The impartial intelligence analyst

The Los Angeles Times describes the efforts of US intelligence analysts to come to terms with politics. Or rather to remain free of it. In particular the article follows Thomas Fingar, who was both designated to lead the overall intelligence effort and is the author of the NIE which estimates that Iran has stopped building a nuclear weapon.

One of Fingar's explicit goals was to free analysis from politics. To achieve this he has created explicit reminders the staff to remain so. "Some of Fingar's first moves were scripted in the legislation that created his job. The law called for basic standards, so analysts now wear cards around their necks reminding them to remain 'independent of political considerations.'"

As an illustration of the new policy to remain free of politics, the LAT provides this puzzling example.

Dissent was encouraged. Attempts to goad students into policy debates were rebuffed. As one young analyst went through the mock exercise of briefing a general who was considering an invasion, she offered a pointed warning. "Once you go into a country and take it over," she said, "it would be best to have a plan."

Based upon what, one wonders? Policy presumably. But there we go again.

Fingar's other initiative was to educate analysts in the correct style of thinking. In order to achieve this the new chief analyst is molding an entire generation of intelligence analysts; building the foundations as it were, of how government will see things. The demographics are impressive.

Nearly half the nation's analysts have joined the government since 2001. To speed their development, Fingar required new hires to take a six-week course called Analysis 101."

In order to impress readers with the scientific bent of the new broom which is sweeping the dusty cobwebs out of the musty old rooms, the LAT writes: "Fingar's team assembled a directory of analysts, the first time that had ever been done. They launched classified versions of the Wikipedia and MySpace websites, so analysts from different agencies could collaborate online."

But the use of social networking tools and Wikis does not in and of itself free analysts from politics. Social networking software like MySpace can in fact be amplifiers of politics. They can create a hermetic universe of group thinkers much faster than manual methods. And while Wikis can place a set of facts in plain view, they do not in and of themselves resolve ambiguities. That's why there are "wars" in Wikipedia and locked threads. One of the more interesting questions that might be asked about Fingar's new approach is whether it avoids politics at all. Very often the goal of 'remaining free from politics' is simply a code word for a project to engage in a rival kind of politics.

"What I liked in him was his analytical style," said Richard Clarke, who was one of Fingar's first bosses before becoming a counter-terrorism advisor to Presidents Clinton and Bush. "He was more open, honest and user-friendly than the intentionally obtuse analysts we sometimes get." ...

The key problem Fingar faces in disassociating analysis from politics surfaced during a recent discussion I had among knowledge management developers who were trying to build a better version of the Yellow Pages. Imagine that you are an immigrant with limited English skills whose washing machine had just gone berserk and flooded the basement. You turn to the Yellow Pages and look under "flood", "basement" or "washing machine" and find nothing that is obviously helpful. You get emergency services, building contractors or appliance stores.

What you really want is a repairman and a plumber. If the immigrant knew the correct category to consult he would find hundreds of tradesmen in his area. But how does he find the right category? The entire Yellow Pages business model is designed around categories, advertising space is even managed by the category. The problem is that a knowledge store founded upon categories is most useful when you already know what the categories mean.

In order to avoid this problem and free users from the tyranny of categories, some developers wanted to work towards creating a network of links among the different ontologies so that the hapless hypothetical immigrant might gradually be led from the idea of flooding to basement flooding and eventually to the right providers. The key problem lay in defining the links between one set of objects and another. The problem was, in practice, of finding ways to progressively constrain the weights among the multiplicity of possible links in order to converge on the right set of associations.

One developer proposed to solve the problem by heading a wiki effort to construct links of appropriate weight between ontologies. In other words, coding a trail of breadcrumbs from one point to another. Others proposed implementing a neural learning network which would track the paths that people traced through Yellow Pages categories in order to empirically learn where, when faced with a recalcitrant washing machine, most sufferers finished up.

But whether deductively or inductively semantics implicitly reared its ugly head. The facts had to be put into a context to make sense. It was never enough to merely observe that a washing machine had flooded a given basement. It order to get to the next step it was vital to establish what the homeowner's utility function was. Depending on the homeowner's preferences, the next call could be to the movers, the demolition man, the travel agent or to the plumber.

In such a universe, what does it mean when an analyst is obliged to wear a card around your neck saying 'I will not think of policy'? And what happens when intelligence estimates, such as Fingar's NIE can produce a political reaction from Iran? The LAT itself hints that Fingar's NIE may have set up a self unfulfilling prophecy. By downplaying the urgency of Iran's nuclear threat, the NIE could have itself weakened UN sanctions against Teheran, making it more likely for Iran to restart what was "stopped".

Even those who defended the report's findings faulted the way it was put together. Fingar's boss, Director of National Intelligence J. Michael McConnell, testified in February that the report had caused such confusion that if he could rewrite it, he would "do some things differently." ...

Fingar said the Iran intelligence report emphasized the halt in warhead work because that was the newest finding. He attributes the attacks to anger among hard-liners that the report didn't conform to their preconceived views.

"The unhappiness with the finding -- namely that the evil Iranians might be susceptible to diplomacy -- adroitly turned into an ad hominem assault," Fingar said. "Why do we have an intelligence community if all you want are cheerleaders?"

The lasting impact of the report on Iran policy has been unclear. Weeks after its release, the U.N. approved new sanctions against Tehran, but they fell far short of what the Bush administration wanted.

Intelligence and policy may prove themselves to be inextricably linked for as long as intelligence influences what policy is made. There may in fact be no such thing as a value-free judgment especially in ambiguous situations, where the value itself may be in the judgment.

The Belmont Club is supported largely by donations from its readers.