As a generalist, I reserve the right to indulge in sharp changes of subject.
I’ll excuse this one by pointing out that, if democracy actually is nonsense, we’re suffering from a serious truth deficit. 9/11 Truthers are one thing—but when you start questioning 7/4, you’re into some heavy alternate reality. Here at UR, Philip K. Dick is always in the house.
So: a social verifier is an institution, authority, Web 2.0 server, etc., etc., which collects and distributes information that its users trust.
Wikipedia is a social verifier. So is the Catholic Church. So is the New York Times. So is UC Berkeley. So are Reddit, Digg, and the new Hacker News (the trusted information being “this link is cool.”) So is the scientific peer-review system. And so on.
This is a pretty wide range. Is the concept overgeneralized? Is there anything interesting we can say about all these systems in general? What could Digg and the Catholic Church possibly have in common?
Well, one possibility is that they both suck. No disrespect to Diggers or Catholics, but neither of these systems looks much like the kind of social verifier I’d like to see. Nothing like my dream SV exists, in fact. And I wish it did.
So on the off chance that any coders with a few spare cycles are reading UR, I thought I’d describe this system. Perhaps someone would be nice enough to build it. Unlike the rotary system (which is a joke, folks), it’s not patented. At least, not by me.
But it needs a name, so let’s call it Uberfact. Of course, who builds it gets to name it, but any system which follows this general design can advertise itself as uberfactious.
Uberfact, or any uberfactious SV, has three unusual features. One, it makes no attempt to separate fact from opinion. Two, its reputation system is factional. Three, its ambitions are unlimited.
Typical SVs today, such as Wikipedia or the New York Times, invest great effort in separating fact from opinion (see La Wik’s NPOV page). Terms such as “objective” are popular.
In my opinion, this reflects a fact which is quite central to Western history, but is seldom expressed as such. The fact is that information is power. In the democratic era this is explicit: who commands public opinion commands the State. Before democracy it wasn’t quite this simple, but ideas have always mattered. Anyone who can persuade others to share his or her opinions is powerful by definition, and very likely dangerous.
The attraction of depoliticized information, of objective truth, is obvious. Facts threaten no one. How could they? An opinion is an interpretation of reality, not an argument with it.
And so the democratic state, which after all is a state and must defend itself like any other, tends to favor objective SVs over those which propagate explicit perspective. Democratic society has integrated this bias so perfectly that it’s hard to imagine life without it.
For example, racism is widespread in Western society. Or at least it supposedly is. And it certainly once was. So why isn’t there, or wasn’t there, a racist TV channel? Surely, in our brave new world of 300 channels, there’s enough audience for Confederate Racist Television? Can you imagine the six o’clock news on CRT?
If you can’t, you can visit the web site “South Africa Sucks,” which I refuse to link to—if you Google it, you will see why. Perhaps CRT’s anchorman would be someone like SAS’s “The Uhuru Guru.” Can you imagine a world in which a child could start in racist kindergarten, continue through racist elementary to racist college, then go to racist journalism school and become a racist reporter for CRT? Which might be part of a whole racist media empire?
If you happen to be a racist yourself, perhaps you find this prospect enticing. Try replacing “racist” with “Communist,” “terrorist” or “jihadi.” There are indeed terrorist kindergartens in the world, but fortunately none of them are in New Jersey. At least not yet.
This is why the idea of “objectivity” is so critical to the democratic system. By attacking opinion and perspective in general, it suppresses all kinds of thought, but the thoughts it suppresses best are the most unusual, and therefore the most dangerous. If your goal is to eliminate POV from Wikipedia, for example, the hardest kind of POV to eliminate is the POV of the mainstream status quo.
So a tradition of neutrality has the inevitable effect of centralizing and standardizing opinion. A European or American intellectual of 1907 would be shocked and appalled by the society of 2007 in many ways, but I think his general impression would be one of great mental conformity. It’s much easier to find popular opinions of 1907 that have no living parallel in 2007, than the reverse. (Gay rights is the only major innovation I can think of.) The process of memetic extinction is quite advanced.
But note that I am thinking in just the same way as the partisans of objectivity. I am describing the aggregate social impact of neutralist social verifiers. Actually, on balance, I think this impact is positive, because I really have no desire to live in a city where there’s a racist kindergarten on one side of town and a terrorist kindergarten on the other. (At least not if that city has a single democratic government.)
For Uberfact, what really matters is that objectivity is not what users want.
As a user, what I want from Uberfact is an infinite extension of my own personality. If I ask Uberfact some question Q, the best answer I can possibly receive is the answer I myself would produce, if I had infinite time to research Q and knew all the information that anyone knows about it.
