MIT Algorithm Sorts Real Smiles from Fakes

MIT researchers have developed an algorithm that determines, with 90 percent accuracy, whether a person’s smile is genuine or not.

The algorithm uses visual cues to calculate the likelihood that a smile is real, and it does it well enough to beat regular humans, who tend to score worse than if they simply left their assessment to random chance.

Researchers say people with autism could potentially benefit from this technology, because they have particular difficulty picking up contextual clues that suggest what another person might be feeling. For example, autistic people are often taught that upturned corners of the lips tend to indicate happiness. That’s true unless the person in question is only smiling with their lips – in other words, if they’re faking it.

The algorithm could simply be used to train people to better recognize genuine smiles, but in an age of tiny computers and Google Glasses, it’s not hard to imagine using it to provide real-time feedback during person-to-person interactions. Then again, it might tell people more than they really want to know.

Related Links