There is an underlying point about the crudeness of using social-scientific methods to predict the conduct of individuals. Even if we had statistics that controlled for the myriad situational factors, different people could not be expected to act the same way in the same situation. This is true in principle if determinism is false. If determinism is true, the causal explanation of conduct will have to be given in terms of neurobiological descriptions of particular brains. This is almost unfathomably remote from social science, and even from most of biology. It is also inapplicable to our actual dealings with one another, unless we devise a way to go around filming people’s brains with real-time neural video cameras.
The relevance of non-situational statistical likelihoods is more debatable if we are deciding how to proceed in a very large number of cases. A police force might practice various kinds of profiling if its sole concern were to stop as much crime as possible, because even tiny differences between the likelihoods of different types of people of committing or having committed a crime could result in significantly different aggregate outcomes when multiplied by thousands or millions of individuals. To this we should reply that stopping as much crime as possible ought not to be our sole concern. The same principle that leads us to think it better that a guilty person should go free than that an innocent person should be found guilty will lead us to eschew profiling.
The second approach to the dark-alley thought experiment is to think of it as an interaction between persons rather than an event type to be studied empirically. This corresponds to what we said above about understanding the meaning of a symbol, with “action” substituted for “symbol” and “purpose” or “intent” for “meaning.” If someone points a gun at you and says, “Your wallet or your life,” you need not reflect on the statistical likelihood that he is being serious rather than playing a practical joke, or that he will shoot you if you refuse the demand, or that by the word “wallet” he refers to your wallet. This would be contrived and absurd. If God were a social scientist (God forfend), He would possess statistics on these things, but we would have no need of them. We already understand the mind of the mugger, and any ambiguity could not be resolved by consulting other cases.
Statistics play a part in establishing the dark-alley fear, but it is small and informal: Your awareness that there is a lot of crime in the neighborhood makes you worry about being a crime victim. For reasons similar to those discussed above, a calculation of the likelihood of crime here and now would yield a very low probability and depend on a great number of simplifying assumptions (though not nearly so many as are involved in attempts to predict the conduct of persons). But really you are not thinking statistically about the likelihood of crime here and now. Rather, as you walk down the dark alley, you say to yourself, “My, what a fine place to mug someone . . .” This way of thinking is fundamentally imaginative and interpersonal, not empirical and quantitative.
A frisson of fear may then run through you if you hear footsteps, but you feel it before you know whose footsteps they are. This “before” is important. It implies that the identity of the individual has only a negative function: It does not establish your fear — the dark alley does that — and can only diminish it. (If you see the person before hearing his footsteps, your frisson is due simply to seeing someone rather than to the particulars of his identity. These things are conceptually distinct even if there is no perceptible delay between seeing someone and noticing the particulars.)