Researchers looked at 1.1 million photos on Instagram and found that pictures with human faces are 38 percent more likely to receive likes than photos with no faces.
They’re also 32 percent more likely to attract comments. The number of faces, gender, or age didn’t make a difference.
The study, conducted by researchers at Georgia Institute of Technology and Yahoo Labs, is one of the first to examine how photos with faces drive engagement on large-scale, image-sharing communities.
On average, pictures of kids or teens aren’t any more popular than those of adults, even though Instagram is most popular among younger people. Men and women have the same chances of getting likes or comments.
A few factors did play a role. As expected, people with more followers attracted more engagement, but only if they didn’t overdo it.
“The more you post, the less feedback you’re going to get,” says Saeideh Bakhshi, a Georgia Tech PhD student, who led the study. “Posting too much decreases likes two times faster than comments.”
Bakhshi also says that the more photos someone uploads, the lower the probability any single one has of getting likes or comments.
Why We Love Faces
She and her team, which included advisor Eric Gilbert and Yahoo Research Scientist David Shamma, used face detection software to scan the photos.
While the study examined how people react to photos with faces, the researchers stopped short of determining why users behave that way. They have some guesses, though.
“Even as babies, people love to look at faces,” says Bakhshi. “Faces are powerful channels of non-verbal communication. We constantly monitor them for a variety of contexts, including attractiveness, emotions, and identity.”
Regardless, knowing that photos with faces drive more engagement could have practical implications. Gilbert notes that social media sites such as Flickr or Pinterest could increase their search ranking and keep consumers onsite and active by featuring human faces in their online content.
“Designers could also use this knowledge to quickly filter, prioritize, and highlight photos shared by followers,” says Gilbert, an assistant professor in the School of Interactive Computing. “Especially pictures that have just been submitted and haven’t had enough time to pick up very many likes or comments yet.”
The team plans to learn more in the future. They’d like to see if pictures of friends are more or less popular than family group photos, or if selfies attract more attention than group shots.
The study will be presented in Toronto at the ACM CHI Conference on Human Factors in Computer Systems on April 26 to May 1. Funding from the Army Research Office helped support the project.