The Power to be Free

I’ve often thought that the classic debates about free will suffer from a conflation of causation and necessitation. More generally, it’s often seemed to me that without a generally satisfying theory of causation, we have no reason to be worried that free will is a mere illusion. Since the reality of rational agency — where ‘reality’ is a matter of our reasoning controlling what we do in a way that cannot be explained entirely in terms of non-rational antecedent causes — seems like a necessary condition of our being able to formulate scientific and philosophical theories that stand a decent chance of being true, we would seem to have powerful reason to be skeptical at best of any philosophical or scientific theory that denies the reality of rational agency (this may be the one thing I think Kant got roughly right). When we notice, in addition to this consideration, that philosophers and scientists have arrived at no generally accepted theory of causation, and that many have even gone so far as to deny that science has any need for the concept of a cause at all, it seems even less sensible to get worked up about the possibility that maybe everything each of us does is determined entirely by antecedent causal factors over which our reasoning has no influence whatsoever.

Happily, I’ve just stumbled across a recent paper by Rani Lill Anjum and Stephen Mumford that argues explicitly that we ought to distinguish between causation and necessitation. Unlike some earlier arguments for that claim that I’ve encountered, however, this one sketches out a theory of causation on which we would need to distinguish causation from necessitation quite generally, and not merely as a logical possibility that may, for all we know, be actual in cases of radioactive decay and rational agency. Anjum and Mumford also nicely show how we need not be led to any kind of wild causal indeterminism either. In fact, one of their main claims is that rational agency needs causation, and that we think about free will wrongly if we think that what we need is to escape from causation. What we need instead is to escape from both necessity and randomness, and we can do that, they think, if we adopt a conception of causation as grounded in dispositional tendencies to produce effects. Causation is grounded in causal powers, and rational agency is one special kind of causal power — special in enabling free will, but not special, on their view, in transcending necessitation.

I’m not sure if Anjum and Mumford are right; I’m especially not sure whether their theory meets the objection that they raise against Aristotle and Aquinas, viz. that they resolve dispositional tendencies into conditional necessities (I’m also not sure that Aristotle and Aquinas fall to the objection, or whether it would be a serious problem for them if they did). But I’m intrigued by their account, which is the first thing I’ve read that not only attempts to sever the link between causation and necessitation, but lays out a plausible theory of causation that is not supposed to apply solely to some special class of entities like rational agents or radioactive material.

PoT readers who know the free will and causation literatures better than I do may not be so intrigued or surprised, but it seemed like a safe bet that at least some of you would share my interest.

You can read Anjum and Mumford’s paper, “Causation is Not Your Enemy,” here:

22 thoughts on “The Power to be Free

  1. I’ve only browsed, not read, the Anjum-Mumford paper, which looks really interesting. My friend Rick Minto defended a similar thesis in a doctoral dissertation he did in HPS at the University of Western Ontario back in 1997, Foundations for a Realist Theory of Causality–same distinction between necessity and causation. But doesn’t Timothy O’Connor make essentially the same distinction in his defense of agent causal libertarianism?

    PS. Just realized that the link above goes to an abstract of Rick’s dissertation, not the diss. itself. This one does, but it’s a big PDF (290 pages).


  2. I’ve looked at Minto’s dissertation, but it was a while ago; what I dimly remember, though, is that he sees all non-rational causation as deterministic. O’Connor’s view, such as I understand it, has the same feature, but also follows Chisholm and Taylor in regarding agent causation as a distinct sort of causation found only in rational beings. I prefer Minto’s view (which is, I take it, a developed version of the “entity-causation” view sketched in Eyal Mozes’ review of Harris in Reason Papers 35.1) in part because it seems odd to treat most causation on an event model and to treat agent causation as fundamentally different and applicable only to mental causation (not least because there are lots of physical things relevant to our actions, and if event causation is what is supposed to apply to the non-mental, it should apply there, too, in which case it would seem unclear why freedom extends to what we do instead of merely what we think or decide), but primarily because it seems to me that an entity- or agent-causal theory makes better sense even when we’re talking about non-living things (indeed, as Roderick Long argues somewhere in his dissertation, arguably the only coherent way to individuate events is by analyzing them in terms of substances manifesting properties at times — and for what it’s worth, Long’s Aristotle thinks that all non-rational agency is deterministic qua necessitating, as well). But all of these theories, so far as I understand them, think that causation is usually deterministic in the sense that causes necessitate their effects. If A&M are right, then causes don’t necessitate their effects in general, so the relatively greater scope of freedom from necessity that we seem to find in rational agency will not be some kind of mysterious exception, but simply a difference in the degree to which rational beings have something that all causal entities have, a difference that seems plausibly explained in terms of ordinary features of the capacity for practical rationality. I don’t find objections from human uniqueness very compelling (often they seem to be nothing more than textbook cases of begging the question), but if necessity applies to everything else in the world above the level of the quantum, it seems at least a bit fishy to suppose that human beings somehow escape from that necessity.


    • Yes on O’Connor, but this is what Minto says (I’m copying and pasting from a PDF, so apologies for any loss of formatting I missed):

      First, entities are profitably regarded as falling somewhere on a spectrum of passivity whose ends are represented on one end by the spherical inelastic billiard ball, and on the other by the rational agent entities for which the “same cause, same effect” principle is roughly true in practice will tend to fa11 near the passive end of the spectrum. Second, and closely related to the first, is the matter of the ability of entities to store and transform energy for the purpose of replication. Those that can do so exhibit novel teleological modes of action (e.g., biological processes) that cannot be accounted for solely by mechanical principles. Third, realists must be open to the possibility that no true, general version of the “same cause, same effect” prînciple can be defended on logical or conceptual grounds. It will likely turn out to be the case that thé extent to which an entity of a given kind behaves deterministically or indeterministically is a function of the persistent and enduring aspects of the entity’s composition and structure, and is an empirîcal matter. Entities that satisfy the determinists’ preferences for mathematically tractable, passive, law-governed motion should not be regarded as ontologically privileged. Similarly, the electrons, persons, and thunderstorms should not be ontologically marginalized. (pp. 253-54)

      This is in the context of a denial that causation entails deterministic necessitation in chapter 5. In other words, he wants what he calls “causal realism” without a general commitment to necessitation, not just in the case of agency, but generally.

