Moral Obligation Applies to Act-Types, Not Act-Tokens

In a couple of conversations that I’ve had recently, I’ve
been surprised to find that several moral philosophers believe that when we
are
morally obliged to do an act, this “act” that we’re obliged
to do is an act-token, not an
act-type. In my view, this is confused: the act that we’re obliged to
do
is always an act-type.

There are two main views of act-tokens out there:

  1. Act-tokens are Davidsonian events, which are concrete
    particulars, individuated by their basic physical properties (including their
    position in the causal order).
  2. Act-tokens are Kim-style entities corresponding to a trio consisting of an agent x, an act-type A, and a time t, such
    that the agent x does something
    of act-type
    A at time t.

On either view, the idea that moral obligation applies to act-tokens runs into grave problems.

First, there are negative moral obligations — e.g. obligations to refrain
from rape and murder and the like. Suppose that you fulfil your moral obigation
not to murder me. Your not murdering me seems to be an omission rather
than an act. Was there really an act-token of your not murdering me? If so, when
and where did it happen? Or were there several act-tokens of your not murdering
me? If so, how many?

This problem is fatal on the Davidsonian view, on which the non-occurrence of an
event is not itself an event. It is not quite fatal on the Kim-style view, since
we could in principle say that you are at every time in your life the agent of
an act-token of not murdering me. Indeed, it seems, over your life as a whole
you are the agent of a vast (presumably infinite) number of act-tokens of not
murdering me (not murdering me today at noon, not murdering me today at 9 am,
not murdering me yesterday at 6 pm,  …). If moral obligations apply to act-tokens, then your obligation not to murder me is in fact an infinite set of obligations not to perform each of these act-tokens. This seems an
uncomfortable position to be in.

It seems easier to suppose that your obligation not to murder me is an obligation to ensure that whatever you do belongs to the
general act-type not murdering me. (Of course, almost all acts that you
could do — singing a song, writing a blog post, paying your bills — would belong
to the very general act-type of not murdering me. But that's all right: one of the convenient
features of negative obligations is that there are so many more ways of fulfilling
them than there are of fulfilling positive obligations….)

Secondly, suppose that you have a positive moral obligation to pay me £20
by midnight tomorrow. You could fulfil this obligation equally well either
by
giving me the £20 today, or by giving it to me tomorrow.

As I shall shortly argue, the act-token A1
that you would perform in a possible world w1
in which
in which you gave me the £20 today cannot be identical to the act-token A2 that you would perform in
any possible world w2 in
which in
which you gave me the £20 tomorrow.

It is clear that on the Davidsonian view, A1
(the act that you would do if you gave me the money today) cannot be identical
to A2 (the act that
you
would do if you gave me the money tomorrow), since their physical properties
and their positions in the causal order are completely different.

On the face of it, the Kim-style view of act-tokens also says that A1 and A2
must be distinct act-tokens, because they are located at different times. Someone might suggest that we could allow
A1 and A2 to be identical to each other if we allowed
the “time t” to be a 48-hour stretch of
time including both today and tomorrow. But that suggestion faces severe problems. For
suppose that in a third possible world w3
you give me £20 today and also give me £20 tomorrow. Clearly there
are two act-tokens in w3.
But each of them corresponds to the trio consisting of the agent (you), the act-type (giving me £20) and the “time” (the 48-hour stretch in
question). Still, these two act-tokens in w3 are two – they are
not one and the same act-token. The best way to avoid this problem seems to be to define the
relevant “time” as the precise temporal location of the act-token,
not a long stretch of time within which the act-token occurs. But then A1 and
A2 must be distinct act-tokens.

If this is right, there is no such act-token as your giving me £20 either
today or tomorrow. There are at least two different
act-tokens each of which would be an instance of your giving me £20
either today or tomorrow. Since you are obliged to give me £20 either
today or tomorrow, your obligation is simply to perform an act of this type.

Of
course, if you fulfilled your moral obligation to give me £20, there
would be a particular act-token
of which
you would be the agent. But you weren’t obliged to be the agent of
precisely that act-token. You were
only obliged to be the agent of some act-token or other of the relevant general
type.

The exercise of generalizing this argument to apply to all
moral obligations (and indeed to all reasons for action, etc.) is left to the
reader!

44 Replies to “Moral Obligation Applies to Act-Types, Not Act-Tokens

  1. Ralph,
    I think I’m convinced! What I don’t understand is why someone would hold the opposing view. When the reasons demand that you perform an act, _they_ don’t seem to demand one action rather than another if these actions are of all the same morally relevant types.

  2. Thanks, Clayton!
    I think the typical motivation for the opposing view is something like this: Moral properties (including moral statuses like obligatoriness) should primarily be instantiated by concrete things, since it is concrete things that really matter in life; act-tokens are concrete, while act-types seem too abstract to be the kinds of thing that matter enough to be the bearers of such crucial moral properties.
    Compare the way in which many philosophers are unhappy with the idea that abstract things like unactualized states-of-affairs can really be good in any way. They think that it is only concrete states that are good. (At most, it may be that if a certain abstract state-of-affairs were actualized, then there would be a corresponding concrete state that was good in the appropriate way.) I think that there is a broadly similar motivation behind the idea that moral obligation applies to act-tokens rather than act-types.

  3. There is also the important problem that obligation is prospective, not simply retrospective. If you ought to do A, and you never do, then there is no concrete event of your doing A, but you still ought to do it. Similarly, there is no trio of an agent, act-type, and time, such that the agent performs an action of that type at that time. But there is still something that is obligatory – so what is obligatory can’t be a token.

