Blog Coordinator

« Featured Author: Alfred Mele | Main | CfA: NOWAR 2 [Moving to the front from Dec. 8] »



Feed You can follow this conversation by subscribing to the comment feed for this post.

I'm with Randy on this one, Al. If we assume the truth of the causal theory of action, all there is to an exercise of control is that the agent's behavior is caused in the right way by some appropriate mental items. If a libertarian is accepting this basic framework, it is not clear to me how the agent's control in initiating and guiding her action is enhanced. It is not clear to me that her control is vitiated, either. I'm not sure what sorts of additions the event-causal libertarian must make to enhance control, but the simple fact that indeterminism is true does nothing. (This became clear to me when I first read John Bishop's *Natural Agency*.) So Jane and Joan seem to be exercising the same variety of control in making their choices. That is not to say that everything is the same. If anything, if Kane is right, assuming that Joan's making her choice is an indeterministic process of the sort he describes in his work, then Joan actually faces more difficulty in making her choice. But given how the process goes and given that it does not involve the exercise of any sui generis causal powers by Joan such as the agent-causalist would posit, Joan is not clearly enjoying any additional control than Jane. So they are in the same boat with respect to their control in making a choice.

I should have said that "at best, they are in the same boat with respect to the control they enjoy." If Kane is right, Joan may be worse off than Jane. But I'm not convinced that determinism or indeterminism makes any difference if we assume the framework given by the CTA as our account of agency.

Hi Al,

Thanks for the post. It nicely illuminates, I think, the crucial need for renewed investigation of the notion of agent causation, specifically how such a notion might help illuminate that of an agent’s causal control over her own bodily and mental actions.

In precisely that spirit, in response to Question A, it seems to me that the correct answer is that *no* event-causal account of action – whether libertarian, compatibilist, or otherwise – is capable of ensuring that the agent has control over her own bodily actions, if we understand the notion of control in causal terms.

The reason why should be familiar: event-causal accounts of bodily action typically explain an agent’s control over her own bodily action in terms of the causal relations between the relevant agent-involving mental events and the subsequent bodily movements. Crucially, it is those agent-involving mental events to which we must assign causal control over the subsequent bodily action, not the agent. Hence, the agent has no control over her own bodily action.

In response to Question B, it seems to me that until we have agreed upon the very notion of control itself, we won’t be in a position to measure its presence. We have to know what we’re looking for before we set out to measure whatever that might be!

Lastly, in response to your remarks about Jane and Joan, it seems to me that in Jane’s deterministic world, we are not entitled to assume that her mental action of assessing the relevant reasons is itself under her control, if we understand the notion of control in causal terms. Given the assumption of determinism, that is, Jane has *zero* control over her own mental and bodily actions, because she is causing neither.

They are all wrong, Professor Mele. Compatibilist control- deterministic event causation of one’s choices by one’s beliefs and desires- is obviously greater than the volitional power accorded to an agent by Prof. Kane’s philosophy, which is nil.

For all his talk about efforts of the will, Kane never succeeds in meeting the Mind objection: the outcomes of his "conflicts of the will" still seem to me lucky or accidental. Yes, a mental effort on the part of the agent is necessary to end such a crisis, but it too must prove indecisive, given Indeterminism. There is nothing about a "self-forming" agent that is responsible for how he ends up volitionally; there is no explanation for it at all- it just happens, beneficially or detrimentally depending upon the circumstances. At least with Compatibilism, we can point to states or events in which they agent is involved or beliefs and desires belonging to him as the proximate causes of his choices. He is at least a part of what brings about his volitions; we can plausibly hold him responsible as the subject of those states. Kane’s view, on the other hand, not only eliminates the executive element in decision making, found only in Agent Causalism, but renders us the victims of luck or chance as we go about trying to do the right thing by ourselves and others. Prof. Clark, for his part, punts (in LAFW, a great book) on the key question of whether probabilistic relations between events could amount to causation. Bottom line: if I can’t be an Aristotelian hand mover, I’d rather be a clockwork orange than a roulette wheel.

Al, I think you’ve clearly identified the fundamental issue: we need to figure out how to measure control.

In order to measure control, I’m thinking we’ll first need to develop the ability to measure the forces that effectively exert the control. In order to accomplish that, we’ll need to determine what “system level” is exerting what force. Due to our natural human references, we only perceive of the *result* of the net sum of forces after the sum has already occurred for each moment of time, and therefore it’s extremely difficult for humans to sense the forces exerted from different system levels. I believe that’s the reason why humans have naturally developed the theory of reductive materialism and have concluded that there are only 4 ea. fundamental forces in reality. Science simply doesn’t believe there’s any such thing as “new emergent forces” – forces that result from something more than a direct sum of existing forces – forces that have an emergent property of life to them. Humans are capable of sensing a *very* limited range of forces directly. In summary, in order to measure control, I’m thinking that mankind needs to first develop the ability to sense forces in additional fields. That’s going to be a difficult task, and I believe the first step in that process is to develop arguments that prove human thoughts exert new emergent forces. (I’ve developed a shot at that argument if anyone is interested.)