      Anyway, that’s how I remember his view; I haven’t read the dissertation in at least a decade, and have just gone in to hunt and peck for confirmation of my vague memories. So I may have forgotten the wider context of his view, but I think this is right.


      • Ok, good. Thanks. That allows for a looser relationship between causation and necessitation than I remember him taking. It does, though, differ non-trivially from A&M’s view. Minto observes that a cause doesn’t have to necessitate its effect, but he still seems to think that’s usually the case, and that whether causation entails necessity depends on the kind of causal agent involved. A&M think it’s never the case that causes necessitate effects, regardless of the kind of causal agent involved. There might also be another, more subtle difference. Given Minto’s language here, he might be taken simply to reject the same-cause-same-effect principle, but to maintain that when A causes B, it necessitates it. If so, that’s again non-trivially different from A&M’s view; they deny that even the stone that causes the window to shatter necessitates the shattering, even when it does in fact cause the shattering. I may be wrong to read Minto that way, and I may be making too much hay out of what A&M say. But in any case, their separation of causation from necessity seems more radical, and for that reason seems to have the virtue I gestured toward initially — the link between causation and necessitation isn’t being severed for some special and rare entities, and so there is nothing whatsoever mysterious about that severance in the cases of electrons, persons, and thunderstorms.

        In general, though, I think Minto’s view is a member of the family of views that Mumford, in particular, has brought into prominence in contemporary analytic philosophy. I’ll certainly take Minto over most other views on the subject that I’ve encountered!

        Liked by 1 person

  3. Thanks for passing along this paper, I’ll take a look.

    I’m sympathetic to A&M’s sort of view. If causation, at the most fundamental, quantum level, involves a kind of bounded indeterminacy, then – at least at a certain level of specificity – a given set of precise conditions always fails to yield precisely the same result. And we can, I think, readily imagine how small-scale differences could have large-scale effects (river goes this way or that, star goes supernova or does not, etc.). So it seems to me that it is quite possible – in more than just a logical sense – that it is not necessitated or determined that this spider or mouse moves right rather left at this obstacle. Or, perhaps somewhat more than analogously, that I “might have done otherwise” in deciding to write this comment!

    How this happens (and how it is possible in a sense of ‘possible’ that is informed by the specific causal mechanisms involved when central nervous systems self-regulate and regulate actions in such cases) seems to me to be a question for science to answer. As is the issue of what the self is that is doing the deciding or choosing.

    I don’t want to say that our concept of free will is entirely non-committal as to what the self and voluntarily-produced action is. There is the “could have done otherwise” element and the “unified self” element. But neither of these elements commit us to metaphysically robust selves that have some special way of influencing an otherwise necessitated causal order (or have some magical way of contravening the causal order itself, whether or not causality necessitates). So it seems unlikely that, when we figure out the science, we will have to substantially revise our concept of free will to the extent that we should say that there is no such thing as free will (i.e., be “eliminativists” about free will).

    On the metaphysical and conceptual issues here, I am sometimes haunted by the idea that I am missing something and should be much more puzzled than I am. So many smart people remain so puzzled. I’ll take a look at the A&M paper for sure.


    • I’m not convinced that considerations from quantum mechanics should play a very big role in our thinking about this stuff. For one thing, it’s a very young and underdeveloped area of science, one that scientists disagree about deeply and in fundamental ways, and so it is virtually guaranteed that it will develop in ways we can’t foresee and that very little of it in its current form deserves to be taken as settled. For another thing, my limited knowledge and understanding of the issues suggests that the issues are often more philosophical than empirical, and while I have no objection to physicists debating the philosophical dimensions of their discipline, I’m not entirely confident that they’ll do it in a philosophically satisfactory way. For a third thing, I’m not sure we should be realists about theoretical physics in the first place. Most basically, though, even if it were conclusively settled tonight that quantum mechanics proves the falsity of universal determinism, that wouldn’t necessarily have any implications at all for anything above the quantum level. So I don’t think it helps, though perhaps it doesn’t hurt.

      I share your feelings about not being mystified to some extent, but I do think that there’s something intuitively compelling about the idea that our only options are causes necessitating their effects or randomness, so I see why the view that A&M ascribe to Dennett at the beginning of the paper is a widespread one. I think Anscombe showed about as conclusively as one can show anything in philosophy that as a conceptual matter, causation and necessitation can come apart. But I also think that that plenty of things that can come apart conceptually cannot in fact come apart as a matter of natural and maybe even metaphysical necessity. So I don’t think Anscombe’s argument does much in the way of showing that determinism qua necessitation is false. Maybe the best way to put it is that I’m not much more puzzled than you are, but I have no trouble seeing why smart people are puzzled.

      Liked by 2 people

  4. Just read the A&M paper – quite good, thanks again for passing that along.

    I agree that conceptually distinguishing causation from necessitation (or, more generally, the making-more-likely modal notion from the determining or necessitating modal notion) does not tell us anything about the sort of causation that exists in our world. However, since at least many causal relationships that we take to exist can be characterized in terms of dispositions, and since dispositions are not or at least need not be necessitating, we should expect that at least many causal relationships are something like making-more-likely-than-not, not necessitating. And so the idea that, in deciding or choosing, we “could have done otherwise” (we face alternative possibilities) does not obviously run afoul of what causation seems to be.