For example, if my question is “where was George W. Bush born?”, an objective SV like Wikipedia will give me a very reliable answer. But if my question is “is George W. Bush a tyrant?”, I am SOL.
Wikipedia cannot possibly answer this question. And if it even came close to trying, I would have no reason at all to trust it. The answer would simply reflect the collective opinion of La Wik’s admins. I’m sure the admins are great people, but why should I care what they think?
What I want to know is: if I knew everything that anyone knows about George W. Bush, and everything that anyone knows about the history and etymology of the word “tyrant,” would I decide that the former is a case of the latter? Surely, if Uberfact can answer this question—and answer it with a mouse click, not a week of research—merely objective questions, like where the tyrant was born, will be no sweat at all.
So if Uberfact can solve both of these problems—if it can deliver the goods both subjective and objective—it can simply walk past the epistemological landmine of distinguishing the two. As an Uberfact user, what I get is my interpretation of reality.
Can Uberfact do better than this? Yes, in fact, it can.
Suppose my interpretation of reality is bad? Suppose I am simply wrong? Suppose my opinions are stupid? Well, of course, most people with stupid opinions are perfectly happy to live with them. Indeed they tend to insist on it.
But some of us are so crazy that we actually like to improve our understanding of the world. Uberfact would certainly be defective if it didn’t assist in this process.
Therefore, Uberfact should tell me not just what I think, but what others think. I should be able to see everyone’s interpretation of why George W. Bush is, or isn’t, a tyrant. Who knows—maybe they’re right and I’m wrong.
I’ve piled a lot of feature requests onto this product. I haven’t said anything about how they’re implemented. The MRD is getting fat and nasty. It takes all day to come out of the printer. How, exactly, can Uberfact produce these magical services?
Enter the world of factional reputation.
The error that most reputation systems make, I think, is that they assume a homogeneous and unstructured reputation environment. The natural impulse of any good programmer is to generalize and simplify. So we see SVs in which every user has a trustworthiness bit (like Wikipedia’s admin flag), or a trust rating / karma as in many discussion boards, or even if you get really fancy a trust graph of who trusts who else, à la PageRank.
None of these has anything to do with the social structures that human groups actually form. Humans are what primatologists call a party-gang species, which means exactly what it sounds like. We have a seemingly irresistible urge to form violent alliances. For the human male, there’s really nothing as fun as getting a bunch of the guys together, swimming across the river, ambushing the two-legged scum who live there, burning their village and enslaving their children. And the human female is even worse.
Furthermore, a graph is a somewhat obtuse representation of the reputation system within these gangs. Sure, every tree is a graph, but if all your graphs are trees, use a tree. Human status systems are without exception hierarchical. They have regal aristocrats at the top, arrogant henchmen right below them, and so on down to cringing, boot-licking peasants.
The whole idea of democracy, which of course comes out of Protestant Christianity, is that we can defeat these tendencies, and emerge into the Millennium, the New Jerusalem in which all are equal. Well, possibly. It would certainly be nice. If you find anywhere that I can place a bet on this one, please let me know.
The idea of factional reputation is that, at least while the New Socialist Man is still stuck at version 0.43, we can actually work with human nature as it is, not as it should be, and build Uberfact around these notorious primate pathologies.
First, we are going to compromise Uberfact’s feature space a little. It will only work, at least work well, for those of us who are basically conformists. For example, I am a formalist and a neocameralist, and while there may be one or two of the former by now, I am quite sure I’m the only one of the latter. So Uberfact won’t work for me, or for other eccentric weirdos.
Second, Uberfact will only answer questions which many other people who think like you care about. If you are the only Sufi who cares whether the Yankees are better than the Mets, Uberfact cannot help you.
Notice these group labels. In Uberfact, these are called factions. Factions are groups of people who see the world in the same way. Factions may form on any issue and for any reason—progressive vs. conservative, Ford vs. Chevy, emacs vs. vi.
Any user can have a reputation in as many factions as he likes. But reputation in one faction has no meaning to another faction. To a Ford-lover, it means nothing that you’re a highly rated libertarian. What do you know about limited-slip differentials? Jack. Until you prove otherwise.
Every contribution to Uberfact must be associated with a faction, and it is judged by that faction and that faction only. If the contribution is good, it improves your local reputation within that faction. If I have something to share about Ezra Pound, I have to decide whether I’m saying it as a modernist, a postmodernist, a New Critic, etc., etc.
Factions are self-constituting—they are responsible for their own reputation algorithms. Anyone can start a new faction for any reason, but generally they form by the usual process of human group formation—one group gets too large and quarrelsome, and splits into parts. The faction’s founders constitute and manage its reputation system.
