be apparent until the very end.]
Some options are more specific than others. One option is
more specific than another if and only if the performance of the one option
logically necessitates the performance of the other but not vice versa. Thus,
sitting down at my desk is more specific than sitting down. And sitting down at
my desk and then checking my email is more specific than sitting down at my
desk. An option is maximally specific (hereafter ‘maximal’) if and only if
there is no other option that is more specific than it.
Now, maximalism is the view that the deontic status of P’s
performing some non-maximal option depends on the deontic statuses of P’s
maximal options. For instance, on typical versions of maximalism, P is morally permitted
to perform a non-maximal option x if and only if P is permitted to perform some
maximal option that logically necessitates performing x. Call this principle D
(because it essentially says that permissibility distributes over conjunction).
So if we were both maximalists and utilitarians, we would employ the principle
of utility to assess only which maximal options are permissible and then rely
on principle D to assess the permissibility of non-maximal options contained
within those maximal options.
In a very interesting recent paper (2013), Gustafsson presents
three cases that each putatively pose a problem for maximalism. The first two
certainly don’t pose a problem for maximalism per se; they pose a problem only
for maximalism combined with a certain (to my mind, implausible) view about
which conjunctive acts count as being performable (that is, count as being
options). The third case, from my point of view, is much more interesting. The
case involves two choice situations: S1 and S2. In S1, P has the choice of
performing either a1 from t1-t2 (at which point S2 will arise) or performing a2
from t1-t3 (in which case P will never face S2). If P performs a1 from t1-t2, P
will then face S2, where she will have to choose between performing a3 from
t2-t3 and performing a4 from t2-t3. Here’s a graph of the case, which I borrow
from his paper.
Now, as Gustafsson notes, there are two ways that P might
end up performing a1 & a3:
- “(i)
P performs a1 in S1 without forming or having an intention at t1 to perform a3,
and P performs a3 in S2 having formed an intention to perform a3 after t1. - (ii)
P performs a1 & a3 in S1 forming or having an intention at t1 to perform a1
& a3.”
As Gustafsson notes, the difference between (i) and (ii)
is just the time at which P decides to perform a3. Whereas, in (ii), P decides
to perform a3 at t1, in (i), P decides to perform a3 after t1. And, as
Gustafsson points out, the consequences of (i) may be very different than the
consequences of (ii). For there may be good consequences associated with
putting off the decision about whether or not to perform a3 until after t1 and/or
bad consequences associated with deciding to perform a3 at t1. Let’s assume
that at least one of these is true.
This case, Gustafsson thinks, raises a problem for maximalism.
He says, “it seems plausible that both possibilities (i) and (ii) are open to
the agent and furthermore that (i) is the only way the agent can achieve the best
consequences in S1. On…maximalism, the relevant
alternative set in S1 is {a1 & a3, a1 & a4 , a2}. First, (i)
is not a way to do either of a1 & a4 or a2. Second, (ii) is a way to intentionally
perform a1 & a3 in S1. Hence there is no way on…maximalism to
prescribe possibility (i) without prescribing possibility (ii),
which is something a plausible version of consequentialism should be able to
do.”
I think that the maximalist can resist this argument by
arguing that if, as Gustafsson claims, it is up to P in the morally relevant
sense whether and when to form the intention (that is, whether and when to
decide) to perform a3, then both deciding at t1 to perform a3 (call is a5) and
deciding at t1 to defer deliberation about a3 until after t1 (call this a6) are
voluntary acts. (And let’s call the decision to perform a3 after t1 “a7.”) If
these decisions are all voluntary acts (as Gustafsson seems to be suggesting),
then the maximalist can rightly deny Gustafsson’s claim that, on maximalism, the
relevant alternative set in S1 is not {a1 & a3, a1 &
a4 , a2}. For if a5, a6, and a7 are indeed voluntary actions, then
a1&a3 cannot be a maximal option given that both a5&a1&a3 and a6&a7&a1&a3
are both more specific than a1&a3 is. It seems, then, that there is a way
for “maximalism to prescribe possibility (i) without prescribing possibility
(ii)”: prescribe a6&a7&a1&a3.