    It does seem funny that there would be no explanation of why (or no reason why) I presently choose to visualize an orange triangle rather than not. (Doing so fits the bill for making my philosophical point about choice, but I need not have chosen to imagine just that. Or so I think!) But it is equally funny though that there is no explanation (or reason why) this unstable atom of uranium decayed at time t1 rather than at time t2 (or so established science, and much good philosophy of science, tells us). Especially when we think of actual cases like this, it becomes obvious that, even if there usually are causal reasons why just this rather than just that happened, it is not incoherent to think that, at least sometimes, there are not. And so the Principle of Sufficient Reason (PSR) is false despite its perhaps initially appearing to be true.

    Absent the quantum-causation cases, I think I would be pretty sharply conflicted. Maybe, on balance, I would choose PSR over the principle of alternative possibilities (PAP). (I guess I’d have to say choose* PSR over PAP, the applicability of the very concept of choice being rejected in the choosing*.) So these cases help me get free of the first horn of the classic dilemma. But I do feel its pull (or poke).

    Here’s why the second horn of the dilemma does not concern me at all (and should not). PAP is incompatible with indeterminism only if we interpret indeterminism in such a way that particular consequent states don’t count as effects of the initial state. Maybe there is this temptation when we paint a picture of “wild” indeterminism, an indeterminism in which may waving my hands might just as well cause a unicorn to appear or the sun to explode as anything else. But an indeterministic relationship in which the initial state produces one of a specific range of consequent states (not just anything) is not as susceptible to the a-causal or anti-causal description. Again, the actual cases of indeterminism help, here: the uranium atom might decay at t1 or even at a distant t2, but it will not turn into an elephant. And, I think partly because I get this at some implicit level, it is not counter-intuitive for me to use the language of causation to say that it is the structure of the uranium atom that caused it to decay at t1 (even though, speaking precisely, its structure did not cause its decaying-at-t1-rather-than-at-t2).

    Regarding the A&M paper…

    (1) The main thing that A&M do (going beyond Anscombe) is account for causation itself in terms of dispositions and analyze dispositions in terms of things making other things more likely (but not necessary). So all causal relations are non-necessary (no causal production is necessitating). In this way, the “could have done otherwise” aspect of free will (the principle of alternative possibilities or PAP) is trivially explained. In this respect, my voluntarily bringing to mind an image of an orange triangle is just an ordinary causal relationship, just like all the others. Yay!

    (2) I’m worried that A&M do not do justice to the idea of causal necessitation, in two respects.

    First, it seems likely that the main idea of causal necessitation concerns relationships between abstract types, not most-determinate types (or tokens or particulars). Like this: given that an event of type A occurred, we will invariably get an event of type B. This is consistent with the particular type-A event, A*, being capable of producing either particular B* or particular B**. If the most fundamental causal relationships are relationships between particulars (as fits nicely with the idea of causation being about one thing causally producing another), then, though there is a kind of necessity in causal relationships, causation is most fundamentally about things making other things more likely, not things making other things necessary. I think this (or something like this) is the idea that A&M mean to defend. I think that they would allow for this type-level sort of necessary causal relationship. But they do not make this clear. More importantly, I think it is this type-level connection that is the home of whatever causal necessity there is (and of at least one of our concepts of causal necessity). And so it is relevant to interpreting the claim that particular-to-particular causal relationships are necessitation relationships. All of this (somewhat complicated) business probably should have been addressed.

    Second, their concept of causal necessitation – it not being possible that there are any conditions that would interfere with or prevent the cause from producing the effect – though interesting, is not clearly the only or best idea of causal necessitation. But they rely on it heavily in characterizing all causal relations (or all causal production) as making-more-likely rather than necessitating. Since all causes can be defeated in producing their effects, there is no causal necessity. Easy! This idea is very strong… but also smells pretty strongly fishy. And here’s a decent objection to it: if a particular cause and its particular effect are framed in terms of their relationship to all other particulars in the universe, there could not be interfering or defeating conditions for the cause producing the effect. In other words, if the relation is between total-world-states or the like, there could be no interfering elements. But why should this oddball move pull the rabbit of causal necessity out of its hat? Moreover, intuitively, such a “necessitating” relationship (between total-world-states) could be of either a making-more-likely or making-necessary variety! If so, things have gone very badly in the attempt to articulate the concept of one thing causally necessitating another.

    Perhaps something like this is closer to being right: A* necessarily produces B* just in case, if nothing were preventing the tendency of A* to produce B*, A* would produce B*. This kind of condition does not hold for the indeterministic causal production of radioactive decay.

    (2) A&M also make some funny and less-worked-out claims near the end, in an attempt to explain – I think – how, when we decide or choose, we “exercise control over” our actions (or are the “ultimate author” of our actions – though I am not entirely clear that this and the control feature are the same). Their idea: we decide or choose, rather than simply react as with a hardwired or learned reflex (or perhaps habit) because we exercise the capacity to have second-order normative judgments about relevant first-order states like our desires. But this is crazy. First, though this sort of feature of the agent or her will is of the right sort to be necessary for the agent to be responsible for her action (in the normative sense, such that it is appropriate that the rest of us hold her responsible), it is not of the right sort to be necessary for the agent to be deciding or choosing rather than acting on reflex or habit. Second, frogs and small children decide and choose, yet it seems likely that they have no second-order normative thoughts. Here’s a better idea: the constitutive account of what it is to decide or choose is probably of the information-processing or neurological sort, not the psychological sort (and not the conceptual sort). On the basis of psychological evidence, all we know is that we do it and that doing it feels a certain way.

    (3) If, on the other hand, we are interested in conditions of the agent or the will that are necessary for it to be appropriate for the rest of us to hold the agent responsible for her actions, then it is more promising that further conceptual work will help. Though A&M’s idea that second-order normative judgment is an inherently social phenomenon is a little vague and probably too strong (Robinson Crusoe can probably judge that his desire to go after the big fish on the off-off chance of catching it rather than hang onto the slightly-smaller fish was greedy and stupid, though I think A&M have to deny this), some conceptual work in the same ballpark might help. After all, holding people responsible for their actions is a social thing, so it should hardly surprise us if there are necessary conditions of the agent or will for holding the agent responsible that include aspects of her social or moral thinking (perhaps even her second-order evaluation of her desires in social or moral terms). But, despite the strong connection between free will and responsibility in the classic discussions of free will, this is really a different kettle of fish, with the (morally) normative tail wagging the (quality of agent or will) metaphysical dog.