  4. Thanks, Mark!
    I’ve heard two responses to the problem that you mention from the act-token people.

    1. Some of them reject actualism, and say that non-actual possible act-tokens exist. I regard this as pretty crazy, but I’ve heard reputable philosophers defend it.
    2. Some of them say that the obligation to F consists in the truth of a pair of subjunctive conditionals: if you were to F, you would thereby do an act-token that had the property of obligatoriness; and if you were to fail to F, you would thereby do an act-token that had the property of impermissibility. This feels very ad hoc to me, but I haven’t got a fully-fledged objection to it.
  5. Hmh. I can see the force of the arguments, but something in me really wants to resist this (not that I am happy with the other option either).
    Here’s a naive worry. Say that I am obliged to help you. On this view, this means that I am required to do an act-type. Well, I have no idea how to do that. I don’t even know what that means. How can I do an act-type? I know I could do an act of a given type, but wouldn’t this be a token?
    So, I guess I would like to hear more about what the act-types are ontologically speaking too. This might be just my nominalism speaking.
    Here’s an alternative picture on the top of my head. There are a lot of act-tokens; some of them are actual, some of them *merely possible*. We also have the quantificational tools to talk about many of these tokes by using the logical vocabulary.
    Start from the negative duties. I have a duty not to murder – this means that *all* the possible (including merely possible and actual if there are any) tokens of me murdering have the deontic property of not to be done.
    In the second case, one can form a disjunction of all the possible tokens of me paying $20 to you by tomorrow. Then, one can think of the obligation as an obligation to do (either token1 or token2 or token3…). I’m sure that there is a way of doing this with existential quantification and ruling out many tokens too.
    I know I haven’t worked out the details. But, I do want to think of types in terms of possible tokens and therefore I do think there must be away of telling a story of requirements to do an act-type in terms of requirements to do tokens with the quantification and other logical machinery.

  6. Jussi —

    1. I think the alleged problem “How can I do an act-type?” is a complete pseudo-problem. We don’t have to use the peculiar phrase ‘do an act-type’ at all. The only point is that your behaviour at a certain time is behaviour of a certain kind. There is surely nothing weird or puzzling about the idea of a general kind of behaviour. E.g. walking, writing, playing the piano are all kinds of behaviour, since there are many different instances of these kinds of behaviour, occurring on many different occasions. All that I mean by an “act-type” is one of these perfectly familiar kinds of behaviour.
    2. Your proposed solution involves quantifying over merely possible non-actual act-tokens. As I said above, I regard this as pretty crazy, but admittedly that reflects a general view in metaphysics, not any special view about ethics in particular.
    3. Act-tokens are concrete things. I don’t understand how there could possibly be disjunctions of concrete things. (What’s the disjunction of my bicycle and my computer?!)
  7. A couple of questions.
    1. What’s wrong with saying the following? I just scratched my nose. In so doing I didn’t murder you. So by performing this act (token), I fulfilled my obligation not to murder you.
    2. You say you don’t like merely possible acts. I wonder, then, what you think act-types are. On one view, they’re sets of act-tokens, including merely possible ones. So, what are types, and what makes their ontological standing any better than mere possibilia?

  8. From what I can tell, there is less here than meets the eye.
    Suppose I murder Ralph. Well, I did something with the property of being impermissible. What did I do? Well, I performed a concrete act-token, namely the murder of Ralph. It is that token which is impermissible, and it is because of that act-token that I will go to jail. So there is some sense, I take it, in which act-tokens bear deontic properties.
    At the same time, there is the reasonable point that it is not an idiosyncratic feature of that particular act-token that makes it impermissible. Any act-token falling under the same type would also be impermissible. So it is also correct to say that there is a true, universally quantified proposition, something like “Any murdering of Ralph is impermissible,” and this proposition will include the property or type of “__ is a murdering of Ralph”. It does not strain English to say that this (true) proposition is the obligation. In that sense, obligations apply to act-types.
    We could ask what “Any murdering of Ralph is impermissible” quantifies over. I would think it quantifies over actual act-tokens: any act-token which is a murdering of Ralph has the property of impermissibility. If nothing is an act-token which is a murdering of Ralph, then (as far as this obligation goes) nothing has the property of impermissibility. So I don’t think it’s necessary to quantify over possible act-tokens; it’s sufficient to say “Necessarily, any murdering of Ralph is impermissible” and “If it were the case that an act was a murdering of Ralph, that act would be impermissible.”
    BTW, I am not nearly as bloodthirsty as this comment makes me sound.

  9. ‘If nothing is an act-token which is a murdering of Ralph, then (as far as this obligation goes) nothing has the property of impermissibility.’

    Nevertheless (and regardless of how bloodthirsty Heath might be), in actual fact, murdering Ralph is impermissible. So something does have the property of impermissibility – namely, murdering Ralph.
    Or are we supposed to understand ‘murdering Ralph is impermissible’ as really having a modalized and quantified structure?

  10. I don’t think there’s anything bizarre about claiming that “Murdering Ralph is impermissible” is quantified. “Life is beautiful,” “Prestidigitation is amusing,” “Triangularity is necessary and sufficient for trilaterality” are all plausibly understood to be talking about instances of life, prestidigitation, and triangles.
    However, I could also live with a view like, “Act-type T is impermissible iff For all act-tokens x, if x is T then x is impermissible.” This uses two distinct though related notions of impermissibility (one for tokens, one for types). It is a separate question which is fundamental.

  11. Campbell —

    1. What is the force of the word ‘in’ in the statement ‘In scratching his nose, Campbell fulfilled his duty not to murder me’? If it is meant to indicate some kind of explanatory relationship, then we can see why this statement sounds defective. The fact that you do nothing other than scratching your nose at time t is indeed sufficient in the circumstances to make it the case that you fulfil your obligation not to murder me at t, but as an explanation of why you fulfilled your obligation not to murder me at t, it seems overloaded with unnecessary detail. The fact that the act-token that you performed at t was an instance of your not murdering me is a much better explanation of why you fulfilled this moral obligation at t than the fact that you scratched your nose at t!
    2. I believe that act-types are general properties (which I think of as universals of a certain kind). We could think of them as properties of a certain type of event (i.e., of act-tokens) or we could think of them as a certain special class of properties that an agent can have at a particular time; I think of that as a house-keeping question rather than a question of substantive philosophical analysis. But the hairy issue of how to think of properties and universals belongs to general metaphysics, not to metaethics, and so I don’t propose to discuss it any further.
  12. Heath and Mark —
    Surely, if I am never murdered, then the statement “Any murdering of Ralph is impermissible” is vacuously true. Unfortunately, the statement “Any murdering of Ralph is praiseworthy and obligatory” is also vacuously true. But surely you wouldn’t accept the statement “Murdering Ralph is praiseworthy and obligatory”?!
    Anyway, I totally agree with Mark’s point that it is grossly implausible to claim that the simple statement “Murdering Ralph is impermissible” has a hidden quantified and modalized structure. (I guess that this point also suggests a promising answer to the act-token theorist’s objection to Mark’s first comment that I mentioned above….) Such butchering of the surface syntax is surely best avoided.