For example, early in Uberfact’s development, there would probably be a libertarian faction. This would then fragment into Rothbardian, Randian, and Kochian libertarians—at least. Various strongly-flavored personalities might spin off their own little factions, and so on.
As a user of Uberfact, you have access to all content produced by all factions. Your process for answering a question, such as “is George W. Bush a tyrant,” is in two steps. One, figure out what faction is both (a) interested in this question, and (b) reasonably aligned with your own perspective in the area. Two, find out what that faction says about George W. Bush.
For example, it’s easy to imagine upgrading Wikipedia to be uberfactious. Instead of one page for George W. Bush, you could read the story of George W. Bush according to libertarians, according to progressives, according to jihadis, racists, Ford lovers, emacs bigots, and so on—anyone who cares enough to have an opinion about George W. Bush.
One might quickly notice that these pages matched in certain details. For example, jihadis, racists, and progressives probably all agree that George W. Bush was born on July 6, 1946. So all of these groups might contribute to a consensus page, signed by a large number of factions, which might even be similar to today’s “objective” page. And since this would probably be the most commonly requested George W. Bush page, it would come up first. An uberfactious Wikipedia doesn’t need to be any harder to use than today’s neutralist Wikipedia.
However, it would be largely free from “edit wars,” because warring gangs would rapidly organize into factions and maintain their own forks of disputed pages. Note the difference between this and the existing bad practice of POV forking, which screws up the Wikipedia namespace. Note also the difference between uberfactiousness and system-level forks such as Conservapedia—there is no easy way to compare the views of Wikipedia and Conservapedia on any topic.
The Uberfact approach should also be effective for much smaller and more ephemeral questions, like “what are today’s top 10 cool links?” This is a subjective question, just like “is George W. Bush a tyrant,” and it demands a subjective answer.
Link sites like Reddit and Digg tend to suffer a kind of democratic degringolade, in which they start out cool and gradually transition to a point of total lameness. That this is the obvious consequence of universal suffrage on the Web should go without saying, at least to anyone who remembers Usenet in 1992.
Paul Graham’s design for Hacker News tries to avoid the degringolade by actually using an oligarchy of human editors, including Paul himself, who will tweak hidden reputation scores. While this will certainly be an improvement on Reddit, I find it excessively algorithmic and antisocial. It has gotten past democracy, but it’s not yet unapologetically medieval.
What I’d like to read at Hacker News is simply the set of links that Paul himself finds cool, or would if he had 80 hours a day to surf the net for links. Paul does not, in fact, have 80 hours a day to surf the net for links. But perhaps he has fifteen minutes to rate would-be toadies and henchmen, of whom he has I’m sure a large supply, who could then rate submitted contributions, and so on, producing a kind of ersatz impression of a massively overclocked Paul.
Eric Clapton may be God, but I refuse to believe that anyone else is. So there must be someone who has better taste than Paul. He or she can rise in Paul’s hierarchy, then defect and form his own faction—teaching the master a lesson in hacker cool. And so on.
The point of factional reputation is that, since all reputation is in-group reputation, a faction that succumbs to democratic degringolade, or to any other social disease, will simply sink in importance and prestige on the system at large. If the Paul faction is doing just fine, and then Paul drops too much acid and becomes a born-again Mormon, and kills everyone’s reputation unless they promote links about the angel Moroni, someone else will step in and feed our need for crack, excuse me, links.
Finally, in the usual tradition of pseudonymous Internet hypesters and vapor-peddlers, I refuse to believe that there are any limits on the power of Uberfact.
The construction of scientific and historical consensus, for example, is a perfect problem for an uberfactious design. Academics have always formed factions, and always will. They have always rated each other, and always will. Academic gang fighting is brutal, and it is conducted with lethal weapons. Get someone’s funding pulled and you can kill their career.
In Uberfact, all this fun backstabbing can be totally open and official. There is no reason at all why an uberfactious design can’t validate original research or rank researchers—within factions only, of course. Peter Woit versus Luboš Motl? Steve McIntyre versus Michael Mann? Bring it on, baby. These fights are simply gorgeous spectacles, and we should see them up close and personal. Ideally, I could click back and forth between the McIntyre and Mann versions of the hockey stick story, for example. Who needs ESPN Classic?
The same is true for literary critics and writers. Again, writers have always formed gang families and will always form gang families. A factional reputation system would just take existing networks of blurb backscratching and make them official.
And then there’s journalism. Ah, journalism! But I’m afraid that’s another post.