I was reading
Dancy’s nonnaturalism entry to Copp’s Oxford Handbook of Ethical Theory. Dancy
discusses two arguments against naturalism which Parfit has presented in some
of his unpublished writings. The first one is called the triviality objection
against analytic versions of naturalism. Initially, I did find this argument
compelling. However, at the same time I was reading Blackburn’s Spreading the Word on the problems of
intensional semantics. I think there is a nice reply to Parfit on
those grounds – it’s the same strategy as used to reply to the open
question arguments.
Here’s how
Dancy puts Parfit’s triviality objection:
“Take
standard version of analytic naturalism: the predicates “is right” and
“minimizes suffering” have the same meaning. Now ask what, if so, could be
meant by saying that this act of minimizing suffering is right. All that can be
meant by this – since, according to the analytic naturalist, the predicates
“minimizing suffering” and “is right” have the same meaning – is that this is
an act of minimizing suffering, and that, as another way of saying the same
thing, we could say that it is right. But, this renders the second half of the
utterance a merely trivial addition to the first; it is a comparatively
insignificant fact that there is another way of saying this act would minimize
suffering. However, we all know perfectly well that second half of the
utterance is not a merely trivial restatement of what the first half said. This
argument works equally well for any other version of analytic naturalism”.
I’ve never
been a friend of analytic naturalism. But, as Mark S rightly pointed out such
intuitions are neither here nor there. I want to then put myself into the shoes
of an analytic naturalist and see if there is a way of answering to the
challenge.
argument hangs on what is meant by the meanings of the predicates and the fact
that two of them mean the same. I don’t know what Parfit’s theory of meaning
is. A way in which a naturalist would want understand
meanings would be to think of the meanings or senses of words as ‘the principles
of classification’ we apply when we successfully use the word. This would imply
that by saying that ‘is right’ and ‘minimizes suffering’ have the same meaning
we claim that when we think of actions as right and as minimising suffering we
are using the same principle of classification in both cases.
What could
then be meant by saying that minimizing suffering is right? Well, one
suggestion could be that this claim is to say that the actions picked out by
the principle for classifying actions under ‘minimizes suffering’ are the same
ones as the ones picked out by the principle for classifying actions under ‘is
right’. Of course, in the situation where we are certain that the principles of
classification are one and the same, it would not be interesting that this one
principle of classification picks out the same actions. But, I wonder if an
analytic naturalist could accept this. If we were in his shoes, then the claim
would be trivial for us.
Of
course, few of us are yet committed analytic naturalists and neither does the
argument hang on that. Rather, what is at issue is whether the claim could be
interesting for us as we are actually even when it turned out that the two
predicates meant the same. At this point, it is worthwhile to go back to the discussions of Moore’s
Open Question Argument. There it was pointed out that the principles of
classification we use to pick out things from the world with a given
term are
not transparent for us. Furthermore, the principles are not always even
propositional knowledge that but rather skills – knowledge of how to go
on about
classifying things.
true, then it is not usually obvious for us what our principles of
classification are. A typical example is our use of the predicate ‘is
grammatical’. We (well, not me really) have a skill of classifying sentences to
grammatical and non-grammatical ones. Yet, the principles with which we do this
are highly complex. Some of them are so complex that without the help of
thorough investigation in linguistics we cannot describe them. But,
it is still the case that those discoverable principles describe what we mean
when we say that a sentence is grammatical. This is so even when we are
presented with the principles and when it still seems like an open question whether
we mean what the principles say.
The upshot
of this is that that two predicates mean the same does not require that we know that
they mean the same, or even that we would without hesitation accept that they
mean the same after careful reflection. In this situation, I think the
naturalist can reply to the Parfit/Dancy challenge. She can say that the
grounds with which go on about making rightness-judgments are not obvious for
us. Doing such judgments is a skill produced by our moral upbringing.
However, once we have the enough data of the judgments we are making, we can
ask which implicitly accepted principle would best explain our judgments. At
this point, it could turn out that this principle tracks some natural property
like minimizing suffering. It could be even a very disjunctive natural property
like Jackson has argued.
