By Sheldon Greaves
There's a book that I think should be on the reading list of just about anyone who is involved in intellectual work of any kind, particularly if it concerns divining information from recalcitrant and multi-variant data. Years ago, a master CIA analyst named Richards J. Heuer wrote a textbook for analysts called The Psychology of Intelligence Analysis. I decided to put it on my nightstand for my bedtime reading recently. I'm glad I did; nightly I get reminded why this is such a good book for anyone who must reason their way though problems or uncover facts.
Heuer's book is designed to identify the cultural, educational, psychological and, in some cases, the physiological limitations and barriers that prevent intelligence analysts from working effectively. He discusses how we learn, how we remember, why we use certain bits of information but not others, how much information is too much, or not enough, and knowing the difference. There are some interesting exercises for identifying biases and overcoming them. The material on creativity and brainstorming is a bit dated (this book was written in 1999), but otherwise it's quite good. Whatever your intellectual work, there is a lot of good advice and practical information that can make you a stronger, more effective thinker.
By the way, it's available as a free .pdf download from the Center for the Study of Intelligence on the CIA's web site. It's worth getting. Your tax dollars at work.
Among other things, Heuer has an excellent section on the use of hypotheses that explains their use more clearly than many other attempts I've seen. Most of us learn in school that the scientific method demands we come up with a hypothesis, then we design an experiment to test it, and then refine the hypothesis based on the results of the experiment. What gets obscured in this explanation is that often the student comes away thinking that the idea is to prove a hypothesis, when in fact the goals is to disprove hypotheses until there is (ideally) just one left. That's an important distinction, one that makes the whole hypothesis thing actually useful. Allow me to quote from Heuer at length, in which he describes a psychological experiment based on this idea:
The experimental design was based on the... point that the validity of a hypothesis can only be tested by seeking to disprove it rather than confirm it. Test subjects were given the three-number sequence, 2 - 4 - 6, and asked to discover the rule employed to generate this sequence. In order to do so, they were permitted to generate three-number sequences of their own and to ask the experimenter whether these conform to the rule. They were encouraged to generate and ask about as many sequences as they wished and were instructed to stop only when they believed they had discovered the rule.
There are, of course, many possible rules that might account for the sequence 2 - 4 - 6. The test subjects formulated tentative hypotheses such as any ascending sequence of even numbers, or any sequence separated by two digits. As expected, the test subjects generally took the incorrect approach of trying to confirm rather than eliminate such hypotheses. To test the hypothesis that the rule was any ascending sequence of even numbers, for example, they might ask if the sequence 8 - 10 - 14 con- forms to the rule.
Readers who have followed the reasoning to this point will recog- nize that this hypothesis can never be proved by enumerating examples of ascending sequences of even numbers that are found to conform to the sought-for rule. One can only disprove the hypothesis by citing an as- cending sequence of odd numbers and learning that this, too, conforms to the rule.
The correct rule was any three ascending numbers, either odd or even. Because of their strategy of seeking confirming evidence, only six of the 29 test subjects in Wasonâ€™s experiment were correct the first time they thought they had discovered the rule. When this same experiment was repeated by a different researcher for a somewhat different purpose, none of the 51 test subjects had the right answer the first time they thought they had discovered the rule.
In the Wason experiment, the strategy of seeking confirming rather than disconfirming evidence was particularly misleading because the 2 - 4 - 6 sequence is consistent with such a large number of hypotheses. It was easy for test subjects to obtain confirmatory evidence for almost any hypothesis they tried to confirm.
So how might one have uncovered the hidden rule? Here are some hypotheses one might test:
All numbers in the sequence are even
The sequence is always ascending
The numbers are always spaced at regular intervals
The sequence is the result of n times 2
Coming up with a sequence to disprove each hypothesis is easier than finding one to "prove" it. So for the hypothesis, "All numbers in a sequence are even", the sequence 6 - 4 - 2 would have "failed" even though it conforms to the hypothesis. That failure would eliminate it from the pool, narrowing the possible set of rules.
The result of the experiment Heuer describes was that very, very few of those tested came up with the correct rule because they kept trying to prove rather than disprove the hypothesis. It's an important tool to keep in one's head. Learn to use it.
Finally, I will close this post with a story of my student days at Berkeley when i was taking a class in Advanced Biblical Hebrew. Our Professor was Rabbi Jacob Milgrom, one of the true giants in the study of ancient Israelite religion. Some of us were discussing another scholar, Yehezkel Kaufman, who had written a book, The Religion of Israel, trying to show that ancient Israelite religion was never influenced by the religions of the surrounding Canaanite nations. After the book was published, it's thesis was gradually demolished. So as we were tut-tutting the fate of poor Kaufman's life's work, Rabbi Milgrom interrupted us and Â told us that Kaufman has in fact made a very important contribution. When we asked what it was, he replied, "Yes, Kaufman's thesis was wrong, and was shown to be wrong. But he made the argument as well as it could possibly have been made, and for that reason, the question is now settled. That's a very significant contribution."
It was a fine example of the value of a failed hypothesis.