  5. There’s a lot there, and I can’t respond to all of it at all or to any of it adequately, but here are just a few thoughts, roughly in order of conviction.

    1a. Your assessment of A&M’s account of choice in terms of second-order desires — that it is “crazy” — strikes me as extraordinarily tendentious. I’m not sure I understand your first argument against it, but if the idea is just that we can exercise the capacity for second-order judgments when we act out of reflex or habit, then I don’t see the problem. First, A&M never claim — nor does anyone who cashes out choice in these terms, so far as I know — that acting on a second-order judgment is sufficient for deciding or choosing in the paradigmatic forming-a-resolution-to-act-as-the-conclusion-of-a-conscious-process-of-thought sense; they quite naturally suppose (or can, anyway) that when we decide in that sense, we exercise the capacity for second-order judgment via a process of conscious thought that yields a resolution to act in a certain way. But it also seems to me — and I think this is just standard Aristotle-Aquinas faire here — that there is a difference between acting on the basis of a habit to choose in certain ways and acting (or perhaps acting*) on the basis of a mere impulse or appetite, let alone doing something quasi-automatically and barely consciously. The difference is just that certain acts that we take not on the basis of some conscious process of thought nonetheless do express second-order desires of ours. It seems pretty natural to me to say that even those actions are ones that we choose, despite their being grounded in our habits, because they do express and in some sense are the products of our second-order judgments. In this respect they differ not only from mere reflex behavior, but also from actions that proceed from desires that we simply happen to find ourselves having (Aristotle and Aquinas disagree about this, famously; Aristotle affirms, and Aquinas denies, that such desires are capable of moving us to action without the endorsement of our will — I’m with Aristotle on this one). So some reflexive or habitual actions aren’t chosen, but others are, and the ones that are are chosen because they express, even when they are not the products of, our second-order judgments. Or so one might say; I have no idea why this should be crazy.

    1b. You also assert that frogs and children choose things. But here you’re just equivocating, or begging the question by denying without argument that there is a difference in kind between rationally chosen action and intentional animal behavior. Aristotle is instructive here again, since he insists that non-human animals and children act voluntarily and intentionally but denies that their actions express any choice (indeed, for that reason he denies that their voluntary and intentional acts are properly actions in the full sense). Maybe this kind of distinction is untenable, but you can’t show it simply by asserting that frogs and infants choose things, or by pointing out that there is some sense in English according to which the verb “to choose” applies to their actions. This is not pre-theoretically guaranteed territory, and empirical psychology is in no position to play the role of a final arbiter here. But in any case I’m not sure how your second argument is supposed to be consistent with your first; your first seems to depend on the notion that we express second-order desires without choosing, but your second seems to reduce the notion of choosing to intentional or voluntary behavior.

    2. Perhaps A&M are sloppy in their formulation of the notion that second-order judgment is inherently social, and I’m just reading them charitably, but even if that’s true, then you should read them that way too. They’re not remotely committed to denying that Crusoe can make second-order judgments. They’re claiming that he couldn’t have developed his capacity to do so had he not been raised in a society whose members communicate such judgments to one another. I think this claim has some pretty damn strong empirical backing from developmental psychology in addition to its intuitive plausibility. There may be some puzzles here about the relationship between language acquisition and primitive self-knowledge, but even if we’d end up having to say simply that individuals isolated from linguistic communities could not fully develop the ordinary capacity for second-order judgments that most normal adults have, that would seem to be enough for A&M.

    I have some more thoughts about your earlier points, but I can only gesture toward them. I don’t think it’s very sensible to think that necessity applies at the level of types rather than tokens, given that (a) the relata of the causal relation insofar as it is supposed to pose some sort of problem for free will and the like seem like they have to be particulars, and (b) the well-known and pretty obvious fact that, at the level of types, virtually nothing we call a cause ever guarantees what we call its effect doesn’t by itself seem to pose any problems for the idea that when A causes B, it necessitates it; we all know that As can fail to bring about Bs if the conditions aren’t right, if something interferes with A’s bringing about B, etc.; but we may well still think — and many causal realists do think — that when an A causes a B, it necessitates it rather than simply making it more likely. In fact, I think this is probably where the most plausible line of resistance to A&M’s view comes in. Granting the obvious fact that a rock hitting a window might fail to shatter the window because some feature of the conditions prevent it from doing so, if we suppose that in those cases where the rock does shatter the window, it doesn’t necessitate it, then we leave it utterly mysterious why it does. I don’t say that this objection is compelling, just that it strikes me as having some force and as the most plausible way to resist A&M’s view.

    Ok, there’s my quick, inadequate response. Thanks!


  6. David – Much thanks for the response/push-back. Your comments made me think really hard and change my mind on some stuff (especially regarding causation).

    ***[reply to your 1a and 1b]***

    I do take ‘decide’ or ‘choose’ to refer to the mental aspect of voluntary action. It is pretty unreasonable to assert that mice and birds and frogs and such do not sometimes act voluntarily. People do sometimes use ‘choice’ (or ‘free choice’) to refer to that quality of will, whatever it is, that is necessary (or that we take to be necessary) for people being responsible for their actions (or such that the rest of us might appropriately hold them responsible). I took A&M to be concerned with the first thing, not the second. For it is this feature that is most obviously potentially at odds with the idea of causal determinism.

    If A&M mean to be addressing the second thing, then their view is more plausible. For it is plausible that the quality of will behind being responsible for one’s voluntary actions (in the normative sense of ‘responsible’ such that others might appropriately hold one responsible) necessarily involves some capacity to be sensitive to the quality of one’s first-order desires or values. I’m not sure this sensitivity needs to take the form of higher-order normative judgments, though. We seem to appropriately hold small children who might not have the concepts and other capacities for such judgments responsible for doing some of the things that they do.

