Thursday, August 09, 2007

Maximizing, Satisficing and Gradability

Greetings from the BSPC, now complete apart from Recreation Day. Soon to follow: BSPC participants sorted into their Harry Potter houses, and lots of photos. But first, some philosophy.

This is actually unrelated to anything that happened during the sessions, and is instead something I have been chatting about with Daniel Nolan (who, incidentally, should get a joint-authorship credit on this post for helping me write up the idea and improve my examples, though I do not have evidence that he is committed to the view itself, nor should any errors herein be attributed to him, etc.).

The idea is that gradability can help accommodate the apparently conflicting intuitions of Maximizing and Satisficing consequentialists.

Maximizers think that only the action(s) with the best consequences are right; all others are wrong (though perhaps to greater or lesser degrees). Satisficers think that all actions with good enough consequences are right, and that there may be several actions, with consequences of differing values, which have good enough consequences. (It need not be assumed that to be good enough a state of affairs has to be good simpliciter; the least worst option may count as good enough even if it is not very good at all.)

My basic thought is that ‘right’ appears to be a gradable adjective like ‘tall’ or ‘flat’. Familiarly, in some contexts, such as when we are talking about basketball players, ‘tall’ is used in a very demanding way, so that someone has to be at least 6’5’’ to fall within its extension. In other contexts, such as when we are talking about children, it is used in a less demanding way, so that someone who is only 3’5’’ falls within its extension.

Another example of gradability may be helpful on the way to the gradability of ‘right’. Consider ‘at the front of the line’. (I’m in the US so it’s a line rather than a queue.) Sometimes, we use that phrase in such a way that only the one person at the very front of the line counts as ‘at the front of the line’. For instance, if we ask ‘Who is at the front of the line?’ because we want to award a prize to the person who is next to be served, we are using it in this demanding way. On other occasions, we use it in such a way that the first few people count as ‘at the front of the line’. For instance, if you and I join a queue of 50 people and I then notice that Ross is in fourth in line, I might say to you ‘It’s OK, we can queue-jump: I know someone at the front of the line’.

The idea about ‘right’, then, is that in some contexts, ‘right’ is used in a very demanding way, so that only the action with the best consequences will be in its extension. On other occasions of use, ‘right’ is used in a less demanding way, so that any action with good enough consequences is in its extension. This is a common phenomenon in natural language; there are other gradable phrases, like ‘at the front of the line’, which are also sometimes used in such a way that only the first thing in some ordering falls within their extension, and on other occasions used in such a way that the first n things in that ordering fall within their extension (for some n>1).

The Maximizers and the Satisficers are therefore both half right; they are each offering a good account of how ‘right’ works on certain occasions of use. Both are motivated by good intuitions, which I think we can accommodate with this gradability point. Comments welcome (including especially, since I don’t know this literature well, comments of the form “wasn’t this said by X at t only better?”).

10 comments:

Robbie said...

There's a bunch of consequentialists at the moment pushing the "scalar" view: that what we should be aiming for with consequentialism is comparative judgements about rightness: that this is more right than that, etc. If all the substantive explanatory work can be done by the scalar stuff, then we shouldn't be worried by "right" turning out to be so context-sensitive. On the other hand, if "right" itself had a serious theoretical role (which I guess is the dominant view?), I'd get nervous about a view that made it so context sensitive...

Incidentally, how do you feel about extending the tactic to other philosophical debates which arguably have an underlying degreed notion and a maximizing/satisificing debate. E.g. credences and belief: (maximizer: credence 1 is belief, etc). Maybe justification fundamentally comes in degrees to... and so on. Are you happy with similar tactics? If not what do you think are the relevant difference between the cases?

One thing I've been thinking of in this sort of connection is the degree theory of truth. Lots of people have the intuition that only truth to degree 1 deserves the name "truth" simpliciter (that's the maximizing intuition). But that may be overdemanding, if most of what we assert is only true to a very high degree: so in most contexts, satisficing seems more appropriate. Again, it seems to me the key is what is doing the explanatory work. If "truth to a degree" is what is explaining norms for belief, consequence and assertability etc then nothings lost by the woosy view of truth. If we really need all-or-nothing truth for some serious theoretical purpose, then we need to think some more...

Daniel Nolan said...

Wasn't this said by Y at t1, only better?

Carrie Jenkins said...

