Comments: My _Guardian_ column on Internet Group Polarization Argument

A thought about how to raise the level of political discourse on the Internet:

Suppose anyone writing a persuasive essay could submit it to a site along with a poll question like, "What was your opinion of Bush's energy policy before reading this post?" and "What was your opinion of Bush's energy policy after reading this post?" The site displays the essay to users, and polls them about whether it actually changed their mind.

The essays that are judged the "best" (where the reward is something like the reward on digg; they get pushed to the top and out to a wider audience), are the ones that score best on the criteria of how many minds they actually changed.

Not every essay that scores well on this criteria is good (trivially, an essay could score well on this measure if it "persuaded" people by telling flat-out lies), but the real point is that any essay that doesn't score well on this front, is useless. If it doesn't persuade anyone who doesn't already agree, what good is it?

For 90% of the political diatribes I see on the Internet, I want to reply to the poster and say, "Look. Do you seriously believe this post would convert a high percentage of people who don't already agree with you? If your answer is 'Yes', then you're delusional; if your answer is 'No', then you should admit that the real purpose was something else (venting, self-aggrandizement, showing off knowledge of lots of statistics, etc.)."

Now, it's true the rating system I described would be vulnerable to cheating. Specifically, if someone reads an essay that they already agree with, they could boost it to the top by responding to the poll saying "I didn't agree with this before, but I believe it now!" I'm not sure of the best way to deal with this. Perhaps every user of the site would register their current opinion on something -- "Bush is a moron, Yes/No" for example. And the site would also display the percentage of users' current opinions. You would want to register your opinion honestly in that case, because if you didn't really think Bush is a moron, you wouldn't want to contribute to the percentage of users who responded "Bush is a moron". On the other hand, if you read an essay that caused you to change your vote on that question, the site would register that and count it in that essay's favor.

But the implementation details are another issue. My point is that a requirement for a "good" essay is one that convinces people who did not already agree with it, and if there were a system for recognizing and rewarding content that did that, maybe we wouldn't see better-quality political discourse.

Posted by Bennett Haselton at July 10, 2008 06:14 AM

Nobody would know the site exists. Nobody would read the essays. And "telling flat-out lies" is actually a key issue, because that's very nearly what drives many (not all, but many) high-audience political sites.

The whole problem is there is no EFFECTIVE, widespread, "system for recognizing and rewarding content" that is accurate rather than popular.

It's not going to be solved with a website _per se_.

Posted by Seth Finkelstein at July 10, 2008 08:21 AM

Too many big words. Not enough punch.

... at least for me, this morning. I'm more in the mood to read:

"Critics blame blogs for ghettoization: They say groups of hyper-angry snarlers only talk to each other on the internet. But that's too easy. Those critics are morons, gang-banging on a soft target."

I know, that's not really your style, Seth. But still... Too many big words. Not enough punch.

Posted by critic at July 10, 2008 12:41 PM

I gotta be me :-).

Posted by Seth Finkelstein at July 10, 2008 01:28 PM

But the whole "nobody will know the site exists and nobody will use it" line could have been said about any site, like digg, or even Google when it got started.

Yes, most sites fail to hit the big-time. But that doesn't mean *all* of it will.

(And I'd mentioned in a previous post that if you find that there are too few people willing to read essays or rate content, you can always give the essay author the option to pay randomly selected people to rate their content, as long as there's no connection between the amount you pay and the rating you get.)

Posted by Bennett Haselton at July 10, 2008 06:02 PM

And there were very few times that line wasn't right - almost everyone won't win the lottery, even if someone does.

Repeat: Go code it if it's such a killer idea.

Posted by Seth Finkelstein at July 10, 2008 11:00 PM

It takes venture cap money to promote it and make it work :) And even then, yes, the odds are against you -- but that's true for *any* category of Internet site.

I think the first step is to promote this as a design for a meritocracy that is abuse-proof.

We used to think an idea for a good meritocracy was one where people could submit links, stories, ideas, etc. and then anybody could go and vote anybody else's content up and down. So people built sites that implemented that idea, and sooner or later one of them was bound to hit it big (digg). But then it became clear that, just like the blogosphere itself, digg favored the "A-list" and "in crowd", and could even be manipulated by buying votes directly.

Well, spread the idea that an abuse-proof meritocratic system would be one that uses a random sampling of voters rather than letting voters self-select on what to vote on. Then maybe a wave of entrepreneurs will try implementing that, and one of *them* will hit it big.

Posted by Bennett Haselton at July 11, 2008 05:10 AM

But maybe you're wrong, and your idea doesn't work, for the reasons I've given above. If you won't devote any effort to real-world testing, that's a powerful argument there's really nothing there.

[Tedious - of course it's not absolute proof in a strict deductive sense, but it's a good heuristic.]

Posted by Seth Finkelstein at July 11, 2008 11:12 AM