MIT Develops a computer program to spot your Frustrated OR Fake Smile

In our every day lives, we sport both genuine and Fake smiles on our faces. Sometimes the world successfully spots a fake smile, sometimes it doesn't. As a result , we are much at ease. But imagine a future, where even your dumb computer snubs you for your fake smile.

That day, you'll definitely scratch your head.

No you will not.

As MIT researchers feel that such a responsive computer will be able to serve you much better.

Nailing down your happy and frustrated Smiles:

A new computer program developed at Massachusetts Institute of Technology (MIT) can guess right nearly twice as often as humans.

The computer program can analyze human smiles. Right now it can tell a genuine smile from a frustrated one--and do so with much better accuracy than we humans can. The researchers are optimistic that some day it may be able to to much more. For them, it’s the first step, they say, towards developing machines that can recognize emotions.

How the Software is created:

Researchers at the MIT’s Media Lab created the software as part of their mission to make “more intelligent and respectful” computers.

The team began by filming a sampling of spontaneous smiles, some delighted and some frustrated, from 27 people who watched the video of an adorable baby. the participants were also made to fill a lengthy, malfunctioning online form. Interestingly, 90 percent of people smiled when they got frustrated.

To develop the algorithm (flow chart), the researchers broke down the smiles into two different types: Frustrated fake smiles came on fast and disappeared quickly; happy smiles rose more gradually and faded over time. The fake smiles were also thinner and stiffer.

Next, 12 people were shown still-frame shots of the two types of smiles and asked which were frustrated and which were happy. These subjects were only able to chose correctly 50 percent of the time--no better than a random guess. The algorithm, on the other hand, was able to gauge the difference accurately 92 percent of the time.

Why the software has a great success rate:

According to Ehsan Hoque, assistant researcher of MIT’s Affective Computer Group and the study's lead author, the reason for computer software faring better than humans is that while the humans usually kind of zoom out and try to interpret an expression, a computer algorithm can utilize the nitty-gritty details of a signal, which is much more enriching.

Possible Applications:

According to the researchers, a responsive and reactive computer is not a bad scenario, as such a computer will be able to serve its human master better. The impact of emotion-sensing computers could be profound: they could sense our feelings and respond more appropriately. Ona more serious note, the software program developed could also prove a boon to those with interpersonal psychological issues, such as autism, in which an individual has difficulty picking up on emotion cues in face-to-face communication.


No comments