Wolfgang Schwarz

Disagreement and Naturalness

If people disagree about whether a sentence S is true in a thought experiment, what could explain the disagreement?

1) They disagree about the meaning of S. Perhaps one party uses 'zombie' for revived corpses whereas the other uses it for people without phenomenal consciousness. The disagreement is 'merely verbal'.

That's not to say it isn't a serious disagreement, in particular if both parties think their usage corresponds to the folk conception, that is, if what they disagree about is whether S is true in the thought experiment according to the common, conventional usage of S in their community. In this case the disagreement can't be resolved by mere stipulation.

2) The parties disagree about what exactly is involved in the thought experiment. They disagree about the facts. In a sense, that's trivial, since by assumption they disagree about whether the fact that S is true obtains in the thought experiment. I'm not quite sure how to make the reason non-trivial. Here is one suggestion: Let's call the disagreement 'merely factual' if it can be resolved by giving more information about the situation described in the thought experiment. This only works if the information that S is true in the situation doesn't always resolve the disagreement. At least sometimes that seems right: If I strongly believe that a Gettier case G is not a case of knowledge, then adding a further clause to the description of G saying that it is a case of knowledge won't convince me that the situation is after all a case of knowledge. It will only make the thought experiment unintelligible to me.

(By the way, do those of you who believe that impossibilities can be true in fictions also believe that impossibilities can be true in thought experiments?)

Another way to characterize merely factual disagreement is to add a restriction on the vocabulary in which the situation is described. Suppose all parties agree that whether S is true can be settled a priori once enough information about the situation is provided in a basic vocabulary V not containing (all) the vocabulary of S. Then the disagreement is merely factual if it can be resolved by telling the complete V-story about the envisaged situation.

3) The two parties disagree about our own world. For instance, if one party believes that water is actually H2O and the other believes it is XYZ, that might explain their disagreement about what is true on twin earth. This is also a kind of factual disagreement. Again, I'm not sure how best to characterize it. If the fact that S is true in such-and-such a situation, or that 'knowledge' actually has such-and-such a meaning, counts as fact about our world, all disagreements end up being disagreements about our world. If we're lucky and all parties are able to consider a situation as actual (experience shows that some people seem sadly unable to do that), the distinction can easily be made by saying that a disagreement is 'merely about actual facts' if it can be resolved by considering the envisaged situation as actual. Another solution would be to resort again to the basic vocabulary V and say that the disagreement is merely about actual facts if all parties agree that if such-and-such V-sentences are actually true, then S is true (false) in the thought experiment.

Are there any further reasons for disagreement? So far, every disagreement that can't be resolved by providing further 'neutral' information about either out world or the thought experiment counts as a verbal disagreement. Some philosophers, like Brian Weatherson (see "What good are counterexamples?"), argue that there must be a further possibility: People can disagree about whether S is true even if they mean exactly the same by S and agree about all neutral facts. For instance, it seems that one-boxers and two-boxers don't just disagree about the meaning of 'rational', but neither do they disagree about the setup of Newcomb's problem. Similarly, when compatibilists and incompatibilists disagree about whether some case is a case of freedom, or when consequentialists and their opponents disagree about whether some action is good, that disagreement is neither merely factual nor merely verbal. (All examples from Brian.)

One further possibility should be rather uncontroversial:

4) At least one of the two parties lacks the cognitive capacity to properly apply their words to the envisaged situation. Consider a disagreement about whether some very complicated sentence is true. It takes a lot of cognitive processing to understand exactly what the sentence says and thus to find out whether what it says obtains in the situation. If somebody now comes up with the wrong judgement, the best explanation might be that he made a mistake in the cognitive processing. Let's say a disagreement is 'mereley cognitive' if it would disappear if all parties had unlimited, ideal cognitive capacities.

Perhaps the disagreement about Newcomb's problem is merely cognitive: If only the one-boxers were smart enough, they'd see that they are wrong.

Brian doesn't like this explanation either. Consider moral disagreement: Would ideal rationality solve the dispute between consequentialism and virtue ethics? Does either partly only make a cognitive mistake? That seems unlikely.

Brian suggests a further reason (which he, wrongly I think, attributes to Lewis):

5) The two parties mean exactly the same by S, but they don't quite know what they mean. That's because the meanings of our predicates depend on the degree of naturalness of various candidate properties, and we might be ignorant of those degrees. The idea is that what my term 'cat' applies to in various possible situations not only depends on which things I'm disposed to classify as 'cat' (and expect others to classify as 'cat'), but also on whether these things comprise a natural class. Some of my classifications might be wrong because they make the resultig class less natural than it would be if these classifications were made differently.

Note that reason (5) probably collapses into reason (4) if ideal rationality suffices to find out which classes are natural to what degree. At any rate, (5) is covered by (4) together with (3) if ideal rationality and sufficient information about our world in principle suffice to know facts about naturalness. So if (5) is really meant to be an additional explanation, degrees of naturalness must be neither a priori knowable nor a posteriori knowable: they must be altogether unknowable.

Perhaps that alone is not a very serious objection. Perhaps some features of reality really are completely beyond our grasp. But could such features really be relevant to the meaning of our words, in the way Brian suggests?

Well, "before we ask what meanings are we have to ask what meanings do". Maybe Brian's inscrutable objectively natural classes can do the job of meanings in some semantic project or other. I don't know. But I think they are ill suited to account for the information we convey by uttering sentences: Suppose there is a very natural class C comprising 95% of the things we're disposed to call 'cat' and 2% of the things we're disposed to call 'mushroom'. (It is absolutely impossible to find that out by investigating the relevant things.) On the other hand, let C* be the less natural class comprising all and only the things we're disposed to call 'cat' (and expect each other to call 'cat'). Now consider an utterance of "there is a cat on the roof". It is commonly expected that people utter this sentence only if there is a member of C* on the roof. If somebody utters the sentence without a member of C* being on the roof, people regard the utterance as false, no matter if there is a member of C on the roof. Uttering "there is a cat on the roof" is therefore a very bad means to communicate that there is a C on the roof unless that C is also a C*: On hearing the sentence, trusting listeners will infer that there is a furry animal on the roof, which is completely wrong if the thing on the roof is only a C. It seems to me that C therefore can't do the most most important job of meanings in a theory of communication -- for the same reason why C-intensions can't do that job: they are not in the head. C* fits the job much better. Who cares if it lacks some occult property of naturalness?

So it seems to me that on a useful conception of 'meaning the same', (5) doesn't work: If two parties really mean the same by S and agree about the facts, and are ideally rational, then they can't disagree about whether S is true.

What about moral disagreement? I think any of (1) through (4), and even better any combination of them, might explain it. That cognitive limitations are involved is revealed by the fact that people often change their moral judgements when carefully thinking about them, in particular by noticing that the situation falls under a certain general pattern. Some argue that "x is good" can be analysed as something like "x is valued in my community". If that's true then moral disagreement is disagreement of kind (3): It can be settled by finding out what is valued in the community. If all fails, one could introduce primitive moral facts and say that any situation described in purely descriptive terms underdetermines the normative facts. That would be an answer of kind (2). At any rate, I don't think we need something as bizarre as (5) to explain moral disagreement.