  13. Ralph and Mark,
    What do you want to say about “Killing is wrong” and “Killing is not wrong”? I am not prepared to say that either is true. What I want to do is say, “Well some instances of killing are wrong, and some are not.” But if I say this, it seems to me that I need a concept of wrongness that applies to tokens, not types.
    Furthermore, what I want to say is well explained on the view that a type is wrong/impermissible iff all its tokens are wrong/impermissible. I guess you could also deny bivalence but that seems extreme.
    Finally, I was explicit, above, that “Murdering Ralph is impermissible” is not a modal claim.

  14. Heath —
    You say:

    ‘A type is wrong/impermissible iff all its tokens are wrong/impermissible.’

    But if there are no tokens of a certain type T, isn’t it vacuously true that all tokens of that type are wrong? Given your biconditional, we can infer that T is wrong. But T might be (e.g.) preventing Martin Luther King from being assassinated — which surely isn’t wrong!
    I guess that strictly speaking, ‘Killing is not wrong’ is true. Of course, killing is usually wrong. But killing is not wrong per se, no more than types like (e.g.) retreating from danger or saying things that some people find upsetting are wrong per se. Of course, in many contexts, it would be misleading to assert ‘Killing is not wrong’, but I’m weakly inclined to think that it’s true.

  15. Ralph,
    Several points, followed by a modification of my previous suggestion, which does indeed fall prey to the problem you’ve pointed out.
    First, “Murdering” (mutatis mutandis for other action descriptions) can have a number of different logical roles. It can pick out an individual act-token:
    (1) Murdering Zelda took fifteen minutes
    It can quantify over act-tokens:
    (2) Murdering a head of state has never achieved its political goals
    It can pick out an act-type:
    (3) Murdering is a species of killing.
    An initial argument that “wrong” (mutatis mutandis for other properties) can apply to either act-types or tokens:
    (4) Murdering Zelda took fifteen minutes. It was wrong.
    (5) Murdering is a species of killing, and it’s wrong.
    How to tell the difference? When we’re talking about act-types, we’ll use tenseless, non-modal verbs, as in (3) or (5) or simply
    (6) Murdering Zelda is wrong.
    Tenses, or modalities, indicate we’re talking about tokens, as in (1) or (2) or (4) or
    (7) Murdering Zelda was wrong
    (8) Murdering Zelda would be wrong.
    In (7) and (8), we can’t very well be saying that act-types used to have properties they now lack, or have properties in nearby possible worlds that they lack in the actual world.
    To make this point further: Suppose I owe Zelda $100 and don’t have much more than that. I might say
    (9) Spending $100 on dinner would be wrong.
    I don’t mean that the act-type spending $100 on dinner is wrong, since there are cases in which I wouldn’t object. I mean that in nearby possible worlds, any act-token with that description is wrong. But suppose I go ahead and splurge on myself. Later, I repent, but I also get my paycheck. I say
    (10) Spending $100 on dinner was wrong back then, but now it would be okay.
    Again, I can hardly be talking about the shifting properties of the act-type. It’s the tokens I’m talking about.
    So here’s my new suggestion, which may not work, for some reasons discussed below:

    (S) An act-type T is wrong iff . Necessarily, for all act-tokens x, if x is T then x is wrong.

    (S) has the consequence that (6) is true, and
    (11) Murdering Zelda is not wrong
    is false, even if Zelda is never actually murdered. It also has the consequence that both “Killing is wrong” and “Killing is not wrong” are false, assuming that some possible killings are wrong and some are not.
    With respect to the arguments in the post, the first is that we have negative obligations. So things like
    (12) Not murdering Zelda is obligatory
    are true. This doesn’t seem to fit with (S), since “Necessarily, if x is not murdering Zelda then x is obligatory” is much too stringent. But obligatory omissions are ontologically peculiar anyway, and if we can massage (12) into
    (13)Murdering Zelda is prohibited
    then “Necessarily, if x is murdering Zelda then x is prohibited” works fine. And then we can say, by (S), that (12) and (13) are predications on act-types, as you initially suggested.
    The second argument is that we can have positive obligations which can be fulfilled in any number of ways. So things like
    (14) Me paying you $20 before Saturday is obligatory.
    This doesn’t initially seem to fit with (S) either, since “Necessarily, if x is me paying you $20 before Saturday, x is obligatory” makes me give you lots more money than I should. However if we can spell out (14) into
    (15)Me paying you a total of $20 before Saturday is obligatory, or
    (16)Me paying you $20 exactly once before Saturday is obligatory
    then (S) works much better. And again, this has the consequence that there is a reading of(14)-(16) which is a predication on act-types. I am less confident here, though, because it seems that what we want is something more like “It is obligatory that: (Ex) x is me paying you $20 before Saturday.”

  16. Thanks Ralph!
    so, “the act that we’re obliged to do is always an act-type” becomes “the act that we’re obliged to do is to behave in certain general kind of ways” by substituting from “All that I mean by an “act-type” is one of these perfectly familiar kinds of behaviour”. That sounds fine to me but not very informative.
    I find it somewhat strange that you set metaphysical questions aside and concentrate on metaethical ones when clearly we are talking about metaphysics of actions, types and tokens. You say that it is crazy to think that there are merely possible token-actions but you are happy with universals and general properties that can exist uninstantiated. Maybe your view in metaphysics is more plausible but prima facie it doesn’t seem less crazy or extravagant.
    Finally, I don’t think my proposal requires disjunctive things or events. For instance, I could say that ‘Obligation to’ is a propositional operator that takes propositions under its scope. These propositions can be disjunctions of descriptions of possible action-tokens when the operator has a wide scope. I can then give the semantics of this by talking about worlds in which these tokens take place belonging to deontically ideal worlds. I don’t think this story requires disjunctive events per se. I believe you have to say something similar unless you think there are disjunctive properties (in cases like the one in which you are obliged to either buy Bill a drink or to come home early).
    Anyway, the main point just was that I’m not sure the two arguments in the original post quite yet rule out the token-based view if we develop it a bit. You are right that there may be metaphysical costs to this down the line.

