|
|
||||||||
Buy Charlie and the Chocolate Factory from Amazon.com and Amazon will suggest related products. The BFG by Roald Dahl. A Godiva assortment. Edward Scissorhands on DVD. You buy, Amazon learns, Amazon recommends.
Tell Pandora your favorite band and Pandora plays you songs. If you like a song, upvote it. If you hate a song, downvote it. Pandora uses your votes to find other bands it thinks you'll like. It creates a personal radio station that constantly adapts to your unique tastes.
You see a story on reddit.com about Barack Obama's childhood. You upvote the story, and now reddit recommends other stories it thinks you'll like. You start seeing more stories about Barack Obama.
We live in the Recommendation Age. Everywhere we turn, the world finds out what we want and gives us more of it.
This is a good thing. It's good for marketers. What we want, we buy more of. And it's good for us. We see more of what we want and less of what we don't.
But there is a hidden danger in the widespread use of recommendation engines. When all you see is what you want, you don't see very much.
Look again at reddit.com. Reddit is a news aggregator. Anyone can submit a story. Reddit shows the story to other users who say whether it appeals to them. Their votes decide whether reddit shows the story to more users or quietly drops it. It's democracy-as-editor.
But reddit recommends news based on what you've voted for, and herein lies the problem. The more you applaud stories about McCain, the more you see stories about McCain. Eventually you only see stories about McCain. You become a member of the McCain news ghetto.
Look up Audacity of Hope on Amazon—what recommendations do you get? Great Speeches by African Americans by James Daley, then five other books about Barack Obama. The customer who likes Obama will find a lot more reasons to like Obama.
Imagine a world where all your information—news, music, books—comes to you through recommendations. That's a world fine-tuned to leave your prejudices intact. Next time you go to the polls, you think, "All I hear about McCain is wonderful. How can anyone not vote for McCain?" The guy next to you, who lives in the Obama news ghetto, thinks, "All I hear about Obama is wonderful. How can anyone not vote for Obama?" One of you sees the world through rose-tinted glasses, the other blue-tinted. You're living in different worlds.
Of course, self-ghettoization is nothing new. We all tend to surround ourselves with Yes-men. Republicans watch FOX, Democrats watch CNN. Human beings generally avoid challenging ideas. We cling to the familiar. That's nothing new.
What's new about recommendation engines is the degree to which they make self-ghettoization possible. When people got their news through papers, magazines, and TV, they saw a lot of stories they never asked for. They probably ignored much of what they disagreed with, but at least they were exposed to it. Now, when you get all your information through recommendations, you agree with everything you see. What you don't see are new or challenging ideas—even those that might help you. You remain cocooned in blissful ignorance.
It's ironic that recommendation engines—such a populist idea, so democratic and egalitarian—present the greatest threat to free speech we've seen in the last hundred years. It used to be that a loud-mouthed maverick whom nobody wanted to listen to could still find an editor to publish him if his ideas were good enough. But in the Recommendation Age, the only writers who get exposure are the writers who say what we want. Writing and marketing have melded into one.
The people who get their news, books, music, and other information primarily through recommendation engines still represent a fairly small proportion of society. But that proportion is growing each year. Even traditionally monolithic news sources like CNN.com are moving into recommendation-based services. It's hard to imagine that ten years from now a majority of voters and consumers will still get their news through broadcast TV and newspapers.
Our society is already riddled with factions. Open discussion is becoming a lost art. As people come to hear more and more of what they already believe and less and less of what they don't, these divisions will only become deeper. If the election of 2008 looks like this, what will 2012 look like?
It's up to us to get out of our ghettoes. We should keep using recommendation engines, sure. But we should make a habit of seeking out new and unfamiliar news and ideas. Go read a book you wouldn't normally read. Visit sites you wouldn't normally visit. Watch a news program you would normally avoid.
