Friday, April 27, 2012

Apples and oranges (and cherries)

"You are comparing apples with oranges". How many times have you heard that argument. Is it valid? Depends. Clearly apples and oranges are different species, and insofar as they are different, it is wrong to make a comparison. Yet they are of the same genus. If the objection is about a fruit-related claim, without regard to species of fruit, it is invalid. For example, you claim that all fruit is completely safe to eat. I object that cherries contain hydrogen cyanide, which is not completely safe to eat. You reply "You are comparing apples with oranges (or perhaps cherries)". Invalid. If cherries are fruit, and if cherries are not completely safe to eat, then it is false that all fruit is completely safe to eat. It is no good claiming that I am comparing apples with oranges, if I am talking about fruit.

The fallacy is obvious in this case, but it is frequently abused, and there some wonderful examples on Jimmy Wales talk page here, in a discussion about the 'paid editing' problem. Wikipedians don't like paid editing, because they see it as a 'conflict of interest'. If you are paid, your interest is to write an article satisfying the interests of the person making the payment, not the interests of a comprehensive, reliable and neutral reference work. That is bad, so paid editing is bad. Now someone has (rightly, in my view) objected that the same principle applies to any case where there is a conflict of interest.
I really, truly don’t understand why someone who gets $1000 to edit an article is more inclined to violate our policies (WP:NPOV, etc.) someone who is a “true fan” committed to showing the world how great their favorite team/singer/restaurant, or someone who is absolutely certain that Ethnic group X is better than Ethnic group Y. I don’t understand why we want or need a special policy to deal with one specific form of motivation, when we have a perfectly good set of policies that govern all forms of biased and improper editing behavior.
Wales replies. "You are positing a competition between two things that are completely different. Apples and oranges." Fallacy. Of course paid editing is different from "favourite band" editing. One is motivated by money, the other is motivated by a fanatical obsession with some stupid 'band'. These are different species of editing, just as apples and oranges are different species of fruit. But they are the same genus for all that. The genus is 'conflicted editing', and one case of conflicted editing is not 'completely different', qua genus than any other case. If it is wrong in the one case, why is it not wrong in the other, at least if the wrongness flows from the genus, rather then the species.

Wales goes on: "the incentives before them and the motivations are different". Certainly, as species of editing. But the question is whether the different incentives and motivations involve a conflict or not, and surely they do. Then he says that one size does not fit all, which is another, frequently encountered, version of the fallacy. Of course one size does not fit all, if the cases are different sizes. But the original claim may have had nothing to do with 'size'. Another editor objects that it is 'self evident' that money is special, and that it is "ridiculous to posit otherwise". How so? It is self evident that oranges are special qua species, and not the same as cherries, apples, bananas etc. But they are all the same insofar as they are fruit. The question is whether the different cases (favourite band editing, nationalistic editing) involve a conflict of interest, and surely they all do. Payment is just one instance of a general, problematic case.

Later on the real reason comes out: unpaid editors would be demoralised if paid editing. That is a valid objection: conflicted editing that is demoralising is specifically different from conflicted editing that is not. But that had to be spelled out.


Anthony said...

A real terms of use is finally out (to become effective May 25). I think it's interesting how they incorporated the rules of the arbitration committee while maintaining the position that arb com is not acting as their agent:

"You agree to comply with the final decisions of dispute resolution bodies that are established by the community for the specific Project editions (such as arbitration committees); these decisions may include sanctions as set out by the policy of the specific Project edition."

Anthony said...

More on topic, a similar fallacious argument arises when people confuse examples for analogies.


A: Oranges are completely safe to eat. After all, they are fruits, and all fruits are completely safe to eat.
B: Cherries are fruits, and they contain hydrogen cyanide, so they aren't completely safe to eat.
A: Yes but that is a bad analogy. Oranges are not like cherries, they don't contain hydrogen cyanide.
B: I wasn't making an analogy, I was giving an example.

Anthony said...

Also somewhat more on topic, I think the argument is that paid editing is a simple, clear, objective criterion which, it is claimed, necessarily shows a conflict of interest.

I think the argument fails because the interest of the person making the payment is not necessarily in conflict with the interest of a comprehensive, reliable and neutral reference work.

That said, it is a complicated case. In traditional journalism, I think the important principle is that any potential conflicts of interest are disclosed (and while a payment is not necessarily an actual conflict of interest, it is necessarily a potential conflict of interest). But the disclosure of potential conflicts of interest doesn't really work in the wiki world, because the work of many different authors (many of whom are anonymous or pseudonymous) gets jumbled together.

The problem, really, is the anonymity/pseudonymity. As you correctly point out, the list of potential conflicts of interest is endless. In fact, aside from someone who the reader pays to write a comprehensive, reliable and neutral reference work, there invariably is some sort of other motivation, which might potentially conflict. We deal fine with these potential conflicts, when they are disclosed, whether explicitly or implicitly.