Computer program can tell if your smile is fake

You're engrossed in an argument with your mother-in-law, but for the sake of keeping the peace, you grin and bear it. Yes, you're frustrated and of course she's wrong, but you flash a smile like someone just dropped a bouquet of posies into your lap and agree to disagree.

Do you think it fooled her? It might have. But researchers at MIT have now developed a computer program they say can detect the true emotion behind that frustrated smile, reports TIME. And they claim it has potential for a plethora of uses.

For the study, participants sat at a computer with a webcam that captured their images throughout. All participants smiled when shown images of babies and 90 per cent didn't smile when asked to look frustrated, says a story in Mashable. They were then asked to complete an online survey that intentionally wouldn't save, which brought out genuinely frustrated smiles from the participants.

Also see: Emotionally intelligent people more likely to be deceived

Afterward, 50 per cent of humans looking at images of the frustrated and genuine smiles side-by-side were able to accurately detect which image corresponded with which emotion, compared with the computer algorithm's 92 per cent.

While this is all well for looking at still images, Dr. Barry Stein, a registered psychologist based in Victoria, says that to accurately detect the emotion behind a person's smile in real life, other factors have to be considered. He cites context as a key determinant.

"If the context is of a person smiling around a cruise ship and they've got a drink in their hand and they're on their way, going to Alaska, I think it's a pretty good guess that they're probably happy," he says. "Whereas you put a person on a basketball court and they've been fouled … you often see the NBA players smiling. And they're smiling out of frustration."

Also see: Having a plan can help you finish one task, but not multiple

He compares the computer program's fallibility to that of a lie detector or polygraph, which, rather than measure whether or not a person is lying as the name suggests, actually measures an individual's stress level, blood pressure, sweat gland activation, reports The Guardian.

"There are some people that can beat lie detectors," says Stein. "Because they don't get nervous while they're lying. So an algorithm will probably be more accurate than most humans in detecting lying. Most of us aren't using a large part of our brain."