A society where information ceases to flow is a dead society. It doesn't matter whether the information stops because of martial law, bad reporting, or a culture of convenience. If we close our ears to everyone but yes-men, we can enjoy a long period of peaceful agreement. But the longer we avoid controversy, the fatter and sloppier our ideas will become. The longer we hide inside our ghettos, the greater the shock when we discover we're dead wrong.
Labels: technology
Comments
Interesting post - I agree with the idea that shutting off information is not a good thing. However, I don't think that people are turning towards recommendation engines simply because they disagree with the sources, but that the sources don't work on level playing fields.
Traditional media (print, TV, and radio) gives us what it wants us to consume, and previously we had no choice other than to turn it off. History, even recent history, is repleat with examples of "old media" witholding or even modifying information to suit their purposes.
And being such an ecclectic music fan, I can never find the music I want in what was traditional outlets. Call me a "grumpy old man", but I have a really hard time trying to go through the chaff on the radio or music stores to find something that is actually good to my ears. Now, Pandora hates me because I'm always shooting down its recommendations, but I do see it as the beginning of a system that can slowly spiral my interests out beyond my current set of artists. That is, if they can survive their legal fights...
You are right, it is too easy to curl up in our comfort zones and not take experiential risks. Our free time (and money) is limited, and we want to make sure we don't blow it on a bad meal, a crappy movie, or a terrible CD, so we oftentimes play it safe. But sometimes boredom and discontent may prompt us otherwise...
Traditional media (print, TV, and radio) gives us what it wants us to consume, and previously we had no choice other than to turn it off. History, even recent history, is repleat with examples of "old media" witholding or even modifying information to suit their purposes.
And being such an ecclectic music fan, I can never find the music I want in what was traditional outlets. Call me a "grumpy old man", but I have a really hard time trying to go through the chaff on the radio or music stores to find something that is actually good to my ears. Now, Pandora hates me because I'm always shooting down its recommendations, but I do see it as the beginning of a system that can slowly spiral my interests out beyond my current set of artists. That is, if they can survive their legal fights...
You are right, it is too easy to curl up in our comfort zones and not take experiential risks. Our free time (and money) is limited, and we want to make sure we don't blow it on a bad meal, a crappy movie, or a terrible CD, so we oftentimes play it safe. But sometimes boredom and discontent may prompt us otherwise...
These are excellent points. It's definitely true that traditional "editor-picked" media had their downsides as well, media bias and all that. But then we also instituted laws to try to curb bias. That's harder to do, and probably too heavy-handed, in the recommendation world.
And I must say that a recommendation engine designing in the right way does exactly the opposite of what I'm warning about here. As you mentioned, Pandora doesn't really ghettoize you: it really analyzes your "ghetto" only so that it can then try to lead you out of the ghetto gently. So recommendation engines can be designed to lead users out into new territory.
Pandora is based on a genetic algorithm. One thing we've learned from working with genetic algorithms are the danger of producing systems that find "local minima"--easy solutions that quick to find but aren't really the best. This is analogous to the human phenomenon of short-sighted thinking, doing whatever feels good now rather than what will pay off more later. From AIG to drug addiction, a rather common mistake. Anyway, this again is the danger of recommendation engines, but Pandora--the makers of which are wise to this problem--seems to have solved it. I just hope other recommendation engine developers understand and fix the problem.
And I must say that a recommendation engine designing in the right way does exactly the opposite of what I'm warning about here. As you mentioned, Pandora doesn't really ghettoize you: it really analyzes your "ghetto" only so that it can then try to lead you out of the ghetto gently. So recommendation engines can be designed to lead users out into new territory.
Pandora is based on a genetic algorithm. One thing we've learned from working with genetic algorithms are the danger of producing systems that find "local minima"--easy solutions that quick to find but aren't really the best. This is analogous to the human phenomenon of short-sighted thinking, doing whatever feels good now rather than what will pay off more later. From AIG to drug addiction, a rather common mistake. Anyway, this again is the danger of recommendation engines, but Pandora--the makers of which are wise to this problem--seems to have solved it. I just hope other recommendation engine developers understand and fix the problem.
<< Home













Post a Comment