Connecting the Dots

Tuesday, March 11th, 2003

In Connecting the Dots, Malcolm Gladwell makes the case that intelligence work is all so easy…after the fact:

In the fall of 1973, the Syrian Army began to gather a large number of tanks, artillery batteries, and infantry along its border with Israel. Simultaneously, to the south, the Egyptian Army cancelled all leaves, called up thousands of reservists, and launched a massive military exercise, building roads and preparing anti-aircraft and artillery positions along the Suez Canal. On October 4th, an Israeli aerial reconnaissance mission showed that the Egyptians had moved artillery into offensive positions. That evening, aman, the Israeli military intelligence agency, learned that portions of the Soviet fleet near Port Said and Alexandria had set sail, and that the Soviet government had begun airlifting the families of Soviet advisers out of Cairo and Damascus. Then, at four o’clock in the morning on October 6th, Israel’s director of military intelligence received an urgent telephone call from one of the country’s most trusted intelligence sources. Egypt and Syria, the source said, would attack later that day. Top Israeli officials immediately called a meeting. Was war imminent? The head of aman, Major General Eli Zeira, looked over the evidence and said he didn’t think so. He was wrong. That afternoon, Syria attacked from the east, overwhelming the thin Israeli defenses in the Golan Heights, and Egypt attacked from the south, bombing Israeli positions and sending eight thousand infantry streaming across the Suez. Despite all the warnings of the previous weeks, Israeli officials were caught by surprise. Why couldn’t they connect the dots?

If you start on the afternoon of October 6th and work backward, the trail of clues pointing to an attack seems obvious; you’d have to conclude that something was badly wrong with the Israeli intelligence service. On the other hand, if you start several years before the Yom Kippur War and work forward, re-creating what people in Israeli intelligence knew in the same order that they knew it, a very different picture emerges.

Gladwell shares an amusing (if frightening) anecdote to illustrate the problem of drawing conclusions from ambiguous signals buried in “noise”:

In the early nineteen-seventies, a professor of psychology at Stanford University named David L. Rosenhan gathered together a painter, a graduate student, a pediatrician, a psychiatrist, a housewife, and three psychologists. He told them to check into different psychiatric hospitals under aliases, with the complaint that they had been hearing voices. They were instructed to say that the voices were unfamiliar, and that they heard words like “empty,”"thud,” and “hollow.” Apart from that initial story, the pseudo patients were instructed to answer every question truthfully, to behave as they normally would, and to tell the hospital staff — at every opportunity — that the voices were gone and that they had experienced no further symptoms. The eight subjects were hospitalized, on average, for nineteen days. One was kept for almost two months. Rosenhan wanted to find out if the hospital staffs would ever see through the ruse. They never did.

Rosenhan’s test is, in a way, a classic intelligence problem. Here was a signal (a sane person) buried in a mountain of conflicting and confusing noise (a mental hospital), and the intelligence analysts (the doctors) were asked to connect the dots — and they failed spectacularly. In the course of their hospital stay, the eight pseudo patients were given a total of twenty-one hundred pills. They underwent psychiatric interviews, and sober case summaries documenting their pathologies were written up. They were asked by Rosenhan to take notes documenting how they were treated, and this quickly became part of their supposed pathology. “Patient engaging in writing behavior,” one nurse ominously wrote in her notes. Having been labelled as ill upon admission, they could not shake the diagnosis. “Nervous?” a friendly nurse asked one of the subjects as he paced the halls one day. “No,” he corrected her, to no avail, “bored.”

The solution to this problem seems obvious enough. Doctors and nurses need to be made alert to the possibility that sane people sometimes get admitted to mental hospitals. So Rosenhan went to a research-and-teaching hospital and informed the staff that at some point in the next three months he would once again send over one or more of his pseudo patients. This time, of the hundred and ninety-three patients admitted in the three-month period, forty-one were identified by at least one staff member as being almost certainly sane. Once again, however, they were wrong. Rosenhan hadn’t sent anyone over. In attempting to solve one kind of intelligence problem (overdiagnosis), the hospital simply created another problem (underdiagnosis). This is the second, and perhaps more serious, consequence of creeping determinism: in our zeal to correct what we believe to be the problems of the past, we end up creating new problems for the future.

Leave a Reply