Analogy
What is analogical argument? It will perhaps be best to begin with a straightforward example of an analogy, and of one that is clearly meant as an argument. Here is a familiar one from Henry David Thoreau's essay, "Civil Disobedience": 'The mass of men serve the State thus, not as men mainly, but as machines, with their bodies. They are the standing army, and the militia, jailers, constables, posse comitatus, &c. In most cases there is no free exercise whatever of the judgment or of the moral sense; but they put themselves on a level with wood and earth and stones, and wooden men can perhaps be manufactured that will serve the purpose as well. Such command no more respect that men of straw, or a lump of dirt."(1)
Thoreau, be might say, is "drawing an analogy" between humanoid machines (today we can say: robots) and human beings who serve the state with uncritical obedience. He is saying: just as machines which otherwise resemble human beings would not deserve our respect, so human beings that serve the state with uncritical obedience do not deserve our respect.
This at once brings out the feature of analogical argument that is most distinctive and at the same time most troublesome. Induction moves from less general statements (call them cases) to more general ones (call them principles), and deduction often goes in the other direction, from principle to case. But analogy seems to reason neither upwards, like induction, nor downwards, like deduction, but sideways. It seems to reason from case to case.(2)
Why this is troublesome can be seen by looking at most attempts to represent the form of argument by analogy, such as the one presented by Susan Stebbing. Such arguments, she says, always have the following logical form:
X has the properties p1, p2, p3, ... and f;
Y has the properties p1, p2, p3, ...
Therefore, Y also has the property f.(3)
I do not wish to deny that this does state the logical form of arguments of this sort - on some definitions of "logical form," it may well be that it does. I do want to point out, however, that this schema does not indicate why the two premises constitute any reason for believing the conclusion. It gives us no reason to think that it is the logical form of an argument.
If this is not obvious, let X and Y be two human beings, let the ps be being male, having a mother named Caroline, and having been born before the election of President Truman, and let f be having written a dissertation on moral psychology. In that case, the two premises are no evidence at all (not even poor evidence) that the conclusion is true. At this point, that most natural thing to say is "Of course not: it makes all the difference what the ps and the f are." That is actually my point, or very close to it. For it implies that no collection of properties shared by two cases can, simply as such, constitute evidence that some other property of one case, picked out at random, is also shared. Something more must be involved than the two cases and their shared properties.
There would be something more involved if, contrary to what I suggested when I first described analogical arguments, there really is a principle at work somewhere in the argument. A closer look at the argument from Thoreau suggests that such is indeed the case.
Although he does not state it explicitly, Thoreau does take some pains to show us that there is a principle behind the analogy he draws. He comments that in the most common ways of serving the state there is "no free exercise whatever of the judgment or of the moral sense." This is given as the reason why such conduct does not deserve our respect. Clearly, there is a principle at work here, which would be something equivalent to "Human conduct cannot serve as grounds for respect unless it involves in some way or other the free exercise of judgment or of the moral sense."(4)
Though not stated, this principle does play a role in the argument. The role that it plays, moreover, is not that of unjustified assumption. He describes what he takes to be the extreme opposite of respect-worthy human conduct - "earth," "stones," "wooden men" - in order to make this principle vividly plausible. The idea seems to be that, obviously, "men" made of such materials would not deserve our respect. If we search for an explanation for this obvious fact, we will find that all the sensible ones rest at least in part on the free exercise of judgment and moral sense.
What this means is that, while the unstated principle supports what he says in the second case, it is supported by what he says in the first. But the sort of support involved in the two instances is quite different. The principle supports the second case as a premise of a deductive argument, of which the case is the conclusion. If the free use of these powers constitute the only grounds for respect, then it necessarily follows that uncritical obedience - which by definition involves their suspension - provides no grounds for respect. On the other hand, the principle is supported by the first case as the conclusion of what Peirce called a "retroductive" or "abductive" inference. It is supported because and to the extent that it is a good explanation of what appears to be the truth of the first case.
This, I suggest, is the actual structure analogical arguments: They always consist of two descriptions of less general putative facts (the cases), the first functions as a premise, and the second as a conclusion. The first case supports a relatively general putative truth, the principle, by retroduction, and the principle in turn supports the second case by deduction.
