Mistaking the Highlight Reel for the Game
Instagram has a problem with statistical sampling.
Posted Nov 26, 2018
Bradley Voytek has a section in his CV called “rejections and failures.” Voytek is a professor at the University of California, San Diego, and in this section, he lists all of the graduate programs he was rejected from, the employers that turned him down, and all of the papers, grants, awards he tried for and was denied. Most people, academics included, don’t keep a detailed record of everything that didn’t go as planned. And if the record is there, then it’s probably not public. Voytek’s inclusion is provocative because it goes against our idea of what a professional synopsis should be. More than that, it goes against what our idea of how we ought to present ourselves in public when we summarize what it is we’ve been up to. We show the highlight reel, not the whole game.
If I asked you about what happened in the big game last night, what I’m really asking is a question about the highlights. I don’t want to know the play-by-play recounted in real time. I want to hear about the game-winning drive, the walk-off home run, or the go-ahead goal scored in the final minutes. If you showed me a YouTube clip featuring a series of failed pass attempts or a few lackluster touches in the middle of the pitch, then I would think you misunderstood the question. Recounting the big game consists in reliving its most exciting moments. The same is true of how we represent our everyday lives to the people around us. And there is perhaps no more pervasive canvas for the highlight reel of life than Instagram.
What kinds of posts do we see from other people on Instagram? We see the best-of, the most exciting moments, the kind of moments that make us sit up and take notice of the incredible lives those around us are living. No one posts a photo of themselves going to the grocery store. Everyone posts of a photo of their recent family vacation to Hawaii. This dynamic is baked into the way we interact with this social media platform, and it presents some problems that are important to consider.
I recently decided to run a little experiment. Instead of posting photos when I did something noteworthy or out of the ordinary, I set an alarm. The alarm went off each day at a random time between eight in the morning and ten in the evening. Whenever the alarm went off, I had to snap a picture of whatever I was doing and post it to Instagram. No highlight reel, just the game. The result of the experiment was that I took more or less the same picture of me working at my desk every day. I had to discontinue the exercise because people were starting to message me to say they were sick of looking at my office space. But that was the whole point—the highlight reel is completely unrepresentative of the actual game. Instagram, in this sense, has a problem with statistical sampling.
Much of what a statistician does comes down to what are called sampling approximations. In technical terms, a statistician wants to know what a full distribution looks like but can only draw single observations one-at-a-time from this distribution. Imagine a bag of red and blue marbles. You want to know what percentage of the bag is blue marbles and what percentage is red. But you can only draw one marble at a time. You need to construct your best guess of what the contents of the bag looks like by only looking at one marble at a time. This is a sampling approximation. Most of what a statistician worries about when trying to figure out what’s in the bag is how she is going to choose her observations. When you’re considering a bag of marbles, it’s easy to figure out how to do this. You just reach in there and grab one. But when you’re picking a series of observations to understand something more complex—say, a representative moment from a baseball game or a career—how you choose the observations has a dramatic effect on what you think the real thing looks like.
You could also think of it as pieces in a puzzle. How many pieces do you need to see in order to know the scene it's depicting? Well, it depends. If someone is drawing the pieces from a pile randomly, you might be able to figure it out once they’ve drawn a few from across the scene. But suppose that person is cherry-picking the pieces to only come from pure blue sky—what statisticians call biased sampling. You might think that the entire picture is sky blue. However, there’s a critical piece of information missing: Are you aware that the person is showing you a biased sample? The problem comes when we mistakenly think that we’re being shown pieces drawn at random, but are really hand-selected from clear blue sky.
Usually, we’re aware that we are being shown a biased sample. We know the highlight reel is only the best moments. But this is sometimes harder to remember when we look at other people’s lives on Instagram or the accomplishments listed on their résumé—using single marbles to reconstruct what’s in the bag. We know what the whole game looks like for our own lives. We’re all too familiar with our own shortcomings and the doldrums of our daily experience. We don’t need a rejections and failures section on our CV because we live through it. But it’s important to keep in mind that everyone has a failures section on their CV, whether it's listed or not. And by leaving out this section—on Instagram as well as our CV—we present those around us with a biased sample of our lives to make it seem better than it really is. The thing we have to keep in mind is that so does everyone else.
And you know what? That’s not necessarily a bad thing. It’s just something you have to recognize. The trick is not to mistake the highlight reel for the game.