Thursday, March 20, 2014

A debate on what faces can tell us.

Security agencies are developing facial emotion profiling software for use at checkpoints, while Apple and Google are working on using your laptop camera to tell them what kind of mood you are in while shopping online. Such approaches are based on the assumption that a basic set of facial emotions are invariant across cultures and universally understood. A large body of work, starting with Charles Darwin and especially since the 1960's done by Paul Ekman and others has substantiated this idea.

In yet another New York Times Op-Ed advertisement wanting to raise the visibility of some basic research, Barrett and collaborators make the heretical claim that this assumption is wrong and point to their articles questioning Ekman's original research protocol of asking individuals in cultures isolated from outside contact for many centuries to match photographs of faces with a preselected set of emotion words. They suspected that providing subjects with a preselected set of emotion words might inadvertently prime the subjects, in effect hinting at the answer, and thus skew the results. In one set of experiments subjects not given any clues and asked to freely describe the emotion on a face or state whether emotions of two faces were the same or different performed less well. When further steps were taken to prevent priming, performance fell further.

A rejoinder from Paul Ekman and Dacher Keltner points out that a number of studies supporting Charles Darwin's original observations suggesting that facial movements are evolved behaviors have avoided the issues raised by Barrett et al. by simply measuring spontaneous facial expressions in different cultures, along with the physiological activity that differed when various universal facial expressions occurred.  It seems reasonable that a universal facial emotional repertoire might in practice be skewed by culturally relative linguistic conventions,  thus helping to explain Barrett et al's observations.

1 comment:

  1. Anonymous12:06 PM

    Derik this petty bickering plays into the hands of a revolutionary like myself — it's as if modern researchers are still questioning the theory of evolution! Fortunately science does move on regardless of such disputes. I think your readers would be interested to know that researchers at UC San Diego have developed software that can accurately identify pain in a subject with 85% accuracy, while trained humans are found to be successful only 55% of the time. And since it's still very new, who knows how much better this software might soon become.

    Your readers might also wonder why a philosopher like myself would take such an interest, though this subject is quite related to my own work. Without accepted understandings to philosophical elements of reality, I believe that "mental/behavioral" sciences will remain "pre-Newton."