In any
case, in this situation the claim that minimizing suffering is right would be
an interesting claim for us. It is a claim that the
principle of classifying acts under ‘minimizing suffering’ picks out the same
actions as the tacit principle of classifying acts under ‘right’. This would
be big news for us because we cannot directly access the latter principle in a propositional form. As Blackburn puts it, predicates with the same sense can have a different way of displaying that sense. A
naturalist has an explanation for why the principles pick out the same
actions – it’s because, contrary to our appearances, the latter
principle we have implicitly used turns out to be the same principle as the former
one. And, this sounds like a highly interesting discovery worth investigating.
This
does not mean that analytic naturalism doesn’t have any other problems. For
one, the view about meanings as classificatory principles is suspect. One might
think that even when to terms classify the same things they might still differ
in meaning because of their different inferential roles. This would get us to
Parfit’s second normativity objection.
Jussi, you write,
What could then be meant by saying that minimizing suffering is right? Well, one suggestion could be that this claim is to say that the actions picked out by the principle for classifying actions under ‘minimizes suffering’ are the same ones as the ones picked out by the principle for classifying actions under ‘is right’
It looks like this proposal is saying that the two predicates mean the same thing if (maybe also only if) they are extensionally equivalent. That would have two predicates F and G synonymous just in case the set of G things just is the set of F things. But those conditions are way too weak (recall Quine’s “creature with a kidney”, “creature with a heart”). This makes me think that I’m not following the proposed analysis here. Do you mean rather that the predicates must be necessarily coextensional? Or maybe something else?
Mike,
I think I meant something else. I was thinking that two predicates mean the same when they have the same Fregean senses, i.e., principles of classification. I thought these would be the epistemic ways in which one conceives of the reference of the term. In the pair you mention, these ways of conceiving the reference are different – one looks at kidneys and the other at hearts. Of course, it happens that from the actual world the same beings are picked out.
I think this means that this view of meanings has the implication that concepts with same meaning are necessarily coextensional. I don’t think it implies the opposite that terms that are necessarily coextensive must have the same meaning.
But, in any case, as the claim minimizing suffering is right is not a modal claim, I don’t think the way the naturalist account for its content should be either. Of course, from the hypothetical naturalist view it would follow that we could truly make the stronger claim that minimizing suffering is necessarily right. This would mean that the acts picked out by the principle of classifying acts as minimizing suffering are necessarily the ones picked out by the principle of classifying acts as right.
I think I see. You are trying to make a case for the non-triviality of the analytic proposition expressed by ‘x if right iff. x minimizes suffering’. The way you do this is to make successful application of the predicates independent of explicit knowledge of their meanings. So if we asked competent language users what the meaning of ‘is right’ is, it’s possible that they would be unable to tell us and yet be able to point to all/most of the things that fall in the extension of the predicate. Is that the idea?
But isn’t Parfit’s triviality objection that we do know the meaning of ‘is right’, or at least we know that ‘is right’ doesn’t mean minimizes suffering? We do seem to know that. Parfit’s position would not be undermined by the discovery that what best explains our use of ‘is right’ is some implicit principle that has us apply ‘is right’ to all and only those things that minimize suffering. That discovery would not entail that ‘x is right iff. x minimizes suffering’ is not a substantive moral claim.
Mike,
I think that roughly sums it up. I don’t think that Parfit can start an argument against analytic naturalism from the premise we know that ‘is right’ does not mean ‘minimizes suffering’. That is the conclusion he tries to argue for. His argument is supposed to begin from the premise: what if it was true that ‘is right’ means ‘minimizes suffering’. He claims that the relevant claim minimizing suffering is right would then be trivial in an unintuitive way. I tried to argue that triviality of that claim does not follow from the assumption about the meaning given the naturalists picture about meanings.
The discovered implicit principle in the scenarion would just be the principle minimize suffering. This would make the claim ‘x is right iff if x minimizes suffering’ non-substantial but still not trivial.
Jussi,
Nice post. As a committed analytic naturalist, I think you’ve described very nicely how to defend the position. (Although I think that in the case of moral use of evaluative language, there’s a lot going on pragmatically, of an expressivist nature). It is also worth noting that a lot of the force of criticism like Parfit’s and Moore’s stems from selecting the wrong candidates. ‘Right’ does not mean ‘minimizes suffering’.
