Just Another Roll of the Dice

It's ba-ack …

Remember the Terrorism Futures Market? Sorry, the "Policy Analysis Market" ("PAM"). That's the one which, as described by CNN, would "allow traders to buy and sell contracts on political and economic events in the Middle East, including assassinations, the overthrow of regimes and terrorist attacks. " As Government Computer News noted, "The market, which is intended to be an analysis tool to track and predict events in the Middle East, was developed for DARPA by Net Exchange of San Diego. The storm of congressional and public protest when the program was revealed resulted in its being shut down within 24 hours " It's now been reconstituted as a purely private venture. So, no problem, right?

Well, wrong, IMHO.

The "Policy Analysis Market" is based on the theory that this mechanism would develop better "guesses" as to likely terrorist activities. "It originally was developed and funded with the assistance of the Defense Department, where officials cited the uncanny ability of other futures markets to predict election results, weather patterns and other complex events." Let me suggest that you don't need any background in statistics or other mathematics to spot the flaw in this theory. Prediction is basically an attempt to discern patterns within a system, or to put it another way, to "model" that system. An electorate or the weather are two examples of systems which can be modeled, with the hope of finding such patterns. The financial markets are another example of such a system.

The key concept, though, is that these are "systems". As such, it is possible for some action to take place within the system, and have an effect. With weather, well, we can try but we can't do much about what will happen; it's classic "chaos theory", with uncountable inputs making up the system. Elections may be more easily manipulated, but in the larger picture it would take a concerted effort to overcome all of the other inputs in the system.

What about terror? Well, first off, it is a system with fewer inputs than the weather or a national election. Suppose one of the intrepid prognosticators does see a pattern developing, such that there is a possibility of a terrorist action? Although in weather there is nothing that can be done, and in an election there is little (if anything) which can be done with any certainty, this is not the case with the PAM. The whole idea is to find a way to determine where a terrorist activity will take place, and prevent it. Thus, action is taken, based on the modeling, and such action then, itself, becomes an input into that modeling.

To put it another way, if PAM is showing that a particular event is becoming possible, then action will presumably be taken to prevent that event. The fact of such action may, itself, influence the potential events which might be imminent. So, where does that leave us? "The observer becomes part of the observed system." It is an inexact analogy, but this is the same dynamic encapsulated in the Heisenberg Uncertainty Principle: The outcome of an experiment depends on the view of the observer. The observer has an effect on the actions of the observed.

Or, to put it another way, unlike the weather or elections, the developers of PAM want to predict actions in a system on which they, themselves have a major effect. Would you take that bet? Another way to say this is, PAM may provide a false sense of security, not because anybody will bet on his own terrorist actions (and clean up), but because the purpose of PAM is to change the inputs, in the very areas it is trying to develop predictions. That may make PAM more like Laplace's Demon, a very theoretical creature. But, that also means that maybe we can't know what is going to happen.