I think the solution isn't to turn personalization off. I'm not romantic about the good ol' days when real people edited this stuff. But, at the same time, I think that right now we're in a dip in the transition between the 20th century ethos of information sorting and the 21st century ethos, where we have these algorithmic filters, but they don't really have built into them any of the best values that are embodied in the New York Times front page. So, for example, there's all sorts of important information in the design of the New York Times front page that really doesn't get replicated in a Like structure. Really what you want to be able to do is signal that kind of thing. In some ways, we're just looking at something that's in not even its adolescence but in its infancy. My concern is that from the companies' perspective, I'm not sure that these problems are at the top of the agenda. Because if there are more profits to be found here, they are very long term profits. I have the sense that when I go to Google versus when I go to Bing, I get sort of better results. That's not something that happens on a click-by-click basis. That's something that happens over months or years.
There were earlier approaches to this, before we saw companies like Google and Facebook doing a high degree of personalization. There was this great deal of interest in agents. So what happened to agents? Why don't I have my own personal information agent who is out there on Twitter and Facebook and everywhere else, finding me what I care about? Why didn't that future come to pass?
My theory on this is that if you make it explicit, it's obviously so dumb that you wouldn't ever want to use it. There were a bunch of products that came out in the mid-1990s, when this was in vogue, that were like, "Here's your personal eButler, and he's going to go find information for you." It was obviously terrible, so you didn't use it. People don't like to be told what they'll like or who they are, but if you just build that in, then they'll click on those things. People don't like to be typed, but if you typecast them, it's actually quite effective. So I think part of the failure of agents was that it was so transparent, and it was obvious how awkward it was. If you bury it under the surface and say, "Oh no, we're just showing you this website," then it's much easier to get people to engage.
So in some ways the fear is subtlety, right? Basically, personalization that wakes you up in the morning and says, "Hello, Eli, here is your Google today. We're working under the assumptions that you are still male, 30 odd years old," etcetera. Perhaps the ability to either alter those or turn those off, that wouldn't be as worrisome as [when it's] super subtle, it just happens, you have no knowledge of it, and you have no control.