Community Structure Cause and Effect

Via Instapundit:

I can tell you exactly how a pointless blog full of poorly written, incoherent commentary made it to the front page on Digg. I paid people to do it. - I Bought Votes on Digg

The profusion of community sites on the Internet, each with their own technical quirks, has enabled an unusual opportunity to study the effects of community structure on the resulting community. I haven't made anything like a formal study out of it, but somebody should; it's strangely interesting in an academic sense, but also very practical as you decide what communities to participate in, and for a few, how to create a community.

An article about Digg is the hook for this blog post because I think for various reasons, Digg is in a full-fledged failure state as a community.... but then, this is a very loaded statement, because it critically depends on what you're looking for "Digg" to be.

In one sense, there's absolutely nothing wrong with Digg whatsoever. The software functions, handles loads that would cripple most sites, and carries on serving its community with admirable technical proficiency. But the implied standard of this statement is "a community is what a community is, and it can not be judged right or wrong". This is probably ultimately as true a statement as you can ask for, but humans can't seem to actually put it into practice and insist on judging things. I am no less human than anybody else, so I can't help but judge Digg by my own standards.

So, I personally consider Digg a failure as a community because the moderation system developed for Digg has resulted in a brutal tyranny of the majority. Digg is a great site for quirky links; as I write this, the number #2 story on the main page is "Man Learns How to Smoke Through Eye Socket, Photo Included!". (I think it was #1 a moment ago, but I'm not quite sure.) Offbeat links that cause no controversy result in a perfectly normal, fun, fluffy discussion; as I write this, only two comments have been "modded into oblivion", one that appears to be an utterly off-topic copy&pasted screed from a paranoid schizophrenic, and one a racist comment that is probably just a troll.

But you switch over to political news and it's a different story. It's not quite as biased as some people would claim, I think, but checking over the pattern of moderations in, for instance, Atheist group challenges Bush at Supreme Court shows a pretty clear preference for one set of views over another, a preference also somewhat more insidiously shown in the stories, and most insidiously of all, in which stories are "buried", that is, made to "go away" because many people are reporting them as abusive or spam.

Is this a failure of the "community"? Is it just that the prevailing Digg biases are so self-evidently correct and truthful that they've actually convinced everybody of the truth? No, it's a fundamental consequence of two simple facts: In controversial domains, people tend to vote for stories and comments they "like" which reinforce their existing beliefs, and that being part of a "successful" vote feels good, while Digging up a story that gets buried feels like getting rejected. The end result is that ideological imbalances in the user population get magnified; the slight majority votes up their stories, chasing away a few of the minority, making the majority that much larger for the next vote. In the Digg system, eventually the majority crossed over to the point where they could actively destroy minority stories.

Now, this is just an analysis of the first-order effects, which are never the whole story. There are some stubborn people in the minority deliberately trying to "fight back", but it's unlikely to "work". You might be able to win small victories on particular stories, but there's going to be no way to "take back the community" against the constant purification force exerted by the moderation system.

If you took the exact same technology platform, and stuffed the initial user population with right-wingers, or conspiracy theorists, or any other group, in three years, it would be extremely likely that the community would be even more firmly in control of the initial group than it was on day one. I think in general the final orientation of these communities is determined by the initial makeup more than any other factor.

Digg is a community that I can see this relatively clearly because I don't identify with the resulting consensus. One that I have to keep reminding myself is not mainstream is programming.reddit.com, a "programming" community that "mysteriously" manages to have nothing but bad things to say about the top five programming languages, and it's not even really happy until it gets to #8 (Python), although the true passion doesn't start until #15 (Lisp) and doesn't get into full swing until it falls right off the bottom of that list into the languages with communities so small that they end up lumped in "misc" (Erlang, Haskell, and even an odd fascination with Factor, though I'm increasingly of the opinion that's more because the Factor people basically spam the system than a true interest). Reddit also works on the "one man, one vote on infinite topics" system, and while the newer Reddit isn't quite as far along the Digg path as Digg is, I see no reason to believe that it won't end up in the same place, since it uses effective the same community structure. programming.reddit.com tickles my fancies a little better, and I recommend it as a great source of mind-expanding news if you've never been exposed to that community... but it is very firmly in control of one particular community; just try to get a story complimentary about Java onto the reddit. Go ahead. $10 says it doesn't even last the first hour probationary period. (I think stories have a 1 hour probationary period to allow opinion to settle before it decides what to do with it, but a strong enough negative reaction can bury it before that period is up; it just takes more negativity than usual.)

I think "one man, one vote on infinite topics" inevitably results in ideological "cleansing".

The Slashdot editors saw this trap clearly, and tried to design their system around it. I find this amusing as their moderation system predates so many sites that have subsequently fallen into that trap, and I think at least some of these systems were created with the explicit goal of being more "democratic" than Slashdot.... Instead, Slashdot select the moderators from the set of registered users that appear to be the "middle of the pack" in terms of usage and dole out the "moderation points" relatively conservatively. Another interesting difference is that comments can not be voted on infinitely even by the chosen moderators, unlike reddit or digg where a comment may go to +/- 1000; Slashdot constrains comments to the range [-1, +5]. I think this range has a biasing effect towards promoting stories, rather than burying them (though comment burying does happen); imagine if they ran on a [-5, +2] range instead, the effects would be radically different.

At the risk of stating an unpopular opinion... oh, wait, you can't really mod this down, can you?... OK, at no risk to me at all, I think the resulting community is actually currently stronger than Digg's or Reddit's in some critical ways. In particular, I find it easier to poke through a Slashdot discussion containing 1000 posts, and get just the best stuff.

