Taking a page from how people learn, researchers from the University of Tsukuba, Japan have created a method of teaching robots via facial expression. Most humans and some animals learn a huge amount of their behavior by the body language of those teaching them — the difference between a smile and a frown tells them if they're doing something right or wrong. By applying these same methods to training robots, you're able to influence their actions in a more intuitive and immediate way than by hitting a button or using a dial.
The technology uses a small wireless electromyography (EMG) head band, which can accurately read smiles and frowns 97 percent of the time — and unlike facial recognition, can be used facing any direction, and under any light. The video below shows how it's used, training a Nao robot in real time whether it should give a ball or throw it. The robot's learning is just about palpable as it figures out what the scientist wants based on if she's smiling or frowning. It's a step towards a more naturalistic way of interacting with robots, one which taps into taps into universal human expressions rather than requiring a specific vocabulary.