  17. Heath —
    You’re obviously right that gerundial constructions like ‘Murdering people’ and ‘Murdering Zelda’ have several different logical and semantic functions.
    Your new suggestion, though, has two problems. First, as Mark pointed out above, importing a quantified and modalized structure into ‘You are morally obliged not to murder Zelda’ seems implausible.
    Secondly, you want to treat ‘… is obligatory’ as a monadic predicate, which can apply both to act-tokens and to act-types. But at least when it applies to act-types, in the cases that you’re focusing on, it seems to me that it really stands for a relation between an agent, a time, and an act (hence the construction “It is obligatory for x to A at t“); I don’t see how to make sense of the cases that you’re focusing on as simply predicating obligatoriness of an act-type.
    Jussi —
    I don’t really need to commit myself on the question of the metaphysical nature of act-types. (I should have made this point clear in responding to Campbell above.) I think that everything that I’m saying could be adapted to any interpretation of what act-types are — even the sort of nominalist view that e.g. Davidson would favour, according to which act-types are really just descriptions of act-tokens.
    My point is simply that the facts about moral obligation cannot be captured by a set of propositions that predicate the property of obligatoriness of act-tokens. We crucially need to include propositions that in some way attach obligatoriness to act-types in relation to the situation of an agent at a time.
    So it seems to me that what you call the “token-based view” is not in fact the view that I was disagreeing with at all, if your “token-based” view implies that ‘Obligation’ is a “propositional operator” attaching to “descriptions of possible act-tokens”. My own view is that an “obligation” is really a relation between an agent, an act-type, and a situation or occasion. But I do think that ‘Ought’ is a propositional operator, and I have no objection to interpreting ‘You ought to leave early’ as attaching the ‘Ought’ operator to the proposition ‘There is an act-token of which you are the agent belonging to the act-type leaving early‘.
    Christian —
    I’m not objecting to consequentialism in my post — that’s true. But if a consequentialist were to say that moral obligation always applies to act-tokens and never to act-types, then my objections would indeed have force against this kind of consequentialist.

  18. Ralph,
    We can consider a form of Consequentialism that does say that moral obligation always applies to act-tokens. But I was being quick. Here’s a brief explanation of why I think that even if you’re right, you’re objection does not have force against this form of Consequentialism.
    Here’s a thesis:
    For any agent S, time t, and action A, S has a moral obligation to perform a token act A at t iff the consequences of S’s performing A at t are at least as good as the consequences of S’s performing any other act-token that S can perform at t.
    How do your objections run against this view?
    First, if performing an act is a way of causing an outcome and omissions are not causes, then the objection from omissions doesn’t apply. I’ve just stated Consequentialism in a way that entails that there are no basic negative obligations. A requirement just is a requirement to do a particular thing, even if satisfying this requirement would entail that other things would not be done.
    Second, I’m supposing that there is only one way to satisfy an obligation to do what is best. That is, either there is exactly one act-token you can perform that has the best consequences and you should perform that one, or else, there are tied acts for the “best” and you are required to perform either of the tied act-tokens.
    If this version of Consequentialism is correct, then I don’t see how your objections have force against it. Moreover, I think this version of Consequentialism is independently pretty plausible.
    I have another worry though. I think ‘ought’ implies ‘can’ but I don’t think someone can perform an act-type. We can only perform act-tokens. So, obligations apply to act-tokens and not act-types. This suggests, I think, that we are obliged to perform act-tokens of a given type.

  19. Christian —

    1. This form of consequentialism is still open to the problem that Mark Schroeder pointed to above. Since an act-token only exists if the act is actually performed, if I fail to do what I am obliged to do, there is no act-token for me to be obliged to do — there is only the act-type.
    2. Even if you are right there are no “basic” negative obligations, there obviously are negative obligations. (E.g. according to consequentialists, I am obliged not to do anything suboptimal.) These cannot be explained on the hypothesis that moral obligations apply to act-tokens, not act-types.
    3. If as you say, I am obliged “to perform either of the tied act-tokens”, then the obligatoriness does not apply to any particular act-token: it is an obligation to do some act-token or other that belongs to a certain set — or as we might say, a certain type.
    4. In general, I’m not objecting to a Davidsonian view of the logical form of action-sentences here. Perhaps action-sentences do always involve quantifying over act-tokens (e.g. “I signed the cheque” = “There is an act-token of which I was the agent, and which was of the cheque-signing type”). Still, this view does not support the view that obligations apply directly to particular act-tokens: if anything, it supports the view that obligations apply to propositions that quantify over act-tokens….
  20. This is something of a side issue, but as a matter of curiosity do you then think that moral obligations be inconsistent? On an act-token view it is impossible for obligations to be inconsistent — an act-token is either obligatory or not. But on an act-type view it seems possible that we could have inconsistent obligations. It would depend, I suppose, on your theory of act-types, but if an act-token can be the token two different act-types (and, at least in the way we normally think of these things they can, e.g., act-types tend to be treated as corresponding to ways acts can full under descriptions, and acts can fall under many different descriptions in many different ways) then, at least prima facie, moral obligations can be inconsistent with each other.

  21. Ralph, Chisholm says some overlapping things about paying a debt in his “Supererogation and Offense” (in footnote 6), but see esp. Stocker’s “Supererogation and Duties”:
    “We fulfill duties by performing acts–i.e. by performing act tokens, for we cannot perform act types. Nonetheless, it is not a duty to perform any act token. For we could have fulfilled the duty by doing another act token of the appropriate type. For example, … …What is it our duty to do since it is not our duty to do any given act token?…
    … The answer …it is simply our duty *to exemplify act types*….” [my stress]
    This may not be just your view, but it has affinities. Both he and Chisholm are discussing issues pertaining to perfect, imperfect duties, and supererogation, Chisholm in passing, Stocker more centrally to his piece.
    (I suspect even Stocker’s proposal is too narrow–too act-oriented, albeit via act types. We have obligations to be certain ways (e.g. to be in our offices from t1-t2 Mons and Weds this semester), from which we derive obligations to bring it about that we are present in certain locations, etc. We might try to argue this is all disguised agential obligation talk, but on its face, we commit all the time to being certain ways-satisfying certain properties (“I’ll be at home Thursday for the delivery”). Call the resulting obligations personal obligations, obligations to be such that p, for some p. We can then think of agential obligations as a special case: personal obligations to be such that we are agential in some respect (my obligation to be such that I bring it about that I’ll be home Thursday for the delivery).