    However, I do feel some of the pull toward explaining this quality of will in terms of higher-order judgments about desires. This is similar to the pull I feel toward explaining rationality or justification in terms of higher-order judgments about the reliability of one’s processes of belief-formation and action generation. The most basic problem with these sorts of accounts is that they are potentially good accounts only of the more-sophisticated ways in which we might exhibit the target rational or quality-of-will property. (Take a look at Arpaly and Schroeder’s IN PRAISE OF DESIRE, chapter 1 for some powerful arguments against this kind of move with regard to explaining what makes basic rational response rational – or, in their terms, what it is to think or act for a reason in a basic way. In their terms, all basic rational response is non-voluntary, non-deliberative and non-reflective – “ND-processes.” Irfan and I are presently reading through this book together.)

    ***[reply to your 2]***

    I should not have used the Robinson Crusoe case – you are right that A&M do not have to deny that he is capable making second-order normative judgments about his desires (for he was socialized in a roughly normal way prior to being alone on an island). A being-raised-by-wolves case would be better. The idea that our self-evaluative capacities develop from being evaluated by others or society is an interesting thought and something in this ballpark might well be right – I’m pretty sympathetic to the idea that social evaluation kicks off at least a certain important and ubiquitous kind of self-evaluation.

    But I doubt that all self-evaluation of this kind. It is at least plausible that some of our capacities for rational self-evaluation (including evaluations of our attitudes) do not have such thoroughly social roots. Also the social hypothesis assumes that others have a prior the capacity to evaluate one – but if they can do this to us then we can do this to them and so why can’t we just exercise this prior capacity on ourselves straight away? It does seem plausible to me that the *kind* of self-evaluation (including the evaluation of one’s desires) that is relevant to being responsible for one’s actions (in the sense of it being appropriate for others to hold one responsible) is both a capacity developed through being evaluated according to pro-social standards by others or society and the application of such standards to oneself.

    ***[reply to your reply regarding causal necessitation]***

    Your comments here are really on-target and prompt me to get clear on some things that I was not clear on before. Here goes. First off, I now realize that in order to properly frame a type-level necessitating relationship, the defeating conditions have to be included in the framing of the cause (I think this is how putatively objective making-more-likely relationships get reduced, on some views, to conditional necessities). So one should re-frame ‘when you have a type-A event you always have a type-B event as well, except when a type-C condition obtains’ as ‘when you have the condition of type-A event with no type-C event, you always get a type-B event.’ So my correlation-style formulation of a kind of type-level causal necessity was wrong. And I suspect that my suggestion for formulating a kind of particular-level causal necessity makes a similar mistake (as well as at least one other mistake – see below).

    Taking this to heart, a special case of this kind of type-level causal necessity would concern only ultimately-determinate types (like qualitatively identical atoms of sub-atomic particles or their basic properties). Such a brand of ultimately-determinate-type-level causal necessity does not essentially involve any process of causal production. Hence, the fully particular (numerically distinct) instantiations of such ultimately-determinate type-level causal relationships – particular causal relationships between particular things – need not involve anything like causes *producing* effects either (even if, among the sorts of causation we typically experience, there is a process of causal production). And you can characterize ultimately-determinate type-level and fully particular causal (bounded) *indeterminism* in just the same sort of way.

    On this way of thinking about causal necessity and (bounded) indeterminism, once you properly specify a cause at a level of generality (including the absence of all defeating conditions), causal necessity or (bounded) indeterminism at that level of generality is itself specified. If type-A events always come along with type-B events (where types A and B are not ultimately-determinate types), then type-A events necessitate or determine type-B events. This schematic scenario is consistent with there being indeterminacy at a more determinate level of generality: perhaps type-A events sometimes come along with type-B1 events and sometimes come along with type-B2 events (where type-B1 and type-B2 are determinates of the determinable type-B). At the level of ultimately-determinate types whether or not there is necessitation is determined in the same way. And thus, via instantiation, you get necessitation or non-necessitation verdicts on fully particular causal relationships.

    I think something like what I have sketched could incorporate the idea that particular causes always *produce* their effects (perhaps making use of an expansive sense of ‘production’ that takes into account any type of transfer of energy or the like, so that additional material processes are not required at least not at the level of basic physics). For the essential point is just that the fundamental modal features here are features of type-level correlations, not features of the types or tokens of any *other* sort of relationship that might necessarily hold between cause and effect.

    I agree that A&M are vulnerable to the kind of objection according to which each uranium atom’s decaying is necessitated (even if one decays X years after its birth and another at X+1 years after its birth) because causal production is necessitating. In other words, A&M need to say more to defend their idea that causal production is always making-more-likely, not necessitating. For reasons already stated (my total-world-state-causes objection from my last comment), I think there may be some problems with how they are thinking of causal production itself being necessitating or not. But more fundamentally, because what I have sketched here is plausible, I’m not sure that they get it right about where and at what level of generality the modal properties of causality inhere. If they are mistaken here, they would also be mistaken in relevant respects regarding what causal necessity is (this would seem to be relevant to my total-world-state-causes objection having some bite).

    This stuff is hard! I’m sufficiently interested and impressed, though, to read more of Mumford’s related material on causation.


    • (Take a look at Arpaly and Schroeder’s IN PRAISE OF DESIRE, chapter 1 for some powerful arguments against this kind of move with regard to explaining what makes basic rational response rational – or, in their terms, what it is to think or act for a reason in a basic way. In their terms, all basic rational response is non-voluntary, non-deliberative and non-reflective – “ND-processes.” Irfan and I are presently reading through this book together.)

      For whatever it’s worth, I’m less convinced than Michael of the cogency of Arpaly-Schroeder’s arguments. It’s clear enough from what they say why basic rational response must be non-deliberative, but not why it should be non-voluntary or non-reflective, or even what exactly is involved in saying that it’s non-reflective.