The crucial difference between Jane and Joan consists in the fact that it is metaphysically possible for Joan to jump left and to jump right, whereas it is not metaphysically possible for Jane to jump left and to jump right. But Joan exercises no control at all over whether she jumps left or right - the indeterministic process settles that, and there is nothing Joan can do about it (she can can - compatibilis - control what the chances are of each, but she has no further control beyond that). For that reason, she has less control than Jane: her control is compatibilist control plus luck. It is nevertheless open for the libertarian to hold that that's enough control, and enough control plus metaphysical openness gives Joan a greater degree of freedom than Jane has, despite her enhanced control.

Thanks for the post Al. I think Kane is right on the money on this issue, though I squabble with him over what is the best version of event-causal libertarianism. How does one measure control? Not sure, but here’s an idea: the degree of control an agent has with respect to his action A is a function of the abilities, skills, opportunities, and know-how he has with respect to A. I have more control with respect to my jump shot the more skill I have in producing certain outcomes. I have more control with respect to how my life unfolds if I have a large (rather than a small) amount of abilities and opportunities to choose and act in a variety of ways. This is at least a place to start (though not a place to end). Thus, one could conclude that Joan has more control than Jane because Joan has more opportunities (or maybe abilities) that Jane (I hope I got the names right).


I don’t see why accepting CTA commits one to believing that “all there is to an exercise of control is that the agent's behavior is caused in the right way by some appropriate mental items.” Presumably accepting CTA commits one to think that control partly consists in the right causal route from the right mental states. But why think it follows that this *exhausts* control? Maybe there are independent reasons for thinking this exhausts control, but I don’t see why it follows merely from accepting CTA. Remember, part of Bishop’s argument that indeterminism is irrelevant his acceptance of FSCs.

Chris, I can see why having more opportunities might mean more free will (that doesn't mean I accept it!) But why think more opportunities = more control? There are more things I can do, but that doesn't obviously entail I have more control over which I do. Nor is it obvious to me that if I do a, I have more control in so doing if I select from options a,b, and c rather than a and b (indeed, I can imagine the opposite being true).

Neil: suppose Ann is locked in any empty room with no way of escape. She (now) has no control over whether she stays or whether she goes. Then someone unlocks the door, giving her the opportunity to leave if she so chooses. Wouldn't you say that she now has more control over whether she stays or goes? If so, then we have at least one scenario in which adding an opportunity adds control. Of course, this doesn't show that having more opportunities entails having more control. But it does support Chris's claim that giving a person more opportunities can enhance control. Any of that sound right?

Justin and Chris--I have to agree with Neil here, on the basis that opportunities are generally externalist factors in freedom, whether it is unlocked doors, elbow room, or even indeterministic metaphysical possibilities. None of those entail internalist abilities to take control/advantage of those opportunities. Whether Ann stays or goes depends on her mental states, and the control has to be located there. If she is demented, hallucinatory, etc. she can't take advantage of the unlocked door. Even if her mental states are indeterministic, those blocking factors must be absent for her to wield some sort of control. If she is rational, perceptually competent, etc. then she counterfactually wields control more than demented, hallucinatory Ann can ("can" as power here) even if her mental states have accessible metaphysical possibilities through indeterminism. If the latter Ann tries the door by accident, she may leave; she exercised zero control in doing so.

Al (if I may)--great question. Although my friends (especially ones named Betty) usually call me Al as well, I'll stick with Alan here on Flickers to avoid confusion.


You are right about Bishop and FSCs. But if the question is simply about control and if the mechanism whereby control is exercised is explicated in terms of the causal role of some appropriate mental items, then whether the process of an agent's producing an action is indeterministic or deterministic seems to matter little. The same sort of control is still exercised whether or not the process is deterministic or indeterministic (I say this as someone friendly to the new dispositionalism and, hence, not altogether convinced by FSCs.)

What I want to know is what more the event-causal libertarian can add besides the process being indeterministic. Parallel processing alone won't help since that can still be deterministic. This gets to the heart of Al's second question. And if Joan enjoys no more control than Jane, then what should we add? I'd say sui generis agent-causal power. But I have my doubts about such strategies.


You should have called it a Locked room (get it?).

Seriously, suppose that Ann's control is enhanced. It is enhanced because of a change in her environment. If that is right, then the case you describe and Al's case seem disanalogous. The reason why is because the case of Jane and Joan is about the indeterminism or determinism involved in their decision-making. If we suppose that the process is the same in both cases (including the outcomes), then that Joan's decision-making was indeterministic while Jane's was not does not obviously matter. The means whereby both exercise control is the same and, if you'd like, neither of them has a special power to guarantee the outcome of the process in a way that takes them outside of the process (in the way that agent's seem to be able to on some agent-causal theories). So since the questions seems to be about whether indeterminism in the decision-making process enhances control, your example does not help.

With all of that said, I'm sympathetic to Neil's claim. So perhaps we can say that Jane enjoys more freedom than Joan. But I'd insist that it is not because Jane has more control.


