Читать книгу This Is Metaphysics - Kris McDaniel - Страница 17

1.27 One of the most obvious ways in which we classify things is by using language. We make use of general terms, such as “dog,” “cat,” and “table” to group things together under common headings. Once we’ve done this, we can use these terms in sentences that communicate very general facts, such as that dogs typically are furry and friendly. In addition to terms like “dog” and “cat” that seem to stand for kinds of things, we can also classify things by way of predicates like “red” and “circular.”

1.28 But there are some interesting apparent differences: dogs do seem to form a kind of thing in a way that, say, red things do not. Suppose I tell you that I have a red thing in a box: it would be reasonable to ask what kind of thing it is, and if you did ask this question, and I replied by saying that it is a red thing, or a colored thing, you’d probably think that my answer was pretty weird. Suppose I tell you that I have a dog in a box: it would be a little odd to ask what kind of thing is in my box, though it wouldn’t be at all odd to ask what kind of dog is in my box. (You should probably also ask why I put a poor dog in a box, but since this is a book about metaphysics rather than ethics, we won’t pursue that question further.) We use words like “dog,” “cat,” “car,” “desk,” “proton,” and “human being” to stand for kinds of things, and we use words like “red,” “tall,” “solid,” “fast,” and “furry” to stand for features of things. There might be an important metaphysical difference between kinds of things and features of things. But there is also an important similarity that is relevant to our discussion of classification: we use both of these types of expressions to mark differences in things: these things are the things that are red, rather than those things that are some other color, such as green; these are things that are dogs rather than those things that are some other kind of animal.

1.29 Reflecting on the various ways we use language to classify objects can open us up to new philosophical problems. Here is an interesting thought experiment suggested by Eli Hirsch.⁹ Suppose we were to encounter a community of human beings who speak a language that is in many respects like English but with some significant differences. First of all, these people don’t have words for colors like “red,” “purple,” “green,” and so forth in their language, and they don’t even have any words that are synonymous with those words. They also don’t have words for shapes, such as “circular,” “triangular,” and so forth. Although these kinds of words are missing from their language, this is not because the people in this community differ from us in what they can perceive with their eyes. (I suppose that if human beings had never evolved with the ability to see things, we wouldn’t have had words for colors either.) Their language does contain some words that English doesn’t contain. For example, it contains the word “gricular,” which stands for a feature that things have when they are either green or circular, and it contains the word “grincular,” which stands for a feature that things have when they are either green or not circular, and it contains the word “ngricular,” which stands for a feature that things have when they are either not green or are circular. Let’s call this language “the gruesome language.”

1.30 The gruesome language is peculiar. I’ll say a few more things about the gruesome language that will probably amplify that impression. There are some words in English that can be explicitly defined. A stock example is the word “bachelor,” which can be explicitly defined as “someone who is an unmarried, eligible to be married, adult human male.” Competent users of words like “bachelor,” which is a word that is capable of being explicitly defined, should be able to at least gesture at its definition. If a person whose first language is not English asks you what “bachelor” means, it would be appropriate to provide her with the abovementioned definition. But not every word in English is capable of being explicitly defined, even by fully competent speakers of English. When you give an explicit definition of a word, obviously and unavoidably you do this by using other words. Some of these words might themselves be capable of being explicitly defined, but at some point the chain of definitions has to come to an end. There must be some words in English that aren’t capable of being explicitly defined, even by competent users of English. I suspect that words like “red” and “green” are good examples of words that we can’t explicitly define. Suppose a person whose first language is not English were to ask you what “green” meant. You could do the following things to help her out. You could show her samples of things that are green, such as American dollar bills, or the color of grass or leaves in late spring, or if you are especially fortunate, some lovely emeralds. In other words, you would help this person learn what “green” meant by pointing at things that happen to be green. That’s a very different activity from giving an explicit definition of the word “green.” I suppose that you could also say that green is the color that things have when they reflect light with a wavelength of around 510 nanometers, but you shouldn’t think that by doing this you are giving a definition of “green.” Definitions of words are the sorts of things that lexicographers discover, but that green things reflect light of a wavelength in the neighborhood of 510 nanometers is a discovery of physics. Perfectly competent speakers of English hundreds of years ago knew what “green” meant but most of them knew next to nothing about the physics of light. (That’s true of a lot of competent speakers of English today too.) Competent speakers of English can’t give an explicit definition of “green.”

