This new computer program can see right through your poker face.
Illustration by Tyler Hoehne
In today’s news of machines taking on eerily human qualities, a study in Behaviour & Information Technology journal reports on a new computer program that can correctly identify an individual’s emotional state around 80 percent of the time. The program analyzed two sets of criteria from each study participant: the person’s keystroke habits and the exact words typed. Each user was then categorized into one of seven emotional classes.
“If we could build any system that is intelligent enough to interact with humans that involves emotions, that is, it can detect user emotions and change its behavior accordingly,” wrote the team of Bangladeshi researchers in the study’s abstract, “then using machines could be more effective and friendly.”
According to the study, it seems that the easiest emotional states to read are happiness and being totally pissed—the program’s algorithms were able to accurately sense joy 87 percent of the time, and had an 81 percent success rate with anger. Guilt came in at 77 percent, disgust 75 percent, sadness at 71 percent, shame 69 percent, and fear was properly identified 67 percent of the time.
While the minds behind the study hope that the technology will help foster better interactions between man and computer, we’re hoping the program leads to an app for deciphering what those cryptic, passive-aggressive text messages from your significant other actually mean.