Jussi,
Dancy’s argument is a reductio, as I see it. Where do we get the absurdity? We assume (1),
1. ‘x is right’ has the same meaning as ‘x minimizes suffering’.
From (1) it is supposed to follow that (2),
2. ‘x is right iff x minimizes suffering’ is trivial.
But according to Dancy we have this knowledge, (now I’m quoting Dancy),”. . . However, we all know perfectly well that second half of the utterance is not a merely trivial restatement of what the first half said”. Make that (3),
3. ‘x is right iff x minimizes suffering’ is not trivial.
Therefore,
4. Premise (1) is false.
There is nothing wrong at all in including (3) in an argument against triviality. What could be wrong with that? Just as I can include in my argument against the claim that “all fire engines are not red” that we all know that there are red fire engines. It just amounts to maintaining that (1) is close to obviously false, which it might well be. In any case, there is no argumentative flaw here.
The discovered implicit principle in the scenarion would just be the principle minimize suffering. This would make the claim ‘x is right iff if x minimizes suffering’ non-substantial but still not trivial.
Wouldn’t it be this principle that explains the usage,
P. x is right iff. x minimizes suffering.
And wouldn’t P explain the usage whether or or not it were a substantive moral principle? This is why I was suggesting that Parfit’s reply might be that you did not discover that they have the same meaning, you discovered a substantive moral principle.
Sorry, last quick point. Suppose your right about the meanings of these predicates. Does it non-triviality of C follow?
C. x is right iff. x minimizes suffering.
I don’t immediately see it, unless you have in mind some epistemological notion of triviality. Compare the triviality of T.
T. (~P -> (P -> Q)) <-> (~((P & Q) v (P & ~Q)) v ~P)
(T) won’t be obvious to everyone, but it is no doubt a trivial truth. So the fact that we have to work a little to discover that (T) is a tautology won’t show that it is not trivial. Similarly, the fact that we have to work a little to discover that (C) is true, won’t show that it’s not trivial.
Jussi, good post. I’m not quite sure what it would be for our “principles of classification” to be the same in the case of each term, or how we would know about it (as you point out in your replies to comments, coextensiveness of reference is not sufficient). In particular, the influence of background beliefs seems to cause problems.
Suppose that my deepest desire is for a job at Oxford. Then whenever I use the terms “an object of Simon Rippon’s deepest desire” and “a job at Oxford” correctly, I use them coextensively. Whether these terms mean the same thing is supposed to depend on the “principles of classification” I employ when I decide whether a thing falls under the term or not. But what are these principles? I perhaps have to find out what my deepest desire is for by examining my own behavior and feelings in response to certain events. Maybe I find myself poring over job listings and feeling disappointment or euphoria according to whether there is an Oxford job available for me to apply for. I’m clearly not using a principle of classification for my deepest desire like that which I use for “a job at Oxford” at this point. Ergo, the terms don’t mean the same. But suppose that now I come to firmly believe that my deepest desire is for a job at Oxford (maybe on the basis of the above experience). Now I don’t need to examine my psychology anymore in order to classify things as “an object of Simon Rippon’s deepest desire” or not. I just need to decide whether those things are in fact jobs at Oxford. And of course, at this point, I look for the same criteria as I would in deciding whether to apply the term “a job at Oxford”. Have the two terms now come to mean the same, by virtue of my holding the background belief? Presumably not. The presence of a background belief may make terms with different meanings seem as though they mean the same, on the account you propose. You would have be able to factor out the influence of the background belief to show that the “principles of classification” are really different.
Can the analytic naturalist tell the difference between different meanings and the influence of a background belief? Suppose that I examine how I go about classifying “is right” and “minimizes suffering”, and I decide that I in fact look for the same naturalistic features in each case. What do I conclude now? Do “is right” and “minimizes suffering” mean the same for me, or do I just have the background belief that actions that minimize suffering are right? I don’t know how to tell the difference between these supposedly different possible worlds.