There absolutely is a gestalt "Slashthink"; all communities will inevitable have some average opinion. But my experience is that it is far easier to get a comment that is extremely critical of "slashthink" up to +5 on Slashdot than it is to prevent a critical comment from simply getting buried on Reddit or Digg. The standards for such a +5 comment are much higher; your logic must be rigorous, your grammar must be largely correct, your points better be well-thought out, where a +5 Slashthink comment need mostly be spelled correctly and not painfully obviously a "karma whore", but I can attest that it's possible. I actually find it a very useful challenge level; if you can get a "non-Slashthink" comment up to +5 with no critical replies, you've accomplished something, I think. Such an endeavor is just hopeless on the more "democratic" sites.

(Democracy vs. Republic is probably a pretty good shorthand to think about the differences, too; Digg and Reddit are pure Democracies, Slashdot is a Republic with an exotic mechanism to select community representatives.)

For sites that have a voting structure, especially ones that use the votes to "bury" things, the effects of the voting system tend to dominate, but they aren't the only interesting parameters a community can explore. The comment structure can come into play too; do you have a flat or threaded comment system? The result conversation has certain characteristics and failure modes.

Threaded discussions are better able to take a point and properly deal with it, because even somebody coming to the conversation two days later can reply to any message. Flat comment systems tend to work more like a conversation; the topic wanders, and just as in a real conversation even returning to a previous topic (such as, say, the point of the original post) can result in something very disjointed. Threaded discussions have the failure mode of a deeply nested quote war between two participants (who aren't going to convince each other any time this century) who can't let anything go. On the web, this often results in comments that are increasingly thin and long as they get formatted further and further to the right. Flat comments often come across as the demented, disjointed ramblings of some sort of mega-entity that can't seem to focus on any one idea for more than ten seconds, and in practice any flat comments beyond about the tenth comment or so are probably pointless; the context has dissipated too far to be useful.

Again, neither is right or wrong, it depends on what you want. I've chosen threaded, because that's what I want here. My solution to the problem of deep threads, if it ever becomes a problem, will be to either cut the discussion off if it's just two people bickering, or if it's a valuable side-discussion, to promote it to a new discussion level in a new post and re-root it. I don't think there's a "solution" for a blog comment system that doesn't involve some time investment for the blog runner, if the comments are intended to be anything beyond a scribbling target for random internet people. (And hey, I suppose there's nothing empirically wrong with that desire either, though I question the utility of Yet Another Comment Box. I know that I'd rather invest the time to have a useful community than just leave put a comment box up and then totally ignore it; in that case, why not just turn it off?)

Then, there are even more exotic community shapes like the one formed by all weblogs, which have some exotic effects.

I don't have the community to play with here, but if I ever do, there are some exotic structures I'd like to play with and see what happens. One thing about iRi is that I intend for it always to be mine; if some sort of self-sustaining community were ever to develop (unlikely, but who knows what the next 40 years holds?), I'd want to spin it off rather than turn this into a group blog, or perhaps spin off my personal blog. On the resulting personal blog, I'd like to play with a voting system where you can vote any comment you want up or down, but you only see your own votes; the function of such a vote is an advisory one, to say "This is a comment that deserves being on the front page" or "This commenter shouldn't even be here", but with the final decision about what to do resting with me. I'm curious if this can harness the benefits of voting, while avoiding the negative consequences that result from automatically burying stories and the "cleansing" that results from publically visible vote counts, and the resulting treating of those vote counts like points in a video game. Maybe you could see the total votes on your own comments, maybe not. (That might still encourage scoring behaviors, but without the score being public, maybe not.)

One thing I've noticed by thinking how community structure shapes communities is that there is no way to get to an "unbiased" community, which isn't a surprising conclusion coming from me since I don't even accept the existence of "unbiased". So as I think about what I'd do if I had more of a community, I don't even try for "unbiased"; no matter what I do, the resulting community will be biased, almost certainly in my favor. The question is "can the biases be made interesting?", not "can they be eliminated?"

I think both Reddit's and Digg's problems actually stem from trying to create "unbiased" sources, to some extent. Trying to engineer the impossible tends to fail more spectacularly than engineering the merely possible, even when the latter fails. It may be "common sense" that handing everybody huge voting powers is the path to some sort of "democratic utopia", but it's a false idea nonetheless.

There's some sort of wonderful Sociology PhD thesis lurking in these ideas. (I'm not saying I've come even close to laying out such a thesis here, just that there is a good 10 years worth of study you could do on this topic. :) ) I've phrased this entire message in terms of the Internet so far because it has such a rich diversity of community structures induced by a rich diversity of technical structures (which I've not even scratched; one should also cover LiveJournal, LinkedIn, Craigslist, MySpace, Facebook, SecondLife, World of Warcraft, and a few others at a bare minimum to truly examine this topic, and I've not got enough experience to do those justice beyond knowing there is justice to be done), but it's also interesting to take these ideas and apply them to the real world and real history. One well-explored subtopic is the effect of television on politics; all politicians need charisma to succeed, but now you absolutely must be able to look good on TV, too. Another well-explored topic is how the acceleration of communication has accelerated our political cycles, an effect still being felt as weblogs take the news cycle up another notch in speed.

Off these beaten paths lie a host of other interesting questions, though; as distance communication improves, are people more likely to move farther away from their families and friends if they feel they can still communicate with them? Do they subsequently actually communicate with them? Do any such micro effects have an interesting cumulative macro effect?

I guess I don't have a pithy wrap up of any of this. I just think it's really interesting stuff, and like I said, surprisingly practical for anybody thinking about what communities they want to participate in, and certainly practical for anybody who wants to build one.

(Oh, hey, closing on the point you opened on; I guess that's an Officially Acceptable essay structure. Bah! I choose to snatch "rambling non-ending" from the jaws of structural cohesion! It's truth-in-advertising.)