So I don’t think that his arguments against maximalism
work. Now, at this point, Gustafsson might insist that making the decision at
t1 to defer deliberation about a3 until after t1 is NOT a voluntary act. Or, if
this is not plausible, he can change the example so that the good consequences
stem not from performing any voluntary act but from forming certain
attitudes in a non-voluntary (and, hence, non-agential) way. But, in that case,
his objection is not so much an objection to maximalism as an objection to theories that are exclusively
act-orientated. So I think that insofar as Gustafsson is getting at an
important truth, it’s the one that I presented in my last post—that is, that the
best moral theory will not be exclusively act-orientated. Gustafsson’s case 3 (when
suitably revised so that a1&a3 is indeed maximally specific) just seems to
be an attempt to come up with intrapersonal version of Regan’s example—the one
that I called The Buttons in my
previous post.
I do not think that this objection works. As I described the case, the only acts that are available in S1 are a1, a2, and a1&a3. Hence a6 is not available in S1. For some reason, while the agent can defer the decision about a3, the agent cannot form at t1 the intention to defer the decision about a3. So the act a1&a6 (or a6&a7&a1&a3) is not available in S1, as I have described the case. Thus the maximalist cannot prescribe a1&a6 (or a6&a7&a1&a3), and then the objection does not work.
One might perhaps object that the case is unrealistic unless also a6 is available. But, if we add a6 to the case, we can distinguish between two ways of realizing possibility (i):
(ia) P realizes (i) and does not intentionally perform a6 in S1.
(ib) P realizes (i) and intentionally performs a6 in S1.
And then we suppose that a6 would be intentionally performed were the agent to intentionally perform a1. Furthermore, we suppose that the consequences of (ia) are optimal but those of (ib) are not optimal. Hence we would like to prescribe possibility (ia) without prescribing (ib). But maximalism cannot do that.
Finally, I do not claim that if it is up to you in the morally relevant sense whether to form a certain intention x, then you can form an intention to form x. Hence to form x need not be an act that you can intentionally perform in the situation, even if it is up to you whether to form x in the situation.
I might add that I am also in favour of theories that are not exclusively act-oriented.
Hi Johan,
Thanks for your reply.
You admit that, in S1, “P can defer the decision about a3,” and you admit that it is up to P whether or not he does so. So it seems to me that P’s deferring the decision about a3 is an option for P is S1. Let’s call the act/option of deferring the decision about a3 in S1 “a8.” Why can’t maximalism prescribe a1&a8 (or a8&a7&a1&a3)?
Perhaps, you think that it can’t because you claim that P cannot form the intention to perform a8, and you believe that P’s performing a8 need not be an act that P can intentionally perform in S1 even if it is up to P whether or not to perform a8 in S1.
But why should maximalism be concerned with only acts that can be intentionally performed in your sense? That is, why can’t the maximalist say that the relevant options are just those that are up to the agent in the morally relevant sense? Some acts like pushing a button are one’s that we do by intending to do them. Other acts (like deciding to push a button) are not acts that we do by intending to do them — we don’t intend to decide to push a button, but just decide to push a button. But, as far as I can tell, maximalism is neutral on how we conceive of an agent’s actions/options. Maximalism is committed to only one thing: the deontic statuses of non-maximal options depend on the deontic statuses of maximal options. It is not committed to the view that the only options are the one’s that the agent can do by forming the intention to do them.
I’m sorry about the typos. Replace “one’s” with “ones,” and replace “an option for P is S1” with “an option for P in S1” in the third sentence.
Doug, did you mean to say that Maximalism is equivalent to the principle that permissibility distributes over disjunction, rather than conjunction?
I think this is true only if the available acts have a certain kind of logical structure. It is true, for example, if they have the structure of a finite Boolean algebra. But it is not true for some other structures.
Here is a very simple (and silly) example of a structure in which it is not true. Suppose there are exactly three available acts, x, y, and z, where x implies y, and y implies z. And suppose only z is permissible. Now consider, e.g., the disjunction x-or-z. The principle that permissibility distributes of disjunction requires that if x-or-z is permissible then either x is permissible or z is permissible. But in this structure x-or-z = z. Thus the requirement is that if z is permissible then either x is permissible or z is permissible, and this is trivially true. The same goes for all other permissible disjunctions (e.g. y-or-z). But Maximalism is false in this case, because z is permissible but is not implied by any permissible maximal act.