  22. Hi, Ralph,
    Interesting discussion. I was wondering whether you could say more in response to Heath, since I’m not quite following how your responses are supposed to go, since,
    1. I don’t see how anything he says is incompatible with ‘…is obligatory’ being relational
    2. I don’t see Mark as saying anything incompatible with prohibition/obligation sometimes attaching to act-tokens. I took him to be giving an argument that they at least sometimes attach to types, which is compatible with what Heath claims.
    3. I don’t see Mark arguing that “importing a quantified and modalized structure into ‘You are morally obliged not to murder Zelda'” is “implausible”. I only see him asking whether Heath thought the contested sentences had that structure.

  23. Ralph:
    “This form of consequentialism is still open to the problem that Mark Schroeder pointed to above.”
    Yes, that is an interesting problem. I’m not sure what to say, though I think that something funny is going on with the relevant construction. Consider:
    (Batman) I failed to see Batman.
    From (Batman) we can’t infer:
    (Crazy) There is an x such that x = Batman, and I failed to see x.
    The same appears to go for others failures like a failure to meet an obligation.
    (Failed Obligation) I failed to perform obligatory act-token A.
    From (Failed Obligation) we can’t infer:
    (Schroeder) There is an x such that x = obligatory act-token A, and I failed to perform A.
    The lesson I’m inclined to draw, though not confidently, is that ‘failure’ contexts are opaque and so they don’t raise a problem for the version of Consequentialism I mentioned above.
    “If as you say, I am obliged “to perform either of the tied act-tokens”, then the obligatoriness does not apply to any particular act-token: it is an obligation to do some act-token or other that belongs to a certain set — or as we might say, a certain type.”
    I want to resist the inference you mention. Suppose I can pick up either a glass of white or a glass of red, but not both. I don’t think it follows that I can only pick up a set or a type. I can pick up one of two particular glasses of wine. I say the same goes for obligations. If I’m obliged to do A or B, then I’m obliged to perform one of two particular act-tokens. But that’s basically just asserting my claim. Do you think we do perform act-tokens or only act-types?
    “Still, this view does not support the view that obligations apply directly to particular act-tokens: if anything, it supports the view that obligations apply to propositions that quantify over act-tokens.”
    I think we disagree over the basis for the obligations. I want to say we have obligations to perform an act-type in virtue of having an obligation to perform an act-token that is of that type. You appear to want the direction of determination to go the other way. Here’s another worry: Determinables are grounded in their determinates. An act-type is a determinable of which an act-token is its determinate. Thus, we should think that obligations to perform act-types are grounded in obligations to perform act-tokens. An analogy with vision: I can see colors in virtue of being able to see particular shades of color.
    I’m still thinking about your second point. I’m not sure what to say to it right now.

  24. Brandon —
    I distinguish moral obligations from all-out moral requirements. In my view, obligations can conflict, but moral requirements cannot. E.g. promissory obligations can conflict (suppose that you simultaneously promise to marry both Chris and Jo…). In effect, obligations are pro tanto moral reasons for action, of a certain kind — they do not always generate an all-out moral requirement.
    Janice —

    1. Perhaps you’re right that I’ve misread Heath on this point. Indeed, towards the end of his last comment, he seems to think that the way to capture the relational character of obligations was by building the agent and time into the act-type itself (which is supposedly what is referred to by the phrase “Me paying you a total of £20 before Saturday”), and then taking obligatoriness to be predicated of this complex act-type. I find it slightly unnatural to call this sort of thing an act-type (it seems more like a possible state of affairs to me), but if he takes that route, I wouldn’t really have any objections on this score.
    2. It seems more promising to me to go for a relatively unified theory of obligation; so I think it’s more plausible in the end that obligation always applies to act-types. But there are some closely related moral terms — like ‘wrong’ — that clearly do apply to act-tokens as well as to act-types. So you might be able to convince me to take a similar approach to “obligation” as well.
    3. I assumed that Mark’s last comment was a rhetorical question! I’m pretty sure that I’m right about this….

    Paul —
    Thanks for your references to Chisholm and Stocker — I must read them when I get the chance!
    I wonder whether you’re right that interpreting all obligations as applying to act-types is too narrow. After all, I have an obligation to be in Ireland next week; but being in Ireland is not strictly an act-type. Perhaps we need a more general theory, according to which every obligation is a relation between an agent and a property (which might be an act-type or a property of some other kind). At all events, this would take us even further away from the view that obligations apply to act-tokens.
    Christian —
    Nothing in the point that I made hinges on contexts that involve ‘fail to…’. The point is quite simple. You were obliged not to murder Zelda, but you did murder Zelda. So there is no act-token of your not murdering Zelda to have the property of obligatoriness!
    The rest of your comment suggests that you may possibly be misunderstanding me. The phrase “performing act-type T” is a weird use of our semi-technical philosophical English, but presumably it’s equivalent to “performing an act-token of type T“. So I am not denying that you ever have an obligation to perform-an-act-token-of-a-particular-type (where the quantification over act-tokens is inside the scope of ‘have an obligation to…’). What I’m denying is that there is any particular act-token such that you have an obligation to do it.

