Shaffer - Data Versus Democracy: How Big Data Algorithms Shape Opinions and Alter the Course of History

As long as weve had information, weve had disinformation . As long as weve had advertising, weve had attempts at psychographic profiling. Ever since the invention of the printing press, weve had concerns about the corrupting influence of mass media. But there are some things that are new in the past decade. Information is abundant in a way we couldnt conceive of just a decade or two ago, and the new science of recommendation enginesalgorithmically selected content, based on personal data profilesdominates the modern media landscape. In this introduction, we will clear away misconceptions and focus on the heart of the problemthe intersection of information abundance, human psychology, user data profiling, and media recommendation algorithms . This is where propaganda finds its way into modern society.

Have you ever used a search engine to find a stock photo? Maybe you needed a slick header for your blog post, some scenery for your family holiday letter, a background for an office party flyer. The results can be pretty good, especially if youre on a professional stock image site (and know how to choose your search terms).

But have you ever used a regular search engine to find images of something generic? Try it sometime. Search for images of doctor , then nurse . Professor , then teacher . What do you see?

Chances are you found some pretty stark stereotypes . White-haired professors, wearing tweed, lecturing in front of a chalkboard. Teachers also in front of chalkboards, smiling at their eager pupils. Doctors in white coats, deftly wielding their stethoscopes or making notes on their patients charts. Nurses in blue scrubs, also masters of their charts and scopes. You get the picture.

Walk through a school, a university, a hospital, a general practitioners office, though, and youll see little that matches these images. Most schools and universities abandoned chalk long ago, in favor of dry erase boards and electronic projectors. And lecturing to seas of students in rows is increasingly rare, particularly with young students. And, by the way, all of these professionals tend to dress less formally, and certainly more diversely, than the subjects of these search result images.

Search engines dont give us reality. They give us the results we expect to see. Using a combination of human programming, data from user interaction, and an ever-repeating feedback loop of the two, the results of these searches gradually become more like the generalizations in our minds. The stereotypes we hold in our minds determine what we click on, and those clicks form the raw data used by the search engines algorithms . These algorithms, in turn, form generalizations of expected user behavior, based on our collective clicks, and serve up the results were most likely to click on, based on that data. We perceive, we generalize, we search, we click, the machine perceives (the clicks), the machine generalizes, the machine returns results.

It doesnt end there. Those search results get used all over the web and in print (isnt that why we were searching in the first place?). Those images become part of the backdrop of our view of the world and further fuel the generalizations formed by our mind . Thus forms an endless loop: human perception human generalization human behavior machine perception machine generalization machine behavior human perception human behavior And in each turn through the feedback loop, the stereotype gets more stereotypical. Reality is lost. Though, because the stereotypes become part of our media landscape, in a very real sense, reality is also formed .

But did you notice something else strange about those image results?

How many of those doctors were men, and how many were women? What about the nurses? According to The Wall Street Journal , 32% of doctors in the United States in 2012 were women, and the proportion is rising. How did your professor search results compare?

Chances are your search results were more stereotype than reality. Thats partly our brains fault. Our brains make generalizations about what we perceive in the world, and those generalizations allow us to make predictions about the world that help us interact with it more efficiently. Cognitive scientists have also found that when we form generalizationscalled schemas we tend to define those schemas , in part, by contrast with other schemas. In other words, we emphasize their differences , often making them more distinct from each other in our minds than they are in reality. While this method of defining ideas and categories in our mind is usually helpful, it sometimes works against us by reinforcing the bias of our environment, including the (stereotype-ridden) media we encounter. And when the feedback loop of human generalizations, machine generalizations, and media representation goes online, that bias gets propagated and reinscribed at near light speed.