      My basic problem is that they deploy a series of regress arguments to the conclusion that Michael cites above (i.e., that basic rational response to reasons for action is non-deliberative), but for reasons that are not clear to me, the regress arguments consistently lead them to a form of non-cognitivism about practical reason. The inference to the latter conclusion seems unmotivated to me, so to speak. On one common interpretation of Aristotle on practical reason, we don’t deliberate about ends, but only about what promotes ends, and yet it doesn’t follow that we’re led from that to any form of non-cognitivism. I don’t think Arpaly-Schroeder are sensitive to that possibility. I’m only 1/3 or 1/2 way through the book, but so far, it seems to me to reproduce a standard neo-Humean response to neo-Kantian-type views, neither of which strike me as adequate.

      Incidentally, we were planning on blogging some of the results of our Arpaly-Shroeder conversations here at PoT, including some conversations on interesting side-issues stimulated by the book, but just haven’t had the time. This, of course, is because I am on blogcation.


        • Yeah, I know. I don’t think I’ve kept a single promise I’ve ever made to the readers of this blog. In fact, it’s gotten to the point that when I make a promise, you can probably predict that I definitely won’t do whatever it is I’ve said I’ll do. The result is like a living example of Kant on false promises.

          ‘It may be that this principle of self-love or of personal advantage would fit nicely into my whole future welfare, ·so that there is no prudential case against it·. But the question remains: would it be right? ·To answer this·, I change the demand of self-love into a universal law, and then put the question like this: If my maxim became a universal law, then how would things stand? I can see straight off that it could never hold as a universal law of nature, and must contradict itself. For if you take a law saying that anyone who thinks he is in need can make any promises he likes without intending to keep them, and make it universal ·so that everyone in need does behave in this way·, that would make the promise and the intended purpose of it impossible—no-one would believe what was promised to him but would only laugh at any such performance as a vain pretence.’

          Mock on, mock on, ’tis all in vain. You keep laughing at my false promises, and I just make some more again.


        • Essentially the argument that Arpaly-Schroeder offer on pp. 40-42 of In Praise of Desire.

          Deliberation does not ground thinking and acting for reasons. But there is another way in which Reason could be given a privileged role over Appetite in the explanation of thinking and acting for reasons. The defender of Reason might say that to think or act for reasons we must represent to ourselves that we are thinking or acting for reasons. It might be said that an action is one performed for reasons just in case it is an act that is believed reasonable, consciously believed reasonable, thought to be justified by the reasons before the agent, or otherwise taken (cognitively) to be appropriated licensed. Call this view “Recognition.” (p. 40)

          They then cite work by Melissa Barry, Christine Korsgaard, Jay Wallace, and Niko Kolodny as representing (to so speak) the view that they’re attacking. But I don’t think they do justice to what B-K-W and K say, and more to the point, they don’t do justice to the broader family of cognitivist views that B, Kors., W, and Kol. represent. Barry and Korsgaard are neo-Kantians, and though it’s been awhile since I’ve read them, I don’t think that Wallace and Kolodny quite capture what’s distinctive to a non-Kantian, non-Humean, neo-Aristotelian conception of practical reason. So the latter sort of view goes undiscussed (at least in what I’ve read so far).

          Near the end of the same section, Arpaly-Schroeder say:

          Whatever makes it possible for us to think and act for reasons, it must be a process that is nondeliberative, nonrecognitional, and nonvoluntary (an “ND process” for short) (p. 42).

          As far as I can see, their arguments secure the first conjunct, but not the second two. “Non-cognitivism” refers to the second conjunct.


  7. I haven’t read enough to have even a half-informed opinion, but from the discussion here I suspect that an awful lot of the trouble here has to do with different notions of what a reason is, and maybe even trivially semantic disputes about how to use the term. I’ve often found myself pulled in incompatible directions when thinking about these things, but what I’ve concluded is that the appearance of incompatibility is just an artifact of a muddled way of using the word ‘reason.’ I used to think, for example, that it was just obvious that non-human animals of mid-level cognitive sophistication (squirrels, dogs, etc.) act for reasons, only to be led to thinking otherwise by the thought that acting for a reason has to have some connection with reasoning. But there is no deep conflict there, because I was simply equivocating on ‘reason,’ taking it first as something roughly like ‘conscious goal to which one searches for effective means’ and then second as something like ‘conceptually grasped consideration that seems to tell in favor of doing something.’ (My confusion here was in part created and in part resolved by reading Anthony Kenny’s Aquinas on Mind up against MacIntyre’s Dependent Rational Animals). I doubt the issues here could be resolved so easily, but I wonder whether there is a similar confusion at work. I find myself thinking, for instance, that “whatever makes it possible for us to think and act for reasons” must indeed be a process that is “nondeliberative, nonrecognitional, and nonvoluntary,” but wondering why on earth I should therefore think that acting for reasons – basic or not — is nondeliberative, nonrecognitional, and nonvoluntary. I’m also pretty sure I don’t know what they mean by “voluntary,” because I think that standard non-rational animal behavior is perfectly voluntary and that this fact is not threatened in the least by the possibility that non-rational animal behavior is entirely deterministic.

    In short, I’m confused. But I should probably read their book anyway. After all, Schroeder’s office is literally a stone’s throw away from mine; I should probably know a little bit more about the relevant work people at my own institution are doing.


    • I agree with that. Having read the “reasons” literature off and on, I find it incredibly frustrating and usually unilluminating: it seems to take a series of substantive moral-psychological assumptions for granted, along with a highly specialized (often extremely counter-intuitive) vocabulary intended to convey those assumptions–and the discussions take off from there. I often find it completely unclear what’s being said in any given book or article, or what’s at stake in saying things one way or another. But it seems to me that that’s an occupational hazard of any quasi-Scholastic approach to philosophy, and the reasons literature is the epitome of one. In my comment, I was mostly just relaying what Arpaly-Schroeder say, along with my rejection of what I take it to mean or entail. But I’m still only about 1/3 of the way through the book.


