Wednesday, July 22, 2009

False Analogy

The Wikipedia entry on false analogy is a bit odd. The following are given as examples of false analogies:

Love is like a spring shower. It brings refreshment to a person's body.
Children are like dogs. They need to be strongly disciplined and housebroken.
You are the star of my life! You bring light and joy to my existence.
This product will make you feel like a kid again!

But not a single one of these is a false analogy; they are all just comparisons. And if we used the method that the article uses to identify them as false analogies, we would not have any true analogies. Analogies do not require flawless similarity in order to work; the fact that analogues are also dissimilar is not a reason for regarding the analogy as false if the analogues are nonetheless similar in the relevant way, the way that matters. Very oddly, the article also goes on to recognize this. One of the citations is of the false analogy page at, which also exhibits the confusions, since it gives the following examples:

People are like dogs. They respond best to clear discipline.
This soap is like a dream. It lifts you up to a spiritual plane.
A school is not so different from a business. It needs a clear competitive strategy that will lead to profitable growth.

Again, not a one is a false analogy, even if one regards the statement as false, because each of these is a comparison plus a reason why one should make the comparison, which is exactly the opposite of an analogical inference: the inferences here, if we even take them as inferences rather than as a figurative and a literal way of saying the same thing, are all non-analogical inferences to a similarity. To have a fallacy of false analogy you have to have an actual inference, and the inference has to be analogical.

Part of the confusion is perhaps that 'analogy' is equivocal; it can just mean a similarity or a comparison describing one, or it can mean an analogical inference. Only the latter can be fallacious; a comparison may be a bad comparison, but to have a fallacy you have to have an argument. So by 'false analogy' you might mean that the comparison is false, which is not a fallacy, or that the inference itself is faulty (and 'faulty analogy' is a commonly used synonym for 'false analogy'). The term is also a bit unfortunate in that analogical inferences can be false and yet good inferences, because analogical inferences do not yield necessary conclusions but simply probable ones.

Which actually raises the question of whether there really is a false analogy fallacy at all, i.e., a form of analogical inference that is a fallacy in the true and proper sense. You can have badly formed comparisons, but that's not a problem with the inference itself, and the same is true if the grounds for the original comparison is false; your inference can get you the wrong conclusion, but that's not surprising because analogical inferences are defeasible. The mere fact that you started with a false assumption doesn't make the inference bad; and when we are dealing with defeasible inferences, the mere fact that your conclusion happens to be false doesn't, either. We'd need a real fallacy: a form of inference that is itself unreasonable to make.

But as analogical inferences really make use of two similarities, and is an inference from one to the other, one can certainly have an inferential failure if the original similarity is not (in context) relevant to the target similarity, which is to say, the original similarity gives no reason in context to conclude the target similarity. So in many contexts this would be problematic:

Person A's name is Ed and Person B's name is Ed, and Person A is tall; so Person B must be tall.

And the reason is that name is not under most circumstances plausibly considered to be relevant to height. So it would be tempted to think that this could be considered an example of false analogy (and it is certainly better than the above examples), although I don't think this is actually right, for reasons I'll get to in a moment.
The Logical Fallacies website
thus seems to be on the right track, since it gives these examples:

Employees are like nails. Just as nails must be hit in the head in order to make them work, so must employees.
Government is like business, so just as business must be sensitive primarily to the bottom line, so also must government.

I don't think the second example would ordinarily be an example of false analogy, even though the conclusion is false; but, however they are alike, hitting a nail on the head and hitting an employee on the head are not actually similar. I think the first example is actually very good, because I would suggest to you that it is the first real case of a false analogy. And the reason is that there is no actual analogy here. It is set up as an analogy, but it equivocates on the word 'head'. Employee heads and nail heads are both heads in only an equivocal sense, so the analogical inference is actually equivocating -- which is why it reads like a joke, since it's the sort of equivocation we'd typically use in order to make a joke. It's not mere falsity. It is irrelevance of a sort; but it's the sort arising from equivocation.

Are all fallacies of false analogy forms of equivocation? When we eliminate analogies that are merely based on bad comparisons (which is not a fallacy but just false premises), and also analogies that merely have a false conclusion (which is not a fallacy but an unsurprising result given that analogical inferences are nonmonotonic and defeasible), that narrows the field of plausible cases considerably. By 'plausible case' I mean one that has all the desiderata, i.e., it is an analogical inference that is a fallacy in the strict and proper sense, which is to say, is unreasonable precisely as an inference, and not merely because the premises or conclusions simply happen to be false.

There are those cases, like the name-tall case, that seem to be bad in a nonequivocating way. The IEP article on fallacies gives another example:

The book Investing for Dummies really helped me understand my finances better. The book Chess for Dummies was written by the same author, was published by the same press, and costs about the same amount. So, this chess book would probably help me understand my finances.

There's obviously a sort of irrelevance here. But is it the inference itself or the assumption that author, publisher, and cost are either causes, effects, or probable concomitants of content? If the latter, it's not a true fallacy; it's a regular analogical inference with a false assumption about books, just as the name-tall case would be a regular analogical inference with a false assumption about names. After all, analogical arguments are evidential arguments, and both these cases seem to go wrong so badly because they show a failure to understand what's good evidence for what. And that's a bad thing, but it's not a fallacy, any more than

If dogs are canines, the categorical imperative is false
Dogs are canines
Therefore the categorical imperative is false

is a fallacious inference (if the conditional is taken as it usually would be in ordinary language, and if the argument were given in an ordinary context), rather than just a nonfallacious inference based on a bad idea of what counts as evidence against the categorical imperative.

So what do you think? What's the best way to handle it?