I predict we will get better at prediction

By Panjandrum / 2 October 2012 / Intel Analysis, Technology

I’m not entirely sure that I agree with Roger when it comes to prediction, so here’s a piece from the opposing viewpoint.

Firstly, we all make some long term predictions. We have to. We put money into the company pension fund based on a largely rational assessment of our chances of living to retirement age. We don’t sell our houses during a recession unless we really have to, because we’re confident that the market will pick up again – sometime. Likewise, as Tom S points out below, any plan on how we deal with (for example) North Korea necessarily contains some inherent assumptions and inferences about what the future holds for that country.

This may seem like a very trite point to make, but perhaps it’s fundamental – I don’t think that anyone is seriously suggesting that we can’t make sensible predictions about the distant future, so the argument is about the nature and accuracy of the predictions we can and do make, and whether an inaccurate prediction is worse than no prediction at all.

Secondly, we’re getting better at the predictive sciences. Social Network Analysis, Game Theory, and Convergent Cross Mappingare just a few areas of scientific development which are increasing our ability to deal with the complexity inherent in the interrelation of multiple causal relationships, and the calculation of high-order effects, particularly when backed up exponential increases in computing power, as predicted by Moore’s law.

Sometimes the best shot is no shot at all.

Lastly, I’d like to draw a (probably slightly naff) analogy with marksmanship. If a trained soldier engages a target at 300 meters with an assault rifle*, you would expect all of their shots to fall within a radius of around 10cm or so. If you extend that range to 500m, it would a good marksman who managed to get the majority of the shots to fall on a man-sized target*. But the principles of marksmanship remain the same. If we want to increase accuracy at greater distance, we can do one of two things: Become better marksmen or get a better weapon system – ideally both. The links above suggest that those better weapons systems are coming into service. We, as analysts, had better start learning how to get the best out of them.

* For the sake of full disclosure: I’ve never been able to hit a man-sized target at more than 100m. I’m not sure if it was this, my complete inability to do drill, my utterly un-soldier-like demeanour or my occasionally ambivalent approach to physical exercise which prevented my meteoric rise to Brigadier. 


3 comments on “I predict we will get better at prediction

  1. Roger Davies

    I think part of the problem is having a mutual understanding of what prediction means. Let me give you exaggerated examples:

    1. The Brigadier who overtook Panjandrum in the career stakes, turns to his intelligence staff and says “ Tell me, O holy ones, what is going to happen on Tuesday next week?” I am only slightly exaggerating – this sort of demand for prediction does happen. Quite often consumers of intelligence have an unrealistic expectation of what is possible. It takes considerable dialog to resolve their understanding.
    2. Another Brigadier asks of his intelligence staff, “ Listen, geeks! I’m worried about next week. Can you research recent incidents and see if there are any patterns which might suggest when any activity may occur”
    3. A third Brigadier turns to his intelligence staff and says “ Listen ninjas, go and talk to your HUMINT sources, take this roll of banknotes, pay them, and get the man on the inside to tell us what’s going to happen next Tuesday.”

    All three ostensibly involve “prediction”. But only the first requires a crystal ball. The other two require resources and skill.


  2. Pingback: http en wikipedia org wiki Correlation does not… « Kathys LinkBook

  • Pingback: http intelmsl com blog 2012 10 i predict… « Kathys LinkBook

  • Leave a Reply

    Your email address will not be published.