There’s No Such Thing as an Objective Filter: Why Designing Algorithms That Tell Us the News Is Hard
We are all immersed in an incomprehensible abundance of available information, and we can only read or watch or consume some meaningless fraction of it. What we see, and what we don’t see, is heavily mediated by information filtering algorithms: Google web search, the Facebook news feed, personalization and recommendation engines of all kinds. Filtering algorithms shape our knowledge and our society. They are now a permanent part of what it means to perceive the world.
So it’s important to design them well, to make good choices here. What makes a “good” filtering algorithm? There’s no easy answer.
How can there be? The question of who should see what when is not like the question of how far the moon is from the Earth, or what the mayor ate for breakfast today, or whether Elvis is still alive. Those questions have simple answers that are either right or wrong. But choosing who sees what information is not like that at all: There’s no one answer, just many possible visions of the type of world we’d like to live in.
It’s not just about code. All filtering algorithms operate on human input.
Yet we still need answers. Otherwise there’s no way to build the algorithms we must have, and no way to critique them.
Different disciplines have different approaches to this problem. At the risk of caricature, let’s say there are two broad camps here. The “technologists” are engineers, computer scientists, people with training in quantitative fields. They are the people likely to be directly responsible for building our filtering systems. The “humanists” are editors, curators, writers, sociologists, humanities scholars. They spend their life on the handcrafted work of deciding that this, and not that, deserves our attention — or examining the consequences of such choices. These two cultures need each other, but they don’t seem to speak the same language.