Thursday, August 1, 2013

The supidity of intelligence

French version : "L'imbécillité de l'intelligence"

Le Monde has published an enlightening article on the Snowden case (Aymeric Janier, " target=”_blank”>Keith Alexander, le « pacha » de la NSA", Le Monde, July 15, 2013).

37,000 employees, a budget of about $ 10 billion, ultra-powerful computers, the ambition of "intercept everything about everything, everywhere" ... We guess in Alexander's behavior a bureaucratic delirium : who could refuse him budget and computation power after September 11 2001, which has spread this obsessive fear that is the victory of the terrorists ?

Intelligence is still based on a tradeoff between observation and interpretation - or, in its language, between collection and analysis. To collect many facts is useless if you do not know how to interpret them, and the skills which are necessary for analysis differ from those used for collecting and processing of data.

Computerization facilitates the collection, data computer itself provides powerful analysis tools. One is tempted to believe that the collection, equipped by statistics, may be enough to know everything.

Finance has been the victim of a similar illusion: while his art lies in the trade-off between risk and return, the power that computer technology brings has reduced the feeling of risk (but not the risk itself). This resulted, this will result in disasters. Similarly, the power that IT brings imbalances the trade-off between collection and analysis, but to observe everything is to understand nothing.

To observe everything is to understand nothing

Looking at yourself, you see that your intellect is necessarily selective: in the complexity of the world we must choose at any time to see what is important to our action, and therefore to do not see the rest. A distracted driver who looks at the details of the landscape is dangerous.

We have to be sure that this selection is relevant, i. e. appropriate to our action, and we have to change it when the orientation of our action changes. The fact remains that a partial blindness is necessary: ​our intellect would be overwhelmed if we did not select the signals in our perception.

Statistics and Economic Theory

I opened in the 70s the course of data analysis at ENSAE (Ecole nationale de la statistique et de l'administration économique). Like the X-ray allows to see what is behind the opacity of the human body, the data analysis allows to see things that the mass of statistics hides. The results of the various methods are accompanied by "interpretive aids" which show in a few hours of work structures that a statistician, without these tools, could only perceive after months of work.

However the "interpretations" that are based solely on statistics are often wrong. Distributions and correlations that data analysis displays are indices that, as in a police investigation, direct towards understanding but are not sufficient to achieve it: indices can be misleading.

Interpretation will be always naive if the statistician ignores the treasure of past interpretations which condenses in the axioms and results of the economic theory. But the temptation is strong: Jean-Paul Benzécri argued that data analysis showed "the pure diamond of true nature." Econometricians themselves – even if their discipline is related to the theory – make sometimes the same mistake by precipitation.

Information and Culture 

This same situation arises in intelligence. Suppose that you are responsible for observing a Middle East country using all the resources of electronic espionage: you receive transcripts of phone calls, emails, Web accesses etc.

To interpret them you need to understand the dialects which are spoken in this country, but it is not enough. Interpretation requires also that you know its history and geography, its literature, its religions, that you are aware of the political situation, the biography of the most prominent personalities, the conflicts between them, the movement of ideas and opinions that are rooted in the past as in the present, etc.. This culture provides the concepts necessary to the exercise of your judgment, the theoretical basis of your interpretations.

  A good analyst does not want an exhaustive observation which would be overwhelming: he is rather an actor in the dialectic between culture and observation. Culture guides the observation to the most fruitful fields; observation leads to specify, complete and sometimes modify the culture, and its most important lessons are those that contradict conceptual prejudices.

The acquisition of such a culture requires hard and long work, which can be bearable and fruitful only if it is driven by a passionate interest. This is among the British that one find the greatest number of people who are interested in other countries because they want to escape the narrowness of insularity: think to T.E. Lawrence, to Leslie Blanch, to Richard Burton, etc.

But such characters are rare in the United States because Americans, believing to live in the best country of the world, have not as much as the British the need to immerse themselves in another culture. Assuming there is nevertheless one of these characters at the NSA, he can not uphold his needs in a bureaucracy that thinks only of collecting more and more observations, of accumulating more and more computing power and algorithms. More than moral indignation, the absurdity of this situation will push him to leave: it is perhaps what happened to Snowden.

*     *
In intelligence, as in finance, excessive confidence in computing power is accompanied by a contempt for common sense: while bad professionals drive out good professionals, delirium sets in more easily when the activity is protected by secrecy and out of control.

The main threat is not that everyone is naked in front of intelligence services: this was already the case under Napoleon. It is that this machine, launched like a hammer without master, begins to hit indiscriminately and make us fall, at the world level, in a new version of McCarthyism.

No comments:

Post a Comment