I should of course have asked “…between sameness of meaning and the influence of a background belief?” in that last paragraph.
Jussi –
It strikes me that your response resurrects a point that was often made in the wake of Moore’s argument against analytic naturalism: The response (roughly) was that Moore wrongly insisted that a naturalistic definition of ‘good’ would have to result in a triviality, in the sense that the analysis would have to strike us as uninformative. But (critics said) this overlooks the paradox of analysis and the fact that correct analyses often are informative (e.g., the Theatetus analysis of knowledge as justified true belief). Later, Ryle (I think?) appealed to the knowledge-how/knowledge-that distinction to explain how we can correctly classify items under a predicate while being struck by the novelty (the non-triviality) of an analysis of that predicate.
Is that in the spirit of your suggestion?
Thanks people for all the interesting, and plentiful comments. I’ll see what I can do (probably not much) in some order that I’m not sure yet of.
Michael,
that is very much the spirit of my reply. Dancy actually brings up the paradox of analysis in his discussion of Moore but says of it only that he is ‘leaving that aside’. He then goes on to the other typical reply to the OQA that uses corefential terms with different meanings. Parfit’s objections are introduced as separete problems for naturalism taken from Sidgwick. But, of course, Sidgwick’s triviality objection influenced Moore’s argument in the ‘same question form’.
Simon,
that’s a good point. I’m not sure what to say. I do agree that when we investigate what someone means, i.e., what principles of classification she uses, we cannot look at applications of individual words at a time. Rather, in the Davidsonian fashion, we must interpret the language of the person as a whole. In this process, one must try to assign meanings for individual words such that these meanings can explain the uses of complex expressions in a compositionally.
This implies that we don’t look at complex constructions such as ‘an object of SR’s deepest desire’ and ‘a job in Oxford’ (my deepest desire too! Anyone reading this in Oxford…) but rather of individual words like ‘object’, ‘deepest’, ‘a job’, and so on. Given that they must have particular meanings so that they can be meaningfully used in other of your complex claims, it will be highly unlikely that the meanings of the complex claims you gave come out as the same even when you use these complex terms in same occasions. Their cognitive significance must then be different because the cognitive significance of their parts is under any good interpretation of all your utterances. Of course, ‘right’ which we are interested in is a simple term. If I remember right Horwich has more interesting things to say about the actual process by which we try to look for the tacit acceptance of which principle best explains the use of individual terms to give their meaning.
Mike,
It’s true that we have the knowledge that the discussed claim is not trivial. This is not the same knowledge that we know that the terms do not mean the same (as you wrote earlier), or so I wanted to argue. I agree that the argument has that form and that it hasn’t got a argumentative flaw. I tried to argue that (2) does not follow from (1) but rather (3) which would not be a problem for the naturalist.
I think you are also right that that is what Parfit might reply. But, I think then it comes down to the *best* explanation. If the principle is not read as a specification of meaning but rather a substantive moral principle, then the nonnaturalist has still two things to explain. First, she has to say what is the speakers have in mind when they use ‘right’ if it’s not ‘minimises suffering. Second, to get true claims they need to assume more ‘spooky stuff’ or end up with wide-spread error. If the cost of avoiding these problems is to take the principle as non-substantial then that seems ok if the triviality still does not follow.
I am happy with epistemic notion of triviality. I think that must be what Parfit had in mind. It would be a rather trivial objection that naturalism implies some non-substantial moral truths.
Thanks Jussi,
I mention the argument not just to point up that urging (3) as an objection to (2) is not argumentatively problematic.
2. ‘x is right iff x minimizes suffering’ is trivial.
3. We know that ‘x is right iff x minimizes suffering’ is not trivial.
Similarly (this is the point we were discussing) urging (2) against (1) is also not a flaw.
1. x is right means x minimizes suffering
2. We know that x is right does not mean x minimizes suffering.
No matter, your point is taken that this is not the route that Parfit takes in Dancy’s reconstruction above.
Just a quick response: If we know that “an object of SR’s deepest desire” and “a job in Oxford” don’t mean the same because their parts must retain the same meaning in other complex constructions, then obviously “is right” and “minimizes suffering” don’t mean the same thing either (which part of “is right” has the meaning of “minimizes”?)! But then your account would have the absurd consequence that your own analysis is obviously false, but “right” means “pleasurable” is not!