      • Ah, but one would hope that a quasi-Scholastic approach would succeed in drawing distinctions clearly! I think the opacity of the literature is only incidentally related to its subject-matter; after all, failure to draw relevant distinctions (or to agree on what the relevant distinctions are) characterizes most philosophical debates to some extent. I’m in favor of giving an important role to the concept of reasons in our thinking about action; I’m just opposed to doing it badly, and I won’t challenge your claim that that’s how it’s often done.


  8. (1) Okay, now I know how you are using ‘non-cognitivism about practical reason’, Irfan. Just to be clear, though: non-cognitivism, in this sense, just means that, in order to be thinking or acting for a reason, it is not necessary for one to represent what one is doing in relevant terms (e.g., involving the concepts of reasons, having reason, rational response, justification, reliably tending to produce relevant outcomes, etc.). That’s consistent with paradigm cases for the application of the concept (or thinking or acting for a reason) exhibiting this sort of performed-under-relevant-description condition.

    (2) David (and Irfan), yes there is a lot of sloppy reasons-talk. There is also some pretty good and precise reasons-talk. Here are two distinctions that I find helpful that are not always clearly reflected in the language that philosophers use.

    (a) (i) If R is a reason (say an objective reason rather than a good rationalizing reason) for agent X to exhibit action or attitude A (ii) then X has reason to exhibit A and (iii) then there is some reason *r* that is a reason for X to exhibit A. The first and third things reason-properties most broadly concern *propositions* (facts or true propositions for objective reasons). The second and third things are necessarily correlated. The second thing is of a different sort from the first and third. It is a broadly normative feature *of the response* (the action or attitude). We could, perhaps, imagine a response that “has something going for it” (relative to the agent) even though there are no corresponding reason-facts (including facts about relevant or valuable ends) that go into making this the case. The standard view is that things like (i) explain things like (ii) (and of course things like (iii)), but how all of this hangs together – and as well the different relationship between good rationalizing reasons and objective reasons – is a matter of on-going conceptual work.

    (b) Reasons and ends (point, purpose) of responses (actions, attitudes) are different. Plausibly, all reasons either are or exist relative to ends. But ends have a special role not shared by reasons that are not ends.

    For example, suppose that I presently have reason to go grab a sandwich. Since this is a practical reason, we would most naturally take the reason here to be a fact (or to be an objective normative reason). The main end (desire, purpose, valuable or beneficial end) that makes this the case might be my having my feeling of hunger alleviated. The fact that sandwiches alleviate hunger would be a non-end type of reason for me to grab a sandwich.

    (c) Actions that one has intrinsic reason to perform can be considered to be ends in a formal sense, yielding a unified framework. And similarly for other responses (attitudes). This yields a kind of agent-relative consequentialism that does not necessarily require the relevant formal ends to be valuable (not even valuable-to-the-agent). So the reasons-that-are-ends, reasons-that-are-not-ends and having-reason-to-response framework can be treated as universal.

    This last element is controversial, but I think only because people get hung up on how the word ‘consequentialism’ is usually used. Say ‘teleological’ if you prefer, but I worry here that what is behind this move is distinguishing Aristotle from agent-neutral-value consequentialist theories (like classical utilitarianism).

    (3) Applying some of the above to the case at hand (what is required or not in order for one to be thinking or acting for a reason), suppose that I go to the kitchen in order to grab a sandwich in response to my being hungry. The end here is alleviating my hunger. That eating a sandwich would alleviate my hunger and that there are materials for making sandwiches in the kitchen are both reasons (of the non-end sort, as it typical when we cite reasons) for my walking to the kitchen. Moreover, I grasp these facts (and my desire not to be hungry) and respond to them in accord with a rule of good reasoning (reasoning that is good at least in a respect) that goes something like this: from (i) my seeking to achieve some desired end of mine P and (ii) my believing that my action A achieves or significantly promotes P to (c) attempting to A. The rationalizing reason here would be the content of my descriptive, instrumental belief (the content of what one is intending to accomplish does not seem to be covered under the concept of a reason, but there is a broadly similar role for both the goal and the belief – they are both premise-like elements in a rational transition in attitude (or from attitudes to attempted actions).

    I take what Irfan calls the cognitivist position regarding this transition to be: absent additional beliefs (that are perhaps background conditions for the transition/inference, not part of the transition/inference) to the effect that the transition is rational, reliable, good, justified, etc., it does not count as rational or as one acting for a reason. (You might also or instead hold that there are other, different sorts of background higher-order beliefs about the pattern of transition itself, what it tend to promote, etc.) I take the non-cognitivist position to be that this kind of transition is rational regardless of whether the agent has any such background, higher-order beliefs concerning the transition (or the pattern that it exemplifies). This seems pretty clear.

    What is less clear to me is whether this is a matter that can be settled by our conceptual intuitions (how we use the phrase ‘think or act for a reason’) rather than via extensions of the use that have good explanatory backing. (The same holds for the question of what does and does not count as a desire – another important topic in IPOD.) I’m at least tempted toward thinking in the latter sort of terms. So the above pattern of transition in attitude, bereft of any higher-order background beliefs the thinker/agent might have about it, is *in some sense* a good rational transition to intention/action, made in part on the basis of some belief. Whether the content of this belief counts as a (good rationalizing) reason might be a matter of the best way to draw precise conceptual boundaries rather than a question of what the boundaries of our present concepts are.

    However, I do think there is a general distinction between transitions in attitude that are correct (or that reliably achieve or promote relevant ends, like beneficial action or true belief) and those that are rational (in a sense of ‘rational’ that does not include mere corrrectness or reliability). This kind of distinction is familiar from debates about reliabilism in epistemology. This does strike me as a conceptual matter, but as much one of good use as one of existing use (that is assumed to be good, to divide nature at her joints). And it is not clear that Arpaly and Schroeder face this issue squarely. They do deny that correctness (and, I assume, reliability) and rationality (or justification) are distinct ways of evaluating transitions in attitude, but their basis for saying this is simply cases and intuitions. This will not do if the relevant conceptual work is as much that of extension or precisification of terms like thinking or acting ‘for a reason’ (or ‘responding rationally to X’) as it is work of constructing relevant scenarios and intuition-pumping (to get definite contours of existing concepts). Strangely, they seem to be quite aware of this general sort of methodological issue in their treatment of the concepts (and properties or phenomena) of ‘action’ and, centrally to the book, ‘desire’.

