Monday, April 04, 2005
Tagonomics, Pt 2: Whuffie vs Tags
Lots of buzz about whuffie going on these days. And tags continue to accelerate - her's a dating by tag site, and a piece in BW. I'm gonna argue that whuffie is a bad idea because it misunderstands resource allocation, and that tags are a far superior mechanism to allocate resources. The resource in question, is, of course, attention.
OK. Whuffie is a long way to describe computational reputation, with all the fun things that entails - it's essentially giving people cool points.
I have a big problem with whuffie. First, we have to critique the trivially contradictory notion of a 'post-scarcity economy'. The premise is that no good is scarce in the near-future fabconomy. Of course, there is a scarce resource even in the fabconomy - attention. So whuffie is a way to allocate attention via quantized reputation.
The problem with this setup is that it ignores history. History tells us that markets are much more efficient (read: fair) resource allocation mechanisms than networks are. You only have to look at the Magna Carta, the Soviet Union, or the Confederate South to understand why: throughout history, networks as resource allocation mechanisms have led inevitably to violence.
Diffused networks and cool points are excactly the wrong way to allocate attention, because they'll give rise to all sorts of inefficient (and nasty) dynamics, like cliques, violence, and wars, etc - because when it comes to scarce resources and open networks, people engage in massively opportunistic behavior to either grab the resources and run, or to win and maintain control over them.
How does this happen? Networks, unlike markets, magnify information asymmetries. Magnified information asymmetries lead to all sorts of hidden behavior and action: you tell someone one thing, but do another. Why does this happen? Because it's much more difficult to lie in a market than it is a network. There's a simple reason why: in a market, you only have one way to reveal your expectations and preferences - via the price mechanism. Lying or hiding information - by bidding on something you don't value, or by not bidding on something you do value - is costly (assuming liquidity).
Put more simply, hiding information/lying is instantaneously costly in markets - either you forgo gains, or directly realize losses. But in networks, the costs of lying can often be deferred, diffused, or externalized. So incomplete or bad information revelation is ex ante costly in markets, but ex post costly in networks. In the real-world, what this means is that lying is less risky in networks than in markets - which increases the relative cost of information in networks a great deal.
So compared to networks, whose dynamics can increase the costs of information, pushing them up to the point where all the info in the network never gets processed, the price mechanism works better because it's a cheap and hyperefficient massively distributed information processor.
Sure, we can argue about the evils of capitalism versus the goodness of networks. Markets do have their own problems - herding, information cascades, and, eventually, bubbles. But, at least in the medium-long run, they're far more efficient than networks at allocating resources - just imagine trying to use a giant LinkedIn to allocate grain futures. The transaction costs, info asymmetries, and opportunism you'd encounter would be stupendous.
What you'd really want, and this is where I think tags come in, is exactly the inverse of whuffie: not cool points for people, but attention points for objects. Which is exactly where tags are heading. Which is why I wrote this post!
Note, I won't call tags folksonomies not least because I think the word is really awkward, but because I think they have the potential to be much more than taxonomical mechanisms (as the above argument should make clear).