1.31 The speakers of the gruesome language are in a similar position with respect to the words “gricular,” “ngricular,” and “grincular.” If you asked them what they meant by “gricular,” they would have to resort to gesturing at things that are gricular and hope that we get the general idea. A few of the more scientifically literate speakers of the gruesome language might tell us that gricular things are those things that either reflect light of a wavelength of around 510 nanometers or are shaped so that the points on its boundary are all some fixed distance from a center point and from a continuous figure. But they wouldn’t be giving you a definition of “gricular.” What they would be doing instead is similar to what we do when we say what wavelength of light green reflects. Gricular is not a word in their language that can be explicitly defined in terms of other words in their language, not even by speakers fully competent in the gruesome language.

1.32 Speakers of the gruesome language classify things in a way that seems odd to us. Just consider all of the things that are grincular. Is this a list of things that objectively belong together? If nothing objectively belongs together with anything else, then their way of classifying objects is not objectively better or worse in this way, although there might be practical reasons to prefer continuing to classify objects in the way that we do instead of changing wholesale to their systems of classification.

1.33 I am going to begin to develop an argument for the conclusion that some things objectively belong together. This argument turns on the idea that certain words are projectable while others aren’t, and then goes on to claim that a classification system is a bad one if the words it uses to classify objects aren’t projectable.

1.34 The key technical term to be explained here is “projectible.” The very rough idea is that a projectable term is one that you can justifiably use to describe things that you don’t directly perceive. Let’s work our way up to a definition of this key technical term. We frequently make generalizations on the basis of a limited number of observations, and if it is irrational for us to do this, the kinds of generalizations we want to make in the course of doing science are irrational as well. Here are some examples. It’s snowing outside and you don’t want to be cold so you put on a heavy coat rather than just go outside in a t‐shirt. Why do you do this? Because you know that thicker clothing is generally warmer than thinner clothing. This is a generalization that you believe, but it is one that you believe on the basis of sampling a relatively small collection of clothing. There are probably millions of heavy coats in the world, and probably millions of t‐shirts as well. You haven’t tried on each heavy coat. And you haven’t tried on each t‐shirt. Does this shake your confidence that, in general, heavy coats are warmer than t‐shirts? I doubt that it does—although you haven’t sampled every coat and t‐shirt, you have tested enough to be justified in believing the general claim that heavy coats are warmer than t‐shirts.

1.35 Here’s a second example. You notice that whenever you drop something heavy on your bare foot, like a laptop computer, it hurts quite a lot. You also notice that whenever you drop lighter things on your bare foot, such as a shoe or a pencil, it hurts a lot less. On the basis of these observations, you conclude that heavier things falling on bare feet tend to hurt more than lighter things falling on bare feet. (You should probably also conclude that you need to wear shoes more often.) You definitely believe this conclusion: were you forced to choose between dropping a feather on your foot or a piano, you know which one you’d choose. But you haven’t observed every possible object that could be dropped on your foot in order to assess how painful that experience would be. You are prepared to form a very general belief on the basis of a very limited number of observations, and you take yourself to be perfectly rational in doing this.

1.36 The general pattern of reasoning we seem to be using in these situations is the following. Whenever we encounter a sufficiently large sample of things that each have a certain feature F, and each member of this sample also has the feature G, we generalize and believe that all things that have F also have G. Philosophers call this kind of reasoning “inductive generalization,” and the method of deriving conclusions in this way, “induction.” It’s hard to see how we can get by in the world if we aren’t justified in believing the conclusions of inductive generalizations. How do I know that this bread will nourish me rather than cause me to explode? Because every time I have sampled bread in the past, it has nourished me rather than caused explosions. My sample size is limited but I nonetheless conclude that I’ll be ok next time I have a sandwich.