  25. Ralph:
    “What I’m denying is that there is any particular act-token such that you have an obligation to do it.”
    I did indeed misread your view! I thought you were claiming that we are obliged only to perform act-types. You said: “I’ve been surprised to find that several moral philosophers believe that when we are morally obliged to do an act, this “act” that we’re obliged to do is an act-token, not an act-type. In my view, this is confused: the act that we’re obliged to do is always an act-type.”
    If I’m interpreting you correctly now, you’re saying that we only have obligations to perform act-tokens of a given type, but never a particular act-token of that given type. Is this the right interpretation? Are you okay with the claim that we are obliged to perform an act-token, but never a particular act-token?
    “The point is quite simple. You were obliged not to murder Zelda, but you did murder Zelda. So there is no act-token of your not murdering Zelda to have the property of obligatoriness!”
    This goes back to negative obligations. Maybe there are such things. I don’t know what to say about these purported obligations exactly. I think this is really hard. But my act-token of killing Zelda has the property of being wrong (or ought-not-to-be-doneness). Maybe there simply isn’t an act-token of my non-murdering-of-Zelda that has some property, but I’m having a hard time seeing why this matters. Let me put this differently:
    You said “The point is quite simple. You were obliged not to murder Zelda, but you did murder Zelda.”
    And I agree. But you then said:
    “So there is no act-token of your not murdering Zelda to have the property of obligatoriness!
    I don’t see how this follows, but maybe it does. Suppose it does. How does this support the view that we are not obliged to perform a particular token-act? I’m having a very difficult time seeing the rationale for the inference here. Apologies if I’m missing something obvious. This is a hard issue that is well-worth working through, I think.

  26. Ralph,
    Re “I wonder whether you’re right that interpreting all obligations as applying to act-types is too narrow. After all, I have an obligation to be in Ireland next week; but being in Ireland is not strictly an act-type. Perhaps we need a more general theory, according to which every obligation is a relation between an agent and a property (which might be an act-type or a property of some other kind).”
    That is just what I meant by saying act types are too narrow–being such that I am in my office [or at home for the delivery] is a property, but not an act type, yet I can be obligated to be in my office or at home for the delivery, so act types would be too narrow as the sole obligation complements.
    Regarding never being obligated to perform an act token, perhaps a key thesis here. Consider the limiting case where the obligation to exemplify an act type might converge on a particular act token–the hypothetical sort of case where a person’s remaining options/circumstances are so limited and control so narrow that they can exemplify an act of type A iff they perform act token x (assume they do), the only token of act type A they can still perform. If what they must do is exemplify A (say they promised) and they can only do so by performing token x (their last chance), then derivatively, they must [i.e. its obligatory to] perform token x. This might suggest that there is nothing wrong generally with deriving from ordinary obligations to exemplify act types (or more generally certain properties), derivative obligations to the effect that you must perform a certain act token, x (albeit invariably in real life cases, it will be to perform some member of some class of act tokens or some member of some class of life-long sequences of act tokens for life-time obligations, like not murdering you, which I cannot fulfill until I am no longer able to murder you.

  27. Ralph,
    I’ll let Heath explain whether he meant it, but I think there is in fact a problem with supposing that my paying you a total of £20 before Saturday is an act type (or whatever sort of thing turns out to be obligatory) with the agent built into it, in lieu of holding that obligation is relational (which I’d think meant it has two argument places). The problem is that we might well want equivalent act-types to be both obligatory or both not. One might even think that equivalent (nominalized) sentences pick out the same act type with different words.
    But here’s an act-type that is equivalent to the given one (in the sense that it is a conceptual necessity that the one occurs iff the other does): your being paid a total of £20 by me before Saturday. But the whole point of the relational approach is that the former might be obligatory even when the latter is not; that is (just to be pedantic about it), the onus of obligation might fall entirely upon me and not on you, no matter which of us is the subject of the (nominalized) sentence.

  28. Christian —
    I’m assuming, with almost everyone else, that an act-token only exists if it is actually performed. So if you’re obliged to do A, but you never actually do A, then you never perform an act-token of A-ing, and so there simply does not act-token of your A-ing.
    Paul —
    I basically agree with everything you say. In the interesting and unusual cases that you’re thinking about, I’d rather say that the obligation is to do something of an unusually narrow and specific type — so narrow that as a matter of fact any two of the “practically feasible” possible worlds in which you do something of that type will be worlds where you perform the very same act-token. (If you insist on saying that in these cases there’s an obligation to do that act-token, then I think you’re extending the notion of ‘obligation’ in a way that serves no very obvious theoretical purpose…)
    Jamie —
    Of course you’re right. (It’s even part of my official view that obligations and reasons and oughts are always indexed to agents. How alarming that that could slip my mind, even for a minute…)

  29. Ralph:
    I’m assuming, with almost everyone else, that an act-token only exists if it is actually performed. So if you’re obliged to do A, but you never actually do A, then you never perform an act-token of A-ing, and so there simply does not act-token of your A-ing.
    This is uncontroversial. What I’m challenging is the idea that this kind of reasoning applies to “negative act-tokens” whatever those might be. In the Zelda case, I murdered her. There was a positive act-token, a murdering, that has the property of being impermissible. I don’t see why there is a need to appeal to a negative act-token, a “non-murdering of Zelda” to explain why I did something impermissible. This is why I quoted:
    So there is no act-token of your not murdering Zelda to have the property of obligatoriness!
    and then asked why this point matters. It seems to me that we can explain the impermissibilty of a murdering without appealing to the failing to obtain of a non-murdering. We only need to appeal to the obtaining of an actual murdering and the fact that, when this token-act obtains, it’s wrong.

  30. Hi Ralph,
    Assuming act tokens are needed for lots of other things, there is no ontological parsimony issue here. Christian, and perhaps others, have made the point that it sure seems right that act tokens will have sister properties of being permissible, impermissible, optional, beyond the call, just as they seem to have axiological properties like being good, bad, better than some others, etc. So it looks like it is only *very hard* to get act tokens to be obligatory for the reasons Stocker and Chisholm pointed out: way too many ways typically to fulfill an obligation for any one of them to be what must be done [even here note that stating this seems to embrace the coherence of the application]; but for the reasons I pointed out it looks like it is possible that there is an obligation to perform an act of type A that can only be fulfilled if you perform a singular concrete act x, and you agree. It also looks like there is an at least defeasible principle that you must do those things the doing of which are necessary for doing what is required. But in the imagined case, you can’t exemplify act A unless you do x in particular. …You can still go on to say that any normative property of any concrete act token is had in virtue of the act token’s being instances of act types that have various morally relevant properties, so that there is an order of priority here, but I’m still not persuaded to stop short of letting concrete acts derive all the normative and axiological properties mentioned above.

