Above is a screen recording of a program written by Greg Borenstein running face detection on a selection of Unicode emoji faces. The program, Eigen Faces, is running an algorithm that loads in test images, analyzes them using vector math and Principle Component Analysis, and is able to understand what a face looks like, remember it, and recognize it.
In that process, the algorithm creates an interesting artifact, the eigen face. An eigen face is the result of obtaining a mean image from your training set, finding the difference between that and the test image, and then performing some covariance matrix math. It ends up creating a ghostly specter that wants to stand on its own as a cultural artifact. Our faces are our identities, and this object is so much like trying to pin point who you are. Like when you sit and think about something, take a second and then ask yourself "who said that?', it creates a dissonance that feels uncanny and a bit liberating.
This experiment is a sketch for the DSM-E, a project I'm co creating with Kate Tibbetts. I wanted to do this because I wanted to see if this process, creating eigen emojis, could decouple emojis from electronic communication, placing them in a space of their own, giving them identity and individuality. Emojis are meant to be vessels and objects of transference. What happens when they start to operate on their own? Can they become unique? Individualistic? Are there affinities across emojis?