1.37 So far, so good. But now we are ready to consider the following puzzle, due originally to a philosopher by the name of Nelson Goodman, called “the New Riddle of Induction.”¹⁰ Once we see how the puzzle works, we’ll be able to define “projectable.”

1.38 Here we go. Let’s think about emeralds. Emeralds have lots of features. Some emeralds are pretty. Some emeralds are covered in dirt, and some emeralds will probably never be seen by anyone, at least not during our lifetimes. Some emeralds have been visually inspected by us, or at least will be, during our lifetime. Let’s say that something is grue if and only if it is either both inspected by us during our lifetime and is green, or it is blue but not inspected by us during our lifetimes. (As you can probably guess, “grue” is another word in the gruesome language.) Because the idea of grue is a little complicated, it’s good to read that definition a couple of times, and then work through some examples. All the emeralds in a jewelry store were visibly inspected by someone during your life and they are green—and so they are also grue.

1.39 Here’s the new riddle of induction. We haven’t seen absolutely all of the emeralds in the world. Yet we have seen a large sample of them. And all of the ones that we have seen have been green. So now we are ready to make an inductive generalization: since we have seen a sufficiently large sample of emeralds, and all of these emeralds are green, we infer that all emeralds are green. However, notice this. All of the emeralds we have seen are also grue. So, we’ve observed the exact same number of grue emeralds as we have green emeralds. So, we should also be willing to make a second inductive generalization: given that all of the emeralds that we have seen are grue, and we have seen a heck of a lot of emeralds, we should be willing to infer that all emeralds are grue too. So, as the result of inspecting the same sample of emeralds, we conclude both that all emeralds are green and that all emeralds are grue.

1.40 The problem is that the claim that all emeralds are green is incompatible with the claim that all emeralds are grue, since there are emeralds that we have never seen and that won’t be observed during our lifetime. Consider an unobserved emerald, which I will call “Eddy.” If all emeralds are green and all emeralds are grue, Eddy is both green and grue. But no one has or will observe Eddy in our lifetime. So, Eddy can’t be green and observed in our lifetime. Since Eddy is grue, and not green and observed in our lifetime, it follows that Eddy must be blue and not observed in our lifetime. So, Eddy is blue. But just a moment ago we said Eddy is green. Nothing can be both blue and green at the same time. So, the same sample set and the same method of forming more general beliefs from that sample set led to inconsistent results. Not good. That’s the puzzle of the new riddle of induction.

1.41 What seems plausible then is that we shouldn’t indiscriminately make inductive inferences. It is ok to infer from our sample size that all emeralds are green. It is not ok to infer from that same sample that all emeralds are grue, even though every emerald in that sample is both green and grue. Don’t believe that whenever we encounter a sufficiently large sample of things that each have a certain feature F, and everything that has F also has another feature G, we should generalize and so believe that all things that have F also have G. Instead, consider something more complex and careful. Only some of those Fs and Gs are ones that license inductive inferences. We are finally ready to define “projectible”: an expression is projectable if and only if it stands for a feature that we can justifiably make inductive inferences about. “Green” is plausibly a projectable expression, but if it is, then “grue” can’t also be projectable.

1.42 But what makes an expression a projectible expression? One answer to this question is that an expression is projectible to the extent that the expression corresponds to objects that objectively belong together. If this is the correct answer, we also have an argument that some objects objectively belong together. The argument is this. We are sometimes justified in making inductive inferences. But we are sometimes justified in making inductive inferences only if some of the words we use to make those inferences are projectible. And those words are projectible only if they classify objects that objectively belong together. So, some objects objectively belong together.