  31. Christian —
    I have claimed in this thread that there are three problems for the view that obligatoriness is just a property of particular act-tokens: (i) negative obligations; (ii) unspecific positive obligations; and (iii) unfulfilled obligations (this last is Mark Schroeder’s problem). The example that I gave in my last post (your failing to fulfil your obligation not to murder Zelda) was — perhaps confusingly — an instance of both (i) and (iii).
    Now, you seem to be saying that what is explanatorily fundamental about this case is the impermissibility of the act-token of your murdering Zelda. But in this discussion, I didn’t really want to get into the question of what is explanatorily more fundamental, and what is explanatorily less fundamental. Even if it is not “explanatorily fundamental”, it is still true that you are obliged not to murder Zelda, and that obligation does not consist in the obligatoriness of any particular act-token. I take it that you would actually you agree with me about this point, wouldn’t you?
    Paul —
    Surely, these “sister properties” — the obligatoriness (or permissibility or impermissibility) of an act-token, and the obligatoriness (or permissibility or impermissibility) of an act-type, in relation to an agent x and a time t — are different properties, aren’t they? (If you deny this, you seem to me to bein effect ascribing these properties a quite implausibly disjunctive or “gerrymandered” character.) So I don’t believe that we can allow that act-tokens “derive” the very same properties, as you put it…
    Admittedly, there is a good question about which is more fundamental — the properties of the act-token or the properties of the act-type. As I said in my reply to Christian, I didn’t really want to get into this question here. As a matter of fact, I’m convinced that it is the properties of the act-types that are fundamental, not the act-tokens. But that issue should probably be postponed for another PEA Soup discussion….

  32. I have an official view close to Ralph’s, and it kind of alarmed me too when I discovered I had not been referencing it in this debate. And it’s set up to include Jamie’s observation that ‘oughts’ apply to agents.
    My initial doubt about Ralph’s post was not the “apply to act-types” part, but the “not act-tokens” part. It seems to me that once we have done an act, we can truly describe that act as permitted, obligatory, prohibited, etc. And then I’ve been groping around for a way to glue the bit about types and the bit about tokens together. In his latest reply, this appears to be pretty much what Ralph believes too.
    Here is a way, then, to begin to think about the “which is explanatorily fundamental” question. Consider another non-factive propositional modifier, hoping: “I hope that my brother comes to visit.” We could say that what I hope for is an event-type, since the event doesn’t yet exist, may never exist, many different events could satisfy it, etc. However, once my brother arrives, we could also say that a hoped-for event has occurred. All these seem correct to me.
    An argument: “I hope that p” is a hope for a particular event iff ‘p’ describes exactly one possible world. Normally, of course, our hopes are considerably less precise than that. Then whether hopes-for-p are fundamentally about event-types or event-tokens reduces to the question of whether possible worlds are to be understood in terms of propositions (hoped-for event-tokens are maximally specific hoped-for event-types), or propositions in terms of possible worlds (hoped-for event types quantify over hoped-for event-tokens). I could easily be persuaded that this is not a deep question.

  33. Ralph,
    Suppose it is obligatory for me that I raise my finger [say at/in t if a particular instant or interval is desired], let’s suppose because I promised I would. Now suppose that as it turns out it is unalterable for me that I raise my finger [at t] iff I perform act token x. I think it follows from this that it is obligatory for me that I perform act token x. Suppose it is permissible for me that I raise my finger [at t] and [unalterably] that will occur iff I perform act token x, then it is permissible (for me) that I perform act token x (at t). If it is impermissible that I raise my finger and that will occur iff I perform act token x, then it is impermissible that I perform act token x. These are inferences I think should go through (defeasibly is enough). Does this entail gerrymandering or a problematic account of deontic notions? Suppose we say that to say an act type A is obligatory/permissible/impermissible for me at t [ignore ambiguity of “at t” here–the doing time or time of binding] is to say something roughly along the lines of it is obligatory/permissible/impermissible for me that I perform an act of type A at t, and to say an act token is obligatory/permissible/impermissible for me is to say something along the lines of it is obligatory/permissible/impermissible for me that I perform that very act token.
    Although we can debate the logical form issues involved, something along this lines escapes the claim of obviously problematic or gerrymandered notions, I think. The fundamental deontic statuses above express relations between a person and agential propositions/states of affairs; act types and tokens would be dealt with derivatively. This is at least a position one might reasonably explore and it has the virtue for me of allowing the sort of inferences I’m suggesting are pre-theoretically plausible.
    Thanks for leading a stimulating discussion, and my apologies if I have overlooked some earlier point made.

  34. Ralph:
    I think now that I was running together quite a number of very different claims without slowing down to explain them.
    Even if it is not “explanatorily fundamental”, it is still true that you are obliged not to murder Zelda, and that obligation does not consist in the obligatoriness of any particular act-token. I take it that you would actually you agree with me about this point, wouldn’t you?
    Yes, I think you are obliged not to murder Zelda. Yes, I think this obligation is not fundamental. But ‘no’ to the last part. I do think either (a) your obligation not to murder Zelda does consist in a positive obligation to perform a particular act-token, or (b) your obligation not to murder Zelda consist in positive obligation to perform a particular act-token from amongst a set, or class of act-tokens. Accepting (b), I think, does not imply that one is obliged to perform an act-type.
    That’s the view. It is meant to extend to all three cases you described, namely, to (i) negative obligations; (ii) unspecific positive obligations; and (iii) unfulfilled obligations.
    There are a number of ways to argue for this picture. I can’t do that satisfactorily here. But here is a line that seem to me to be promising.
    (1) The truthmaker for claims about obligations picked out in (i) – (iii) involve facts about positive fundamental obligations. Thus, you’re obliged not to murder Zelda is made true by the fact that you have a positive fundamental obligation to maximize utility by performing a particular act-token A. This basic obligation to perform A, in conjunction with the empirical facts, makes true this claim about your negative obligation. The idea then is that this picture can be extended to unfulfilled obligations and unspecific obligations.
    I also think there is a problem with respect to individuating act-tokens. If we take a coarse grained view about how to do this, but add that there are different ways to instantiated a particular act-token, then we can save the idea that we have obligations to perform particular act-tokens in some way or other. That would be another way to go, though it’s unclear to me whether this way of going wouldn’t be just a notational variant on the view you are proposing.

