What the social web can learn from Lehman Brothers: social systems architecture

We did a car boot sale on the weekend.. and amongst the pile of books that we keep trying to sell but never do was one I bought a couple of years ago.  The topic of that book was how to price various kinds of exotic derivative contracts, as used in a range of complex ways to buy, sell, spread and mitigate risk and make loads on money for investment bankers.

Derivatives start out as nice simple ideas, but before you know it, things get very complex.  And if you trade lots of derivatives with lots of other parties, you end coupled to your trading parties in lots of weird ways.

So, what started out as a possibly sensible risk management exercise becomes so complex and you end up so tied to everybody else’s success that you’ve got a Lehman-sized problem.

The trouble here is:

Complexity:  it gets too hard for humans to understand what is going on. Everybody kind of pretends they do, but nobody really does.

Coupling:  it gets very hard to work out how to untangle the knot of relationships you’ve created.

Trouble. So the investment bank solution was to create more and varied derivatives to cover all of that, making it worse.  Pass the coke and let’s do some business.

What has this got to do with the social web?  Lots.

If we layer API on top of API, and couple all our social web services together without a lot of thought about systems architecture we run the risk of making something complex, over-coupled, and utterly unstable.

And when we start building businesses on top of all of that, things get risky for those businesses indeed. And if we start relying on complex, over coupled web services, what happens when something’s business model fails and it goes into the deadpool.  Could the whole lot fall over and be hard to get working again?

So, I’m thinking we need to start thinking about the discipline of social systems architecture, where we manage and look at the interconnected web as we do large distributed systems, taking note of single points of failure, instability, reliability, complexity and coupling.

Sure, we can integrate lots of social web stuff, but we need to keep a systems engineering mind while doing it.  Otherwise it is a bit like running your billion dollar derivative contracts out of an Excel spreadsheet.