Method, not ideology

Over the last few months, I’ve done a succession of posts which can be read under the tag TWOS, for the war on stupidity, which explore various forms of ideology and consensus. More related posts will be tagged as I go. But is this just an exercise in pooflinging? Beyond tirades about managerialism, Tony Blair, bad engineering, and other stupidity-generating institutions, does TYR offer anything? Daniel Davies, of course, would argue that poo-flinging is indeed enough, for a variety of reasons.

He’s right, up to a point. After all, when the forces marshalled behind incredibly bad ideas are so powerful, who would want to waste time discussing alternatives when you could be concentrating your fire, or rather poo, on the enemy? And his rationale – that most people can identify the flaws in a proposal, but coming up with ones that have fewer requires ability – is reasonably persuasive.

But I think this doesn’t go far enough. Stupidity in organisations is like noise in information systems. Claude Shannon worked this stuff out at Bell Labs in the 1940s, when he theorised that the factor governing the informational throughput of any communication channel, all other things being equal, was the error-rate. Therefore, for a given bandwidth, the fastest link is the one with the better error-cancelling procedure.

We can see similar processes at work throughout the natural world, and throughout society. Evolution, markets, debate – these are all processes that create a big pool of errors, and then use a stupidity-elimination process to sieve out the least silly. Then shuffle, recombine, iterate, and destupidify. The persuasive force of this is well shown by simple computer simulations – like this one, ICE, which aims to defeat the argument against evolution from irreducible complexity. ICE sets a simple challenge, to catch as many randomly dropped balls as possible using crosses on a grid. Its organisms are randomly-generated, then tested and ranked in order of fitness. Then they are recombined, with random changes, and the whole thing is run again, with those below a threshold level eliminated. Evolution is visible within two or three iterations.

So, here we are at my first point. The Redwood consensus, as we identified in this post, relies on the creation of anxiety about security issues that the core executive of the state can offer relief for, as a substitute for anxiety about economic issues that the state will offer no relief for. It further assumes that a managerialist elite consensus knows what to do on all issues.

Clearly, this is highly stupidogenic. Managerialism relies, after all, on the use of pseudo-scientific methods to enforce compliance with the managers’ a priori beliefs. The deliberate exemption of a large sector of the political sphere from normal debate is at the heart of the consensus – UK-US relations, control of drugs and borders, the workplace. Where de-stupidising processes are not at work, stupidity accumulates.

Therefore, we need a more hostile memetic environment.

But that’s not all. If we want faster memetic evolution, as well as sharpening the stupidity-remover’s blade, we need to increase the size of the pool of errors behind it. I’ve said before that I can’t understand why conservatives, and for that matter right-libertarians, think that innovation is best encouraged when the cost of failure is maximised and the barriers to entry high. Consider the experiment Jonah Lehrer describes here, in which monkeys were raised in three environments of varying richness. Poverty of experience had visible effects on the monkeys’ brains. Interestingly, though, the benefits of a richer childhood showed diminishing returns – above a certain point, the monkeys derived no further benefit.

Now think of society. Can anyone seriously argue that a few percentage points shaved off Bill Gates’ income would deter any significant innovation? Can anyone seriously deny that a few tens of millions of dollars wouldn’t have a seriously beneficial effect on a significant number of children in, say, Africa? (Bill Gates certainly wouldn’t, after all, as he is giving substantial amounts of his money to them.) Just as importantly, trying to do anything new needs space, time, and freedom. And, as we pointed out, there is a sense in which greater equality is greater freedom.

But wait. Isn’t this contrary to our first principle? Might there be some awful social failure mode concealed in a left-libertarian utopia? You’d be right. If there’s one philosophy that has achieved more than any other and is still to cause any pyramids of skulls, it’s scepticism.

Now, scepticism may not tell us very much about what to try, but it does have a built-in stupidity-reduction process. Even if you’re a fascist, if you are a coherent sceptic you won’t be able to do too much damage. Whatever your ideology, it doesn’t matter, so long as you get your methodology right. Rather than thinking about end-states, utopias, and anti-utopias, wouldn’t it be a more robust practice to think about processes, methods, and principles that minimise stupidity and maximise creativity? Another lesson from evolution is that incremental steps towards problem-solving are more likely to hit the target than revolutionary change.

This brings us back to the importance of negativity as a creative force. If democratic participation, evidence-based policy, and other nostrums are to have any meaning, they must have one vital feature – they must be able to force the government, the management, or whoever to change course. This is what Tony Blair’s friends fail to realise whilst havering about “engagement”, “community” and other pabulum – it won’t gain anyone’s trust whilst the only result of a negative answer is that Blair’s office sends out millions of e-mails to tell the citizens that they are stupid.

One of the most important reasons we need stupidity-removing institutions is control lag, coupled with the salience heuristic. As a rule, people overestimate the importance of the loud, the obvious, the dramatic, and the immediate. Equally, they find it difficult to manipulate anything when the response to their actions is delayed or ambiguous – an excellent example is Goodhart’s law. Lag tends to cause exaggerated control input – the longer you wait, the greater the temptation to press the button again. (Two words: John Reid.) The end result can be a positive feedback loop, with the deviations getting bigger and bigger as you struggle to get ahead of the cycle.

It’s another reason why politics should be difficult. It’s also an argument against hierarchy. John Boyd’s concept of the OODA loop, drawn from his experience as a fighter pilot, argues that in any competitive activity, the actor with the fastest process of observation, orientation, decision, and action will win. Boyd argued that this implied a flatter command structure for the military, among many other things. Similarly, David Stirling originally thought that the SAS’s four-man teams would prevent a leader emerging in each. Empirical data shows that small teams capture most of the benefits of aggregating information. At a lower level, lag and information loss are always less the shorter the link. People who actually do the job usually know how it works, and anyway will find out first if they are wrong.

To recap briefly: ideas are not the problem, as they will be generated in conditions of freedom and maximised horizontal exchange. Stupidity elimination is the problem. Hierarchy is the problem, management is not the problem. Final goal targets are not the problem, psuedo-statistics is the problem.

And most importantly of all, if we’re serious about a new left-wing consensus, we ought to install it on top of a sceptical operating system.