We can both agree that we need filters. We have hit a level of content where there are countless petabytes of data on the internet. Is your feeling that we shouldn't have automated filters, or that we shouldn't haveopaque automated filters?

My feeling is that the filters should be transparent in a few critical senses. People should know that the filters are on. Ideally, they would also have a sense of how they're evaluating information. I mean, when you think about the Facebook News Feed, I can make guesses about how that works, and Facebook has said a few little things, but really nobody has any idea how this system works. What does it prioritize? Also, ideally, you want to give people some control over this stuff, so that they can choose how homophilic their search results are. One of the thought experiments that I like is: Imagine that you have a slider at the top of Google that allows you to slide it from "Most narrowly like me" to "Completely different." That's less about people using it everyday than about calling attention to the fact that it's possible and that it's going on behind the scenes.

There are a couple of interesting directions to go from there. One of the dangers is asking for a purely unfiltered internet. Or even asking for the ability to filter the internet by yourself. Those are tasks that, at this level of development, are probably beyond us. I mean, a search engine, in a basic way, is kind of like our first filter.

I think the key is that you want filters with a certain level of integrity, and then you have to talk about what that means. And I would say that it would draw on some of the signals that are available besides simply, "What do people click on the most?" So, for example, on Facebook, the one signal that's really available is the "Like" button. I know you think a lot about foreign news and especially how that propagates to American audiences, and I just have to believe that it's way easier to like "I just ran a marathon" than it is to like "There's a civil war breaking out in the Congo." And so that story doesn't get propagated on Facebook, and you lose it. If you really want to build this so that people can share useful information, it wouldn't be so hard to build in some other signals, like an "Important" signal, and mix that in. To go a little further out on a limb, how about a "It took awhile to get into, but then it changed my life" signal. That's hard to capture if all you're looking at are "Likes."

In both examples you've just given, you've basically said, on search engines, it would be really good if we could have a slider to control how like us or how unlike us our results are. And you're sort of pointing to this idea of a Google that's independent of who Google thinks you are. You're asking for more options there. On Facebook, you're (I think very correctly) saying that the ability only to vote up rather than to vote down, and having no nuance in what it means to vote up — no difference between "you go girl" and "this is really important, pay attention to this" [is problematic]. If we had more control and nuance over these automated filters would that get us out of the problems that you're worried about? Can you imagine a Facebook with enough dials and levers that you would feel okay about the filtering, or do you just want it to shut off?

< prev  1  |  2  |  3  |  4  |  5  |  6  |  7  |  8  |  9  |  10  |   next >
  Topics: Books , Technology, Facebook, Harvard Book Store,  More more >
| More


Most Popular