  35. Christian (and Ralph), some of the features of this dispute are baffling.
    Obligation is a relation between a person and… something. What is the something?
    Ralph says it’s an act type, Christian says it’s a set of act tokens. Plainly, the set will contain some merely possible act tokens. My bafflement: how can we even distinguish these two views? What is the difference between an act type and a set of possible act tokens (whose members are all and only the tokens of the type in question)?
    A related obscurity for Christian: what does it mean to speak of an act token’s being instantiated? I have always thought that a token is an instantiation of its type. Act tokens are concrete particulars, with a spatiotemporal location. I can’t see how they can be instantiated any more than I can see how Ralph could be instantiated.
    All this is really to say that I suspect there is some terminological muddle surrounding this dispute.

  36. Jamie,
    This is close to the same issue Campbell raised earlier. The answer is that Ralph thinks act-types are like universals, something over and above the set of tokens of that type. So, there is a distinction between the two views, so long as you add commitment to something like universals.
    But Campbell raised another issue which I think I don’t think got a response: what are act types (given that they are something like universals) such that they are not metaphysically suspect, though possible act-tokens are?

  37. I see.
    But Ralph, you said:

    There are at least two different act-tokens each of which would be an instance of your giving me £20 either today or tomorrow.

    I don’t think that was crazy. Do you?

  38. Hi Jamie,
    Obligation is a relation between a person and… something. What is the something?
    I’m not sure this is right (or wrong). Perhaps an obligation is a non-relational property of an act-token. For example, suppose that a particular act-token is a promising. Suppose that what makes it a promising are only intrinsic features to the act. Finally, suppose that being a promising is sufficient to make the act-token prima facie obligatory. If these suppositions are right, my performing this particular act-token promising, by being its agent, has the property of being obligatory in virtue of its intrinsic features, and so, being prima facia obligatory will be an intrinsic property of this act-token and not a relation between a person and something else. I’m not sure that this view is incorrect.
    A related obscurity for Christian: what does it mean to speak of an act token’s being instantiated?
    This question does sound strange. What I meant was simply that by performing a particular act one brings into existence (tokens) a particular state of affairs. So, by brushing my teeth I bring into existence (instantiate, in my sense) some particular state of affairs, an act-token. So, I think an act-token is the instantiation of an act-type, but I think it also makes sense to say that one brings into existence an act-token, and thus, to instantiate it. Anyway, the idea of an act-token simply being the instantiation of an act-type seems to be consistent with the idea of an agent instantiating an act-token which is the instantiation of an act-type.

  39. Christian,
    Whether there is any serious problem with thinking of obligation as a property of acts (instead of a relation between agent and act) depends on what you think acts are. Suppose acts just are (or have as much structure as) pairs. Then there is no problem; a monadic property of pairs is the same thing as a relation between agent and event. Suppose acts do not have that much structure; they are just the event, a concrete particular. Then there is a big problem with taking obligation to be a property of acts, namely, the one I mentioned earlier.
    But my real question was supposed to be the other one: what is at issue between someone who thinks a relatum (or bearer of a property, whatever) is a set of act tokens and someone who thinks it is an act type?

  40. Ralph,
    Couldn’t it be the case that there is a set of act-tokens, not generalizable to an act-type, such that your obligation is to do at least one of them? If this is coherent, then it would take care of your debt scenario. You have a very particularist kind of obligation, to pay the guy back by a certain date. There are just a bunch of ways you can do it. But it’s not clear this would commit us to an act-type.
    But maybe it would, and I’m not making sense.
    Joshua

  41. I’m sorry I haven’t been checking PEA Soup for the last couple of days. (I’m sure you all know how it is….)
    I may try to answer some of these more recent comments in more detail. But for the time being, I thought that I would just make the following point.
    There are two issues that we should separate. One of them is metaethical; the other is more metaphysical:

    1. Metaethical: Obligation is a relation between an agent and what? Pretty well everyone agrees that it’s not a relation between an agent and a particular act-token; it must be a relation between an agent and something more “general” or “type-like” in nature.
    2. Metaphysical: What is the correct theory of the nature of this more “general” or “type-like” entity? Is it a set of (actual and non-actual) possible act-tokens? Or a universal? Or what? As a matter of fact, I am strongly in favour of an actualist Platonist metaphysics, and so I am firmly against quantifying over non-actual possibilia, while I am quite happy (indeed, enthusiastic) about quantifying over universals. Still, this seems to be a metaphysical issue rather than a metaethical one.

    Incidentally, Jamie, when I said

    There are at least two different act-tokens each of which would be an instance of your giving me £20 either today or tomorrow

    I think I was only talking about what was true in world w3 where you give me £20 twice, once on each of the two days.

  42. Jamie,
    I thought I posted a response to your question a few days ago. I may have forgotten to push the “post” icon.
    …what is at issue between someone who thinks a relatum (or bearer of a property, whatever) is a set of act tokens and someone who thinks it is an act type?
    If you want value theory to be consistent with the right metaphysics, it matters. I myself think neither of these views is plausible. I don’t think we are obliged to perform either sets or types. We cannot perform sets or types, and yet, we are obliged to do things. Thus, those things are neither sets nor types.

  43. On many views (including nominalist views), types are abstracta. Ralph, can you explain how you can be obligated to perform an abstractum? I don’t even think that question makes much sense. It seems sensible to say that we are obligated to perform some token-action of a particular type-action. For instance, if I am obligated to save the child from drowning in the shallow pond, I am obligated to perform a token-action of that the type-[save the drowning child]. It is a familiar fact from action theory that many token-actions can be of that type. And I might not be obligated to perform any particular token of the type but I am obligated to perform SOME token of the type.
    Why isn’t it just incoherent to say that we are obligated to perform a type?

Comments are closed.