News

What Makes a Face Memorable?

by Andrea Garcia-Vargas

Selfie fans of America, you're going to love this: researchers at MIT are figuring out out how to digitally adjust an image of a person's face to make it more memorable. How? By making small, specific changes to images of yourself using photo-editing technology — and we're not talking about Photoshopping your blemishes, enhancing your eyes, or making your hair look smoother. It's a whole lot more complicated than that.

Yes, some of your face's "memorability factor" has to do with whether it feels familiar to you. But sometimes certain faces just look memorable, as NPR explains.

There are subjective factors affecting how a face sticks in your memory — for example, if you know someone else who looks similar, you might find a new face more familiar. But researchers found that there is also a strong universal component to memorability. Some faces are just consistently more easily remembered.

Researchers found that certain associations help make a face memorable: familiarity, kindness, trustworthiness, uniqueness.

But what does a familiar, kind, trustworthy, and/or unique face even look like? Well, MIT put a computer algorithm on it to find out. From one of MIT's research articles on the subject:

To overcome the complex combination of factors that determine the memorability of a face, we propose a data-driven approach to modify face memorability. In our method, we combine the representational power of features based on Active Appearance Models (AAMs) with the predictive power of global features such as Histograms of Oriented Gradients (HOG), to achieve desired effects on face memorability. Our experiment show that our method can accurately modify the memorability of faces with an accuracy of 74%.

OK, time for plain English. Researches used a computer algorithm (the Active Appearance Model) in this experiment, a method often used to match and track types of images in medicine and forensics.

Then, they combined it with another algorithm called Histograms of Oriented Gradients, which tracks images in a slightly different way to AAMS. The point: This combination of methods were pretty darn good at knowing which faces will be memorable, and which weren't.

It's interesting that trustworthy and friendly faces are inherently memorable — but in the words of one NPR commenter and a couple of Twitter users: "Hmmm... based on those sample photos, what makes a face more memorable is being thinner."

Plus, it's hard to know how reliable those supposedly "attractive" and "friendly" faces really were, since those are such subjective measurements.

Tweaking your Facebook photos so people remember you more: Coming to an app near you. Probably.

Images: NPR, MIT