|
Article Excerpt
They all laughed at Wilbur and his brother, When they said that man could fly. They told Marconi wireless was a phony; It's the same old cry.
--IRA GERSHWIN
WHY DO WE MAKE ERRORS? ARE THEY BLUNDERS CAUSED BY THE limitations of our cognitive system? Or are errors indispensable parts of every intelligent system? From the first perspective, all errors are at best unnecessary and at worst harmful. Consider an error commonly made by children. When asked to find the sum of 1/2 and 1/3, the answer is often 2/5. This is called the freshman error of adding numerators and adding denominators (Silver, 1986). But blunders are not limited to children. After the invention of the telephone, a group of British experts concluded that this invention had no practical value, at least in their country: "The telephone may be appropriate for our American cousins, but not here, because we have an adequate supply of messenger boys" (Sherden, 1998: 175). In 1961, President John F. Kennedy is reported to have asked himself "How could I have been so stupid?" after realizing how badly he had miscalculated when he approved the Bay of Pigs invasion planned by the CIA (Janis and Mann, 1977: xv). Blunders like these seem to be unnecessary as well as embarrassing, and every intelligent system would work better without them. In this view, to err is not to think.
From the second perspective, there are errors that need to be made--that is, errors that are indispensable and functional. I call these "good" errors. Children are known for good errors. Consider a 3-year-old who uses the phrase "I gived" instead of "I gave." A child cannot know in advance which verbs are irregular; because irregular verbs are rare, the child's best bet is to assume the regular form until proved wrong. The error is "good"--that is, useful--because if the 3-year-old did not try out new forms and occasionally make errors, but instead played it safe and used only those words it had already heard, she would learn a language at a very slow rate. The characteristic of a good error is that a person is better off making the error than not making it--for reaching a goal more quickly, or attaining it at all. In this view, every intelligent system has to make errors. Making no errors would destroy the intelligence of the system. There is a close parallel to Darwinian theory, where random variability and mutation--copying errors--are essential for evolution by natural selection. Not making these errors would eliminate evolution. Trial-and-error learning, at the ontogenetic or evolutionary level, is one source of good errors in an uncertain world.
In this article, I deal with the study of human errors in experimental psychology. The problem that researchers try to resolve is this: How can one infer the laws of cognition--of perception, memory, and thought? One answer is to study the systematic errors people make. At first glance, this program looks like a straightforward extension of Francis Bacon's plan for studying nature's errors, or of Freud's strategy to analyze repressed memories, slips of tongue, and abnormal neurotic behavior. The idea is to catch nature when it does not pay attention--creating strange facts such as blood rain in Bavaria and an Irish girl with several horns growing on her body (Daston and Park, 1998). However, there is an important difference. We can easily see what is wrong with a goat with two heads or a man with obsessive-compulsive hand washing, and understand that it is not to the benefit of the animal or the human. Cognitive errors, however, are not as clear, as we will soon see. Here, one has to define rather than simply observe what an error of judgment is. In this article, I argue:
1. The study of cognitive errors has been dominated by a logical definition of errors. But this narrow norm tends to mistake forms of human intelligence that go beyond logic for stupid blunders, and consequently fails to unravel the laws of mind.
2. An ecological analysis, in place of a logical one, instead reveals the existence of good errors, which open a window into the mind. The prototype of an ecological analysis is the study of visual illusions.
The method I use in this article is to document both points by illustrative examples.
VISUAL ILLUSIONS
Let us first see what a visual illusion is, and what one can learn from it. Consider the dots on the left-hand side of figure 1. They appear concave, receding into the surface away from the observer. The dots on the right side, however, appear convex: they project up from the surface, extending toward the observer. When you turn the page upside-down, the concave dots will turn into convex dots, and vice versa. But there is no third dimension, and there are no convex and concave dots. Seeing things that systematically deviate from the relevant physical measurements is called a perceptual illusion.
[FIGURE 1 OMITTED]
What can we learn from this illusion about how our brain works? First, that the world, from the perspective of our mind, is fundamentally uncertain. Our brain does not have sufficient information to know for certain what is out there, but it is not paralyzed by uncertainty. Second, the brain uses heuristics to make a good bet. Third, the bet is based on the structure of its environment, or what it assumes the structure to be. The brain assumes a three-dimensional world and uses the shaded parts of the dots to guess in what direction of the third dimension they extend. By experimentally varying factors such as the location of the light source and the shading, and documenting their effect on the illusion, Kleffner and Ramachandran (1992) concluded that the assumed ecological structures are that
1. light comes from above (in relation to retinal coordinates), and
2. there is only one source of light.
These structures describe human (and mammalian) history, where the sun and moon were the only sources of light, and only one operated at a time. The first regularity also holds approximately for artificial light today, which is typically placed above us, such as street lamps (although there are exceptions, such as car lights). The brain exploits these assumed structures by using a fast and frugal heuristic: If the shade is in the upper part, then the dots are concave; if the shade is in the lower part, then the dots are convex.
Shading is phylogenetically one of the most primitive cues, and so is the principle of countershading that conceals animals' shapes from predators, as in the pale bellies of swarm fishes that neutralize the effects of the sun shining from above. Helmholtz (1962 [1856-1866]) used the term "unconscious inferences" for this type of heuristic, and he and his followers (e.g., Brunswik, 1934) thought that the cues were learned...
|