My hypothesis is that the proposed counterexamples to Maximalism all involve cases where the logical structure of acts is such that distribution over disjunction does not entail Maximalism.
Hi Campbell,
The principle that I have in mind is the principle that says that permissibility distributes over conjunction. More specifically, this:
P(S, ti, [x1, x2, …, & xn]) → [P(S, ti, x1), P(S, ti, x2), …, & P(S, ti, xn)], where ‘P(S, ti, xi)’ stands for ‘S is permitted at ti to xi’.
But that principle doesn’t entail maximalism, does it?
Simplifying a little, the principle is this:
(DC) If P(x and y), then P(x) and P(y).
Here’s a case in which DC is true, but maximalism is false. Suppose there is some non-maximal act x such that for all acts y, P(y) iff x implies y. Now, because implication distributes over conjunction (i.e., if x implies (y and z) then x implies y and x implies z), it follows that DC is true. But maximalism is false in this case, because no maximal act is permissible, right?
I never said that it did entail maximalism. All I said was that, on one version of maximalism, we use the principle of utility to assess the deontic statuses of maximal options and use the principle that holds that permissibility distributes over conjunction (viz., Principle D) to assess the non-maximal options whose performance is logically necessitated by the performance of some permissible maximal option.
Where do you take me to have said or implied that Principle D entails maximalism — that is, “the view that the deontic status of P’s performing some non-maximal option depends on the deontic statuses of P’s maximal options”?
Ah, I misunderstood.
I had in mind this passage:
I took you to be saying that D and DC are equivalent (i.e., they say the same thing). I take it now that you meant only that D entails DC, which is true, I think.
Right. Sorry. That could have been more clear.
No worries.
Actually, I now realise that part of what I said earlier was incorrect. Let me try again.
Consider these two principles:
(1) Permissibility distributes over conjunction, i.e., for all acts x and y, if P(x and y), then P(x) and P(y).
(2) Permissibility distributes over disjunction, i.e., for all acts x and y, if P(x or y), then P(x) or P(y).
(3) Every act is the disjunction of a finite set of maximal acts, i.e., for every act x, there exist maximal acts y1, y2, … yn such that x = (y1 or y2 or … yn).
Together these imply maximalism, by which I mean this principle:
(M) An act is permissible iff it is implied by a permissible maximal act.
The right-to-left part follows immediately from (1), because (1) is equivalent to the principle that permissibility is closed under implication. The left-to-right part can be proven by induction. Here’s a sketch of a proof. Begin with the case of a disjunction of two maximal acts. Suppose x = (y1 or y2), with y1 and y2 both maximal. Then it follows from (2) that P(x) only if either P(y1) or P(y2). Next consider the case with three maximal acts, i.e., x = (y1 or y2 or y3). In this case (2) implies that P(x) only if either P(y1) or P(y2 or y3), and it also implies that P(y2 or y3) only if either P(y2) or P(y3). So we have P(x) only if either P(y1) or P(y2) or P(y3). It should be clear that the same reasoning could be used for any finite number of maximal acts.
So opponents of maximalism must deny either (1), (2), or (3). The first two seem quite hard to deny. So that leaves (3). And in fact this does appear to be false in the proposed counterexamples to maximalism. Consider Johan’s example which you discuss above. Here, in S1, there are only two maximal acts available, a2 and (a1 and a3), and one non-maximal act, a1. But it cannot be the case that a1 = (a2 or (a1 and a3)), because it is possible to do (a2 or (a1 and a3)) without doing a1, i.e., by doing a2. So a1 is not the disjunction of any set of maximal acts.
Your response to Johan’s example is to redescribe it. I suspect (though I’ve not thought about this carefully) that on your redescription (3) will turn out to be true. This would explain why maximalism is plausible for your description of the case, but not for Johan’s.