Simon,
I’m not sure the consequent of your conditional follows from the antecedent. It’s not that in complex constructions with same meaning the meanings of the elements must correspond even when their combinations do. So, ‘is a widow’ and ‘had a husband, who died while stile married to her, and has not since remarried’ mean the same as a consequence of the meaning of the elements even when the elements themselves do not correspond in meaning. So, ‘right’ and ‘minimizes suffering’ are still candidates to share a meaning if the combined meaning of ‘minimizes’ and ‘suffering’ comes to be our principle of classification as things as ‘right’.
I just thought that in your example it would be difficult to find word meanings such that the complex expressions would mean the same and the other complex expressions constructed from the meanings of the parts would still be sensible.
Jussi, thanks for the clarification. I misunderstood you because I can’t see why you would otherwise think that your point (let’s call it the “combinatorial restriction”) throws the equivalence of the complex expressions in my example into doubt if it doesn’t raise a similar doubt for is right/minimizes suffering.
I don’t of course think that “an object of SR’s deepest desire” and “a job at Oxford” mean the same, I just made the point that given the background belief that they are non-accidentally extensionally equivalent, we might use the same “principles of classification” when we apply either of them. The common “principles of classification” in this case would have very little to do with the meanings of the individual words “deepest”, “desire”, etc. I think your combinatorial restriction was supposed to rule out the idea that these principles reflect the _meaning_ of the expression and show that there’s a background belief influencing our behavior here.
The problem I raised for the analytic naturalist is to show that if we use the same “principles of classification” for “is right” and “minimizes suffering”, this isn’t because of a background belief that the two expressions are non-accidentally extensionally equivalent, but rather because they mean the same thing. The trouble is that the combinatorial restriction seems to rule out this equivalence too. The analytic naturalist claim is that we use the principles of classification for “minimizes suffering” when we apply “is right”. But these principles seem to have little to do with the meaning of the individual word “right” (consider, e.g.: “right view”, “right instrument”, “right answer”, “right choice”, “right page”). So the combinatorial restriction seems to show that the expressions don’t mean the same and there’s a background belief operating to influence our behavior here too: namely that the morally right actions are non-accidentally extensionally equivalent to those that minimize suffering.
Simon,
that’s good. That didn’t fix the problem but made thinks worse! Thanks. I think I understood your example slightly differently. I’m starting to think of the Quine’s ‘being with a kidney’ and ‘being with a heart’ case. Once you knew that in the actual world the were the very same things, you could begin to use of the classifications as an epistemic cue for the other. I think you are right that this wouldn’t change the meaning of that other expression.
But, in this case there is the modal difference. If the speaker was presented with possibilia where a being had a heart but not a kidney, even if she used looking at hearts for a cue for a being having a kidney of actual beings, she wouldn’t make the mistake here to call the being with a heart a being with a kidney too. So deep down she has to be classifying things differently.
Maybe something like this could be said about the objects of our deepest desires and jobs in Oxford. Even if actually seek for jobs in Oxford to satisfy our deepest desires, maybe when we were presented with counterfactuals where Oxford was quite different, we wouldn’t make the mistake of classifying one with the other. This would again show that deep down we have different concepts and classifications even when in some circumstances given contingent beliefs they seem to come to applying a same principle. So, maybe the distorting factors of background beliefs could be teased out with thinking about possible scenarios enough so that we find the fundamental grounds with which we classify things.
You are right that the combinational restriction brings problems with it. Compositionality for one doesn’t happen everywhere – a cat fish is not a combinations of meanings of cat and fish. So sometimes combinations of words is its own term and not a term which meaning is composed of the meanings of the elements. I also think that ‘right’ might be ambiguous in the cases you give in the same way as good is in attributive and predicative uses. Morally right seems like a predicative use whereas the others may be attributive.
Hi, Jussi.