Hi Robbie,

Well, as I think we've discovered before, I'm much less worried than you about being a contextualist about matters that play a serious theoretical role. The question of whether to carry something similar across to credences and beliefs is interesting - I have actually been toying with some similar ideas about that, but I hadn't spotted the connection. So thanks! I'm less tempted by the truth version, but I'm suspicious of degrees of truth in general.

Robbie said...

Fair enough. The thought was just that *if* you were already prepared to do the work with the scalar stuff, then doing the thing in this sort of way would seem extremely natural and unobjectionable to me.

Of course, you could endorse the more ambitious thing of being contextualist about an ethical notion *that matters*. (My argument against that is: Yuck :-)

Campbell Brown said...

I've pursued a similar idea, though not in print. My idea was that, because ought implies can, we might be able to explain the gradability of "ought" (or, I suppose, "right") by appeal to the gradability of "can". I understand John Hawthorne has independently pursued this idea.

Mike said...

For what its worth, Sorensen talks about similar cases (the standing in line case in particular) in 'Moral Dilemmas, Thought Experiments and Conflict Vagueness' Philosophical Studies 63 (1991) 291-308. What in your view indicates a difference in meaning of 'being at the front of the line' or 'being right', Sorensen argues is in fact an instance of vagueness (i.e., conflict vagueness). There is a correct resolution of that vagueness--no surprise there, for Sorensen--so being in the front of the line resolves correctly in just one way. Of course, we can't know what way that is. He'd likely say the same for 'right', where, for all we know, it resolves correctly to the maximizing side.

jonny said...

Couldn't the ambiguity be resolved by the use of "the", since "The right thing to do" does imply uniqueness, that there is only one right action. Gradability would allow the phrase "That was a right thing to do", which sounds wrong to my ear.
I tend to think that "right" can be reduced to good, better and best, "best" replacing the maximisers "right" and "good" replacing the satisficer's "right", and we've still got "better" left over to compare actions which are both good.
That was a good thing to do
that was a better thing to do,
that was the better thing to do
that was the best thing to do.

visioneerwindows said...

Am glad ye mentioned 'context' because that is the key word here......

Brian King (who capitalises his initials, damnit) said...

I have yet to be convinced by examples that permit us to have multiple but morally disparate right acts. But maybe this is because satisfaction consequentialists aren't interested in the vulgar notion of 'the right' so much as a more general notion of 'righteousness'. That seems to me a more, a ha, satisfactory if not maximally satisfying explanation of one's ambivalent intuitions.

Julien said...

Hi Carrie,

I'm discovering this a bit late, I hope you get the comment!

Kennedy and McNally (Scale structure, degree modification and the semantics of gradable predicates) have a few interesting tests that you might want to apply. First, the scale of some adjectives is upper-bounded, and the test for this is "completely":

#He is completely tall
The window is completely open
He was completely right

Second, some have a lower bound, and their test for this is that the opposite adjective takes "completely":

#He is completely short
The door is completely closed
He was completely wrong

(For instance, "dry" has an upper bound, because you can say "completely dry", but no lower bound, because you can't say "completely wet". Conversely "wet" is lower-bounded only.)

When the scale is both upper-bounded and lower-bounded, you can use "half" and "partially":

#He is partially short
The door is partially/half open
He is partially right

Now their important claim is this: when an adjective has a boundary (or two), it's not semantically context-sensitive. It refers either to the maximum or the minimum point of the scale. (As they note that's a point Unger made by distinguishing absolute vs relative gradable adjectives.) So if you get context-sensitivity there it's going to be through some kind of pragmatic mechanism (e.g. Lasersohn's "pragmatic halos", cf the eponymous paper). Given that "right" seems to be bounded, the prediction is that it's not context-sensitive after all. It would either designate a maximal degree of being right or a minimal one (more precisely, any degree above 0 - for instance, something is "wet" as soon as it is minimally not-dry).

The way they test whether the adjective designates the upper or lower bound is like Unger's: if it's odd to say that O is F but not completely, then F designates the upper bound:

??The glass is empty, though there is some water left in it.
??The towel is dry, though there is some water on it.

That test does not always give clear results, though. But assuming it works, I guess "right" will come out as a "maximum" adjective:

?What he did was right, though it was a bit bad/wrong.

That would be in favour of maximizers - at least as far as the semantics is concerned. But you might argue that even when a gradable adjective is semantically maximal, it is subject to pragmatic context-sensitivity.