Hi Campbell,
I don’t take myself to be responding to Gustafsson by redescribing the case that he gives (Case 3). I want to go with his description of the case, as it is given in the paper and in the above. I just question whether, given his description, it is true that (a1 and a3) is a maximal act, which is crucial to his argument. I think that, given his description of the case (specifically, that he describes it as one where it is up to P where or not he defers the decision about a3), then it seems that so-called “maximal” act (a1 and a2) are not maximally specific.
Why do you think that I’m redescribing the case?
Hi Campbell,
One more thing: I’m having a hard time understanding your (3). I take it that “=” doesn’t stand for “is identical to” but stands instead for something like the following: “it’s logically impossible to perform one of the maximal disjuncts to the right without performing the non-maximal act to the left’. Do I have that right?
In this sense, a1 = [(a1 and a2) or (a1 and a3)]. Or am I misunderstanding.
It’s just seems weird to me to let the equal sign to equate a non-maximal option with a disjunction of maximal options.
Of course, I would also like to know whether (and, if so, why) you think that I’m redescribing Gustafsson’s case.
I should of said “=” stands for “performing the act to the right logically necessitates performing the act to the left.” That’s simpler. Is that the idea?
Campbell: The more I think about it, I can’t fathom how you would come to use the = sign for an asymmetric relation such as the one that I suggest above. So you must take “=” to stand for symmetric relation such as “performing the act to the right necessitates performing the act to the left AND vice versa.” Is that it?
Now it’s intuitive to think that if you’re going to do some non-maximal option N (e.g., raise your hand) you must do so in some maximally specific way (e.g., slowly, fastly, before waving it about above your head, etc.). Thus, for every act that you can perform there will be at least one disjunct in the set of the disjunctions of all your maximal options that necessitates performing that act. But why think that it goes in the other direction? It’s possible that one of my maximal options includes my raising a hand at t1 while another includes my not raising a hand at t1. So it would seem that performing one of the disjuncts in the set of disjunctions of all my maximal options doesn’t necessitate my raising my hand at t1. So why think that “for every act x [such as raising a hand at t1], there exist maximal acts y1, y2, … yn such that x = (y1 or y2 or … yn)”?
Let me just add one more thing, and then I promise to go to bed.
The intuitive idea behind maximalism is, I take it, that because it’s impossible to do x (e.g., raise your right arm) without doing so in some maximally specific way, then it must be that if it is permissible to do x, this must be because there is at least one maximally specific way of doing x that is permissible, where doing y is a maximally specific way of doing x just in case doing y logically necessitates doing x and there is no z that logically necessitates doing y but not vice versa. Thus, the deontic statuses of non-maximal options depends on the deontic statuses of one’s maximal options, for you can’t perform a non-maximal option without performing some maximal option whose performance necessitates its performance.
But note that the idea is not that for any non-maximal option, there exists a disjunctive set of all of one’s maximal options such that performing that disjunctive set necessitate performing that non-maximal option.
Hi Doug,
1. By “redescribing” I just meant that your description of the case differs from Johan’s, because, e.g., you use names for actions, “a5”, “a6”, and so on, which Johan didn’t use. I did not mean to say that you were changing the case, only that you were suggesting a different and (to your mind) more perspicuous way of describing it.
2. By “=” I do mean identity. I think that, e.g., the act of going to the movie and the act of either going to the movie and eating popcorn or going to the movie and not eating popcorn are identical. The italicised passages are different descriptions of one and the same act. As an analogy, consider the statement “1 + 3 = 6 – 2”. Here “=” denotes identity. The expressions “1 + 3” and “6 – 2” are different descriptions of one and the same number, namely, 4.
It is important to remember that, in the principles stated above, expressions like “(x and y)” or “(x or y)” are not sentences (though they do look like the sentences one finds in, e.g., propositional logic). Rather they are definite descriptions: “(x and y)” means something like “the act of doing both x and y”.
I think it would be odd to hold that logically equivalent descriptions denote numerically distinct acts. Do you think, e.g., that x, (x or x), (x or (x or x)), and so on, are all numerically distinct?
