Imagine a robot that can decipher your emotions by listening to your voice, reading the expressions on your face, interpreting your body language, and then respond accordingly. Machines that can understand human emotion have been imagined and described since the dawn of science fiction. Sometimes they are portrayed as helpful and good (Robby the Robot in the movie Forbidden Planet); sometimes as flawed or sinister (Hal in 2001: A Space Odyssey); and sometimes as both (Isaac Asimov’s I, Robot, and Commander Data in Star Trek: The Next Generation).
Scientists are working to make fiction a reality, and this year’s Siemens Competition in Math, Science and Technology was awarded to two high school students who developed an algorithm that analyzes speech and figures out the emotional state of the speaker. These budding young scientists from Oregon – Matthew Fernandez and Akash Krishnan – developed a program that can analyze audio from a human voice and figure out the emotional state of the speaker. They point out that their program has good ability to recognize basic emotions like happiness and sadness, but is less accurate for complex emotions like anger and fearfulness. They hope to use this technology to help autistic children recognize the emotions of others, and even to assist workers in call centers to respond to the emotions of their customers.
Advances like this are exciting for many reasons. What could be better than a couple of teenagers who develop a revolutionary therapeutic tool that helps autistic kids? And if it helps me get through to a human being at my least favorite airline the next time I can’t redeem miles using the website, all the better. But it also makes me think about how we assess behavioral style, and the next wave of psychological measurement. Scientists are trying to determine personality traits through brain imaging, but it’s a little hard to put a room full of managers into an fMRI machine. Voice recognition, on the other hand, is something that can be realistically and meaningfully used.
SOCIAL STYLE is based on a range of verbal and non-verbal behaviors. While not perfect, imagine a program that could listen to phone calls, decipher the non-verbal aspects of Assertiveness and Responsiveness, and generate an estimated Style of the caller. Better yet, imagine a tool that builds on information it’s already learned about a caller, and each time the person calls the tool refines its analysis to more accurately pinpoint the person’s Style. This would be powerful indeed. For instance, it could help people interact more effectively when working with one another in geographically dispersed teams. It could be especially effective for cross-cultural teams where nuances of emotion may bypass humans but could be accounted for by the algorithm.