy the speaker.
Currently, there is no perfect, or even good, way of identifying the cues that truly matter, but that's the point of research: to find out answers to dilemmas like this.
to read more or just watch the video embedded above. We have to say, we like her voice.