3. Perhaps it will be helpful for me to explain what I mean by a disjunction of acts. This can be defined in terms of the relation of implication (or, if you like, logical necessitation) between acts: the disjunction of acts x and y, i.e., (x or y), is least upper bound of the set {x, y} with respect to implication. That is, z = (x or y) iff (i) z is implied both by x and by y, and (ii) any other act that is implied both by x and by y is also implied by z.
This is just what mathematicians call a “join” in lattice theory. See, e.g., http://en.wikipedia.org/wiki/Lattice_%28order%29
For the purposes of the argument I gave above, I think it would be sufficient to assume that every act is an upper bound of some set of maximal acts (i.e., it needn’t be the least upper bound). That would be a slightly weaker premise than my (3).
4. I agree with what you say in your last comment:
Hi Campbell,
Okay, that’s very helpful. Thanks. If I understand things now, (3) just says:
Every act is identical to the disjunction of all the maximally specific ways of performing that act.
What got me confused was this statement of yours: “So that leaves (3). And in fact this does appear to be false in the proposed counterexamples to maximalism. Consider Johan’s example which you discuss above. Here, in S1, there are only two maximal acts available, a2 and (a1 and a3), and one non-maximal act, a1. But it cannot be the case that a1 = (a2 or (a1 and a3)), because it is possible to do (a2 or (a1 and a3)) without doing a1, i.e., by doing a2. So a1 is not the disjunction of any set of maximal acts.”
Because why can’t Gustafsson just say that a1 = ((a1 and a3) or (a1 and a3))? That is, since he thinks that there is only one maximally specific way of performing a1 — that is, by performing (a1 and a3) — he holds that a1 is equal to the set of maximal acts that includes only (a1 and a3).
So I don’t think that Gustafsson is denying (3) or (M). His objection is instead the following:
(A) There are two ways to perform a1: “(i) P performs a1 in S1 without forming or having an intention at t1 to perform a3, and P performs a3 in S2 having formed an intention to perform a3 after t1. [And] (ii) P performs a1 & a3 in S1 forming or having an intention at t1 to perform a1 & a3.”
(B) “Both possibilities (i) and (ii) are open to the agent and furthermore…(i) is the only way the agent can achieve the best consequences in S1.”
(C) Given (B), any plausible (consequentialist?) theory must prescribe (i).
(D) (i) and (ii) are not maximal options — that is, they are not maximally specific ways of performing a1. For he says above: “As I described the case, the only acts that are available in S1 are a1, a2, and a1&a3.”
(E) If (i) is not a maximal option, then maximalism cannot prescribe (i).
(F) Therefore, maximalism is not a plausible theory.
Premise E just follows from the fact that moral theory can only prescribe some member of the relevant alternative set in conjunction with Gustafsson’s definition of “Maximalism: The relevant alternative set for P in S consists of all acts that are maximal for P in S.”
But I don’t see why the maximalist shouldn’t just deny premise D. (i) and (ii) are certainly maximally specific ways of performing a1. And if we are to accept that “both possibilities (i) and (ii) are open to the agent,” then it seems that we should also accept that they are options.
Gustafsson claims that possibilities (i) and (ii) are two distinct possibilities that are open to the agent, but that they are not two distinct available options.
He can’t say this, I think, because a1 and (a1 and a3) are supposed to be distinct acts. More exactly, a1 is supposed to be a proper part of (a1 and a3), i.e., a1 is contained in, or implied by, (a1 and a3), but not vice versa. But ((a1 and a3) or (a1 and a3)) = (a1 and a3). (It follows from the definition of disjunction I gave above that (x or x) = x, for any x.) So if, as you suggest, a1 = ((a1 and a3) or (a1 and a3)), we get the bad result at a1 = (a1 and a3).
This shows an odd thing about the example. Normally, we would expect that if x is the only maximally specific way of doing y, then x = y. This would be so in a Boolean algebra. But in the example, as I understand it, (a1 and a3) is the only maximally specific way of doing a1, yet it is not the case that (a1 and a3) = a1.
I’m not sure what to say about the argument, but I agree that premise (D) looks suspect, for the reasons you give.
Hi Campbell,
I see; that makes sense. Thanks for your patience. This has been helpful.