I doubt that either Parfit or Dancy intends ‘trivial’ to be given an epistemic gloss. Both, after all, think that the triviality objection is going to apply to synthetic reductive views just as much as to analytic reductive views (though Dancy warms up, as critics of reduction always do, by criticizing the easier target and then claiming that everything is pretty much the same for the more elusive target).
Moreover, if you read closely, Dancy makes a big deal out of emphasizing that what the analytic reductivist (analytic naturalist) is committed to, is that some claim is normatively trivial; not just that it is trival. This argument, after all, is not supposed to be simply another version of the Open Question argument.
Now, this makes it very puzzling to me exactly what either Parfit or Dancy could mean by the charge of ‘triviality’. On one reading – suggested by some of Dancy’s explication – it sounds like the charge is simply that it is obvious that being wrong is not the same as being xyz, no matter what you supply for xyz. I don’t know how this sort of thing can be obvious, though, and my confidence in the probity of Parfit and Dancy’s insight is not encouraged by the fact that the reductive view they are actually considering is one that I think is hopeless on many independent grounds.
Mark,
I agree about that and I’ve been thinking about it too. It’s funny that Dancy doesn’t use normativity on the presentation of the argument but only after it. If *normative* triviality is the essence of the argument, how can it be presented without mentioning normativity? One way a naturalist could try to deal with the normativity side of the objection would be to use pragmatics as Steve suggested.
A worry is that if the charge of triviality is about normative triviality, then this objection collapses into the second objection – the normativity objection. There is even evidence for this. Dancy writes that:
‘For each form of analytic naturalism, then, there is an evaluative utterance that it renders normatively trivial…Note that the triviality objection does not show that *all* analytic naturalisms render *all* normative utterances trivial; that is the business of the normativity objection.
This makes it sound like there is one objection in more general and particular forms. It also threatens to make the triviality objection redundant.
I also wonder if it is a question of normative triviality, then why would Dancy think that it gets no purchase against Sturgeon’s non-reductive synthetic naturalism. On that view, the claims we are making are about actions that have one natural property having another natural property. That claim seems *normatively* just as trivial saying twice that an action has the same natural property.
Jeff King has some interesting work on the problem of analysis that is of relevance to the issues discussed here. According to King, simple predicates like ‘is right’ express (or mean) properties and relations. In this case, King would say that the property expressed is being right. On the other hand, King claims that complex predicates like ‘minimizes suffering’ do not express properties. Rather, they express certain complex semantic entities that have properties and relations (and sometimes individuals) as constituents. In this case, King would say that the complex semantic entity expressed by ‘minimizes suffering’ contains the relation of minimizing and the property of suffering as constituents.
In addition, properties and relations are sometimes complex; that is, sometimes properties and relations have other properties and relations as parts. According to King, an analysis is successful iff the property mentioned on the left hand side of the analysis has as its parts the properties and relations that are constituents of the complex semantic entity mentioned on the right hand side of the analysis. Thus, according to King, there can be successful analyses in which the left hand side of the analysis does not mean the same thing as the right hand side. (In fact, King seems to think that all successful analyses are like this.)
So, in this case, a proponent of naturalism could claim that the following is the correct analysis of ‘is right’ (or the property of being right):
Necessarily, for all x, x is right iff x minimizes suffering.
And he or she could say that this is the correct analysis although ‘is right’ and ‘minimizes suffering’ differ in meaning and so ‘Necessarily, for all x, x is right iff x minimizes suffering’ is not trivial.
Greg is right – more metaethicists should know about King’s work on philosophical analysis. But it’s hard to know whether it is actually supposed to help with the triviality objection that Dancy and Parfit intend to be making – if we take Dancy’s word for it, he claims that the triviality objection is supposed to be something totally different from the Open Question argument, which is problematic because of general considerations about the paradox of analysis of the kind that King’s account provides a (nice, I think) solution. That’s part of what makes it so puzzling (to me) just what the triviality objection is supposed to be.
If Jussi is right in his speculation that the problem really turns out to ride on the normativity objection, then I got the diagnosis right in my paper, ‘Realism and Robustness’. Ironically, Dancy goes on to endorse exactly the kind of account of normativity that I argue in that paper makes normativity turn out to actually be very easy for a reductive theory to account for.