Technology trained to judge faces like humans do

Technology trained to judge faces like humans do


We humans have a terrible habit of using our eyes to instantly size up those we meet, quickly attributing character traits to them that may be wildly inaccurate.

That doesn’t stop us from doing it, of course.

Researchers at Stevens Institute of Technology have taught an artificial intelligence algorithm to accurately predict how someone will be perceived based on a photograph of their face.

The researchers’ work is outlined in the Proceedings of the National Academy of Sciences.

The scientists asked thousands of people to offer first impressions of more than 1,000 computer-generated photos of faces, ranking them on a host of criteria, such as how electable they appeared, or how religious, trustworthy or smart they were.

Those responses were then used to guide the algorithm to think like a human and make similar judgments.

Some of the algorithm’s uses are obvious: Say you want to curate a set of photos likely to make others view you most favorably. For instance, you might look super cute in the photo where you’re wearing sunglasses, but because no one can see your eyes, people may not trust you on first glance.

Or you might want to know which photo you definitely shouldn’t add to your resume, because it makes you look shady as heck.

Down the road, the researchers believe they can create an algorithm that will accurately predict how one individual will respond to another person’s face.

That technology could go a long way in helping us check our own biases and level the playing field in job searches or even criminal cases.

Until then, just know that while a picture might be worth 1,000 words, they may not be the most forgiving.

Related Episodes