"Proof by analogy is fraud." - BjarneStroustrup
- A is like B.
- A has property P.
- Therefore, B has property P.
. You can't prove anything from an analogy, only hope to improve informal understanding. Abuse of analogies can teach people to hate analogies completely (See: StopUsingMetaphors
). However, sometimes they can be used to good effect. Let's sort this out.
An example analogy:
- Stuff in RAM is like the stuff in your head.
- Stuff on Hard disks is like stuff written down.
It is true that there are many properties of RAM that brains don't have, many properties of hard disks that scratch pads don't have, and vice-versas. Unlike RAM, our brains can compute as well as store. A hard disk can be more easily reorganized than a notebook. Et cetera.
However, this very analogy has helped us explain the difference between RAM and disks to countless users. When your computer turns off, whatever you didn't save to disk goes away -- just like how when you die the stuff you didn't write down goes away. It has been said that we can only learn what is on the EdgeOfOurKnowledge
. When we speak in analogies, we bring things that were very distant from the edge of our audiences knowledge (in this case data persistence) much closer.
The solution seems fairly simple:
- Use analogies where they work, and not where they don't. When they don't work, figure out the essential differences; explain.
Another way to think about it is that the ArgumentByAnalogy
is approximately as valid as the two comparators are alike. The likeness drives the validity. If the argument relies on a property which the things do not really share, then the argument is invalid. Screws are like nails (size, shape, material), but you won't get anywhere very far trying to screw in a nail (threaded/non-threaded). The method of use relies on the threaded/non-threaded property.
Arguments of this type are often seen as using inductive reasoning. Specifically Modus Ponens
Basics for Argument By Analogy [Warning: Pet Peeve to Follow]
For an analogy to be useful, it must be well understood by all parties involved, and should be better understood than the issue under discussion.
Most of the analogies used in software discussions tend to fall short of this standard. Typically, the parties to the discussion have a strong understanding of software, but feel compelled to use analogies to areas that few (if any) of the participants understand. The result is usually the main point is lost and arguments ensue regarding the analogy.
Unless one is very certain that a particular analogy is well understood by all, don't use it.
Also, if one participant disagrees with the applicability of the analogy, then there is little or no point in the other participant(s) persisting with it -- no matter how well understood it is.
I disagree. I think you should NEVER use analogies in certain places, such as forums. The problem, as you mentioned, is all analogies break down at a certain level of comparison (they have at least one weakness). On a forum, there are always people who will advance their assertion by discrediting yours. You prove your point by analogy and instead of defending their point, they attack your analogy in the weak spot(s). There is no analogy ever made that cannot be attacked in a forum, derailed, and be made to discredit the author. It allows people to escape defending their assertions.
This is why you have to be very careful questioning politicians. The second you mention something like 'Iraq is like Vietnam' you allow them 2 choices instead of one: 1) to defend Iraq, or 2) attack the weakpoints of the Vietnam analogy. All lesser intellects (and most young people) will choose 2, since it's easier. Always, the better approach is to ask them to justify their assertions.
You also allow two other choices: 3) to defend Vietnam, or 4) identify a different analogy (such as El Salvador or the Philippines).
The discussion of the second choice and fourth choices (the weaknesses of the analogy, or the differences between two analogies) can be very informative.
I disagree with the above. As stated in the first section, the issue is "whether the issue(s) in question share the important and discussed characteristics". If both parties agree that they share the characteristic relevant to discussion, then there is no use arguing about other unrelated specifics. For example, if both parties agree in the qualities evaluated, then externalities may be ignored. For example, most would agree that there is a relation between a hand and a palm that is similar to a foot and a sole. Applying this to the Iraq-Vietnam analogy, so long as the characteristics being discussed are not in dispute, then so long as the conversation does not deviate from this initial premise (that there is a particular relation) all deductions are valid. However, the analogy maker must be careful not to overstep and draw conclusions not dealing with the relation, or criticisms are appropriate and valid.
Who's the selfish idiot who disagreed by deleting opinions they don't like? Bake in hell or it's Earthly equivalent, you Mental Soviet!
I am. The removed "content" was barely more than gibberish. Neural nets "rely on ArgumentByAnalogy"? Utter hogwash. Do you know anything about neural networks? Please, please, please at least make an
attempt at scientific credibility. Once you've managed that, you can start to work on actually achieving it.
Yes I do know a bit about them. No, I am not an expert, but I do know that they generally are arranged to use ArgumentByAnalogy
, or something very similar, to make "decisions". They have even been used to emulate large-variable-quantity regression because formal regression is often too resource-intensive. If you want more details, use the tool known as "asking", not the Delete key. Deletion is rude. Don't be an Asperger asshole.
I moved the material you deleted to AnalogyProcessing