    Liked by 1 person

  9. Two more items relevant to my above comment (one of which is a terminological correction or clarification):

    (A) I was just reading this in anticipation of going to hear a talk at BU by the author on the topic of the paper (Matthew Silverstein). This is worth reading to get up-to-date on (some of) the literature on normative reasons. The thesis is that normative reasons are a function of sound reasoning (basically, premises in sound reasoning) – where soundness is construed as a descriptive property. Because soundness is non-normative, this position facilitates a non-naturalistic reduction of normative reasons to non-normative elements (non-naturalistic because soundness, like other things logical, are generally viewed as genuine mind-independent abstracta).

    I don’t buy this view because I think that though the soundness of sound deductive reasoning (and analogous features in other doxastic and practical reasoning) is non-normative, reasons are a function of the rationality of (good) rational response, not sound reasoning. So in order to reduce reasons and get to the non-normative stuff you have to reduce (good) rational response. Logical relationships between attitudes go into constituting basic (good) rational response, but they do not suffice for it. Silverstein has nothing to say about what more (above and beyond objective-relationship features like soundness or reliability) is needed in order to get (good) rational response – because he holds that nothing more is needed at all. My criticism here is closely related to my objection to Arpaly and Schroeder’s account of basic (and good) rational response: their account equates basic (and good) rational response with reliable response *that is caused by the logical features displayed in the pattern of response*. Though they are in a better position than is Silverstein to explain basic (and good) rational response (and hence at least normative reasons of the good-rationalizing-reasons sort), I worry that all they can succeed with in making this move is something like distinguishing chancy from non-chancy reliable production of relevant outcomes for a bit or type of reasoning (whether or not, or to whatever extent, logical relationships between propositions are “Platonic”). Anyhow, Silverstein’s paper is a good read, if a little deep in the weeds of reasons-esoterica at points.

    (B) In reading this paper, I came across – then immediately remembered – another, more common sense of ‘cognitivism about practical reason’. The view that commonly goes by this name is simply the view that practical reasoning always terminates in judgments (say, about what one ought to do), not intentions or actions. Hence, practical reason is a species of theoretical or doxastic reasoning. That is different from the idea that practical reasoning (or any other sort of reasoning) requires (is partly constituted by) relevant higher-order judgments about that reasoning – what I at least took (perhaps mistakenly) Irfan to be expressing. We can of course use the term as we like (and I can misinterpret and be confused as I like), but we are probably better off sticking to the standard use in any case.


    • Can you say a little more about how soundness is non-normative? I’m finding that difficult to understand. Reasoning in general seems about as normative as it gets; some reasoning is good, some reasoning is bad, we ought to affirm the conclusion of a valid argument with true premises, we ought not to accept the conclusions of invalid arguments on the strength of such arguments, etc. I wouldn’t deny that claims like “that argument is sound” and “that argument is unsound” are descriptive; but short of some ambitious metaphysical arguments, there is no reason to think that the descriptive and the normative are mutually exclusive. So I suspect I don’t understand your use of the terms, or perhaps I’m too optimistic in thinking that we can truly describe arguments as good or bad or claim that we shouldn’t accept the bad ones and should accept the good ones.


      • It also seems to me to be odd and perhaps question-begging to say that if practical reasoning terminates in judgments about what to do, it is therefore a species of theoretical reason. No doubt the two will have much in common, but that’s just because they’re both species of reasoning, not because they’re both species of theoretical reasoning.

        For Aristotle, this distinction is easy and obvious, though the details are not. Theoretical reason not only terminates in judgments or beliefs, but takes as its objects propositions that are true either necessarily or for the most part, and aims not merely at discovering true propositions of that sort, but at acquiring explanatory understanding of them. Practical reasoning is ultimately about what to do in particular situations, and the true judgments it reaches are not always true always or for the most part; nor does practical reasoning as such aim at explanatory understanding, however necessary such understanding might be. There are, of course, longstanding disputes about whether Aristotle thinks the conclusion of practical reasoning is an action, a decision, or something else. But even if it is always an action or decision, judgments about what is good and what one should do are a necessary step along the way to an action or decision, and in any case the space between the judgment and the action or decision, if there is such space, is not bridged by any further reasoning (when you conclude that this is what you should do right here and now, you just do it, and if you don’t, it’s because external or non-rational factors prevent you from doing so). So for Aristotle, practical reason’s distinction from theoretical reason has nothing to do with practical reason’s not being directed toward the formation of truth-apt judgments and beliefs; all practical reasoning necessarily, on his view, is directed toward the formation of truth-apt judgments and beliefs, even though these are ultimately for the sake of action; the fundamental difference lies in the modality of the propositions judged and the role of explanatory understanding.

        That’s not to say that you or anyone should distinguish between theoretical and practical reason in the way Aristotle does; it’s just to observe that the guy who more or less invented the distinction does not think that reason’s being practical is inconsistent with its essentially involving or even terminating in judgments and beliefs. I suspect that any impulse to regard practical reason as not essentially aiming at the formation of truth-apt judgments and beliefs that are to be put into action would flow from some sort of skepticism about the possibility of true judgments about things being good; at any rate, subjectivism about the good is the only thing I can think of that would make me inclined to deny that practical reasoning is essentially concerned with reaching true judgments. But I may be misunderstanding, and I’m sure I’m missing something, so that’s why I’m asking for clarification.


  10. Pingback: Tenth Annual Felician Institute Conference: Tentative Program | Policy of Truth

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s