Further, even a fully determinate symbol does not necessarily imply anything determinate about an individual’s likely conduct. The India-ink tattoo is an excellent example of this. What it means is “I have been to reformatory.” What it connotes is “I am not to be trifled with.” And what that means is anyone’s guess. Caution in dealing with such a person is understandable, but we should remember that his character cannot be reduced to, nor his likely conduct inferred from, his history.
There is, finally, a lesson here for the individual who wishes to avoid being stereotyped: If you want others to see you as sui generis, you must first see yourself that way. An extreme individualist would avoid the realm of sociocultural connotation as much as possible, and might therefore end up looking quite conventional.
Although it seems to me that some of Mr. Dalrymple’s conclusions depend on attributing a false determinacy to ambiguous symbols, I think his essay is an important and humane corrective to the misuse of statistics as a guide to interpersonal dealings. A word in this connection about everybody’s favorite thought experiment, “Whom would you rather meet in a dark alley?”There are two ways of approaching that question. The first is to make statistical arguments about what a certain type of person is likely to do. This amounts to applying inductive methods to human beings and studying them as pieces of mechanism. If the resulting conclusions are supposed to guide our conduct, it is the actual likelihood of some outcome that matters, not the mere existence of a statistically significant correlation. There are all kinds of statistically significant correlations we ignore because the likelihoods associated with them in any particular case are too low to worry about. For example, there is a statistically significant correlation between rock-climbing and falling to one’s death, but the likelihood of falling to one’s death is low enough that rock climbers practice their sport regardless. This is not irrational. What would be irrational is to live your life in a state of paranoia about even very small likelihoods.
From statistics about crime by demographic group, how are we supposed to calculate a likelihood that that guy will mug us here and now? Our statistics are about crime in large populations over long periods of time, so we’ll have to do something like this: First we calculate the (low) likelihood that a hypothetical average person of that guy’s profile will commit a certain type of crime over some long period of time; then we choose some very small unit of time to serve as a rough approximation of now and calculate the number of units in the long period; and then we divide the likelihood by the total number of units (thus breaking our initial continuous distribution into a very large number of shorter continuous distributions). Here drops out of the picture, since our statistics are not sorted by type of location, dark alley vs. busy street vs. grassy park. The final likelihood will be minuscule, and completely unilluminating. That is because our calculation has abstracted away from the myriad situational factors — themselves unrelated to demographic profile — that we would have to take into account in order to say with confidence that someone was likely to commit a crime. Statistical methods are not fine-grained enough to capture these things.
If there is a statistically significant correlation between demographic group and crime, we could adopt a vague principle of “extra caution” in our dealings with members of the group, but we must admit that we have no idea how likely even the hypothetical average member is to commit a crime in some particular situation, let alone the actual human being who stands before us. Nor can we say how much likelier he is than some other type of person to commit a crime in that situation. We have simply translated the demographic correlation into an abstract character trait such as “prone to violence,” predicated the trait of everyone in the group, and let our imagination run wild about what someone “prone to violence” might do. There is nothing scientific about this. It is not statistical common sense. It is a pretext for prejudice.