If you want more “likes” on your Instagram pics, you should probably pose with a panda in a red bikini — at least, according to MIT PhD candidate Aditya Khosla. Also, according to me: I would probably also like that picture (depending on the panda).
But Khosla has data to back him up. By crunching the stats on 2.3 million Flickr photos, he’s isolated a set of variables he says explain why some images go viral and others get totally ignored. And more importantly (at least, for my personal interests), he’s created an algorithm that can predict how popular your picture is going to be before you even post it. The formula considers social factors, obviously — even the best cat picture on Instagram won’t get any traction if there’s no one to see it — but it also measures content-based factors: things like texture, color, gradient, and objects present in the photo. Spatulas, for instance, are terrible. No one likes your pictures of space heaters, either. Cheetahs and llamas are pretty good, but people in miniskirts are better. People in bright colored miniskirts are the best.
With Khosla’s algorithm, you can avoid the humiliation of posting stuff no one cares about and maximize your chances of Internet stardom. To test it out, I used a tool on Khosla’s site to upload an old picture of my dog wearing rain boots. The algorithm said that no one was going to look at it. Empirically, this has proved to be true: I have tried to show people this photo in person, while literally standing next to them, and I still cannot get them look at it. The algorithm works!
I’m still posting the picture, though, because — c’mon you guys — she’s wearing rain boots.