We can now see why, if an argument by analogy is to have logical force, "it makes all the difference what the ps and the f are." The ps in the first case must but such as will support a principle which can in turn support the second case. If the principle does support the second case in the requisite way, then these properties will also appear in the second case, but the brute fact of resemblance between the cases - the mere sharing of ps, however many there are - is irrelevant.(5) The principle, together with its logical relations with the two cases, is what gives the ps their evidentiary force.
Once we see the actual logical structure of analogical arguments, we can see that (just as we may have suspected) they are respectable arguments. Thoreau's argument seems to me, in particular, a reasonably good one. Both parts of it seem sufficiently cogent to be merit a thoughtful response from the people at whom it is aimed. At the same time, we can also see why analogical arguments impress us generally as logically "soft" ones, as arguments that yield probability and not certainty. While it is true that the second part of the argument, on my account, is deductive, and such arguments are paradigms of reasoning that demonstrates its conclusion, the first part is retroductive, and arguments of that sort, even good ones, often yield something that falls far short of certainty.
It is natural to see this as a deficiency of analogical arguments but, supposing it is a deficiency, it is not as severe as one might think. Such arguments can be very strong ones, including in particular, the first, softer, part of it. I think it is very unlikely, for instance, that at whom Thoreau's argument was aimed would attack it by do try to demolish the first part. That is, they probably would not deny that robot-like mechanisms fail to deserve the respect we give to people who are conducting themselves well, nor would they deny that a good explanation for this can be found in the fact that such mechanisms do not use judgment and moral sense. They would be much more likely to attack the deductive part of the argument, by denying the truth its unstated minor premise. That is, they would most likely deny the implied assertions about their own conduct in virtue of which the principle is thought to apply to them. They might claim, for instance, that their obedience and refusal to criticize actually constitute a moral choice and rest on ideals that are lofty and authentically their own. This is probably their most promising avenue of escape. The retroductive part of it is probably not the part most likely to suffer a breach.
The same sort of thing is true of this argument from Mencius, which also seems to me a fairly good one;
"Tai Ying-chih said, `We are unable in the present year to change over to a tax on one in ten and to abolish custom and market duties. What would you think if we were to make some reductions and wait till next year before putting the change fully into effect?'
"'Here is a man,' said Mencius, 'who appropriates one of his neighbor's chickens every day. Someone tells him, "This is not how a gentleman behaves." He answers, "May I reduce it to one chicken every month and wait until next year to stop altogether?"'
"'When one realizes that something is morally wrong, one should stop it as soon as possible. Why wait for next year?'
Here the first case is the description of the chicken thief's conduct and his proposed reform, together with the notion, unstated and taken as obvious, that this proposal is not what he should be doing. The principle is stated by Mencius in his penultimate sentence. It does seem to be a good explanation for the truth of the first case and, consequently, a reasonable conclusion to draw from it. The Mencius does not tell us who Tai Ying-chih was or give us a context for this exchange, but it seems likely that, if he wishes to resist Mencius' conclusion, he probably should not try to refute this part of the argument. He is likely to fare better by resisting the deductive part. He could try to argue, for instance, that the custom and market duties are not morally wrong, only inconvenient, or he might try to show that a mere reduction in them makes some sort of morally qualitative difference, such that it is not morally wrong. Whether such a strategy would succeed or not, the second, deductive part of the argument seems more vulnerable to attack, or closer to being vulnerable, than the first and logically "softer" part.
1. "Civil Disobedience," in Walden and Civil Disobedience, ed. Owen Thomas (New York: Norton, 1966), p. 226.
2. Described in this way, analogical reasoning sounds very much like what Martha Nussbaum calls "fancy," which she defines as consisting in part of the ability to "see one thing as another and to see one thing in another." Martha Nussbaum, Poetic Justice: The Literary Imagination and Public Life (Boston: Beacon Press, 1995), p. 4. Nussbaum's readers might find it illuminating to see my account of analogy as another account, sharply different from hers, of what "fancy" is. I will discuss her views at greater length in Chapter IV.
4. I hope it is obvious that Thoreau is not thinking of the Kantian sort of respect, which is really a respect for the rights of individuals, and is grounded, simply, in the recognition that they are rational agents. What Thoreau has in mind is plainly a sort of admiration, and thus is something that must be earned.
5. Here I am disagreeing with what Stebbing seems to be saying when she asserts that "The force of the argument depends upon the resemblance between X and Y with regard to the p's." Thinking to Some Purpose, p. 113.