Its sounds strange to my ears to say Joan has more free will with respect to her action than Jane, but not more control. But maybe I am thinking of this wrong. I’d be interested to hear why (and how) you think they might come apart. I agree that more opportunities may not necessarily add more control and even may in some cases diminish it (here I think of Frankfurt’s interesting claim that freedom must have some constraints). However, Justin’s case nicely illustrates why I find it plausible to say opportunities can increase control. Moreover, Justin’s case illustrates how sometimes adding a single opportunity can make a significant difference.

In your earlier post you wrote: “But Joan exercises no control at all over whether she jumps left or right - the indeterministic process settles that, and there is nothing Joan can do about it.” Why not think that Joan’s exercise of control over whether she jumps left or right consists in the indeterministic causal process of her mental states bringing about the relevant action? If Joan exercises control in her jumping left, then doesn’t it follow that she exercises control in her jumping left rather than right? After all, jumping left is sufficient for jumping left rather than right. Why then isn’t Joan’s exercising control over her jumping left sufficient for her exercising control over her jumping left rather than right?


There may be two different meanings for the word “control” that need to be considered, just as there are likely two fundamentally different meanings for the term “free will”.

I’m thinking there’s a “weak sense” and a “strong sense” for the meaning of the word “control”. In the weak sense, a person’s thoughts clearly have the ability to cause something to happen – to exert control. In your example, Ann clearly has the ability to leave the empty room if she desires, once the door is unlocked. She effectively acquires more control in the weak sense – she can cause a wider range of things to happen.

The “strong sense” of control is different. To suggest that someone has control in the strong sense, we're claiming that their thoughts must be capable of exerting new emergent forces (i.e., forces that don’t result solely from a direct sum of existing forces). To relate that to your example, I don’t think Ann has more control in the strong sense just because the door becomes unlocked – her thoughts don’t exert more forces within her physical brain when the door is unlocked vs. when the door is locked. If a person believes that Ann has control in the strong sense, then her control isn’t a function of the lock on the door.

Alan, Andrei, and James: Neil questioned whether having more opportunities means having more control. My example was simply intended to illustrate that sometimes having additional opportunities does give one more control. I didn't claim the example did more.

I do think it's interesting to note, though, that merely extrinsic and environment features can affect control. Take a standard Frankfurt-style case, to use a different sort of example. Jones can't avoid killing Smith because of Black, etc., etc, but Jones kills Smith "own his own" (i.e., without Black's help). Notice, Black's presence is an environmental factor. By hypothesis, it's not supposed to significantly affect Jones's intrinsic features or play any causal role in the production of Jones's action. But adding this environmental factor can make it so that Jones can't (the can here is that of ability plus opportunity) avoid killing Smith. That is, it can make it that Jones has no control over whether he kills Smith. Now, if we grant that adding such environmental/extrinsic factors to the agent's circumstances can diminish control, why shouldn't it be that removing them can enhance it as well?


I think I know what John would say here: in a regulative sense of control Black affects Jones, but not in a guidance sense. Jones kills Smith with an actual sequence guidance control, and Black cannot affect that since no factor that Black can control (by itself guidance?) has any actual effect on what Jones does. Only if one assumes that Jones possesses regulative control can one conclude that Black affects that control. But that sort of control is that very thing at issue here.

Justin, I find the example at lest somewhat plausible. Indeed, she now has more control,over whether she stays in the room. But it doesn't show that more opportunities = more control. Add another two doors,,or three, or 28, and she has just as much control,over whether she stays in the room as in the original case.

Chris, it is agents who possess control. Agents have minds, and minds consist of s multiplicity of discree and dissocisble mechanisms. These mechanisms are not reasons responsive or reactive on their own; at least they are responsive to too narrow a range of reasons for their activity,,taken one by one,,to be exercises of control. That's one reason why I don't think the causal,process can be identified with an exercise of control. Another reason is the standard luck objection to ECL, which I set out, endorses and tweaked a little in my book last year.

Your other point, that more free will without more control sounds odd, is a good one. I choose modus tollens.

Thanks for this nice question, Al.
I think I'm basically in agreement with Chris and perhaps others, although I have a slightly different way of putting the point. I would distinguish two kinds of control: regulative control (which involves freedom to choose/do otherwise) and guidance control. Bob is right insofar as causal determinism is taken to rule out regulative control (a controversial, but not implausible position); thus, causal indeterminism at least opens up the possibility of more control--regulative, as well as guidance control. But as regards "the exercise of control", this might simply refer to the exercise of guidance control, and here indeterminism arguably does not give us more control.

So: Bob is right about regulative control, but Derk and Randy are right about guidance control.

Here is another place where perhaps at least it is helpful to distinguish the two importantly different kinds of control.

Al’s Question B concerns measuring control. In my view an agent exercises control *over* some state of affairs or event by exercising control *in service of* some relevant motivational attitude (paradigmatically, but not necessarily, an intention). So in many cases we have a straightforward way of measuring the exercise of control: an agent exercises control in service of an intention I to the degree that the behavior for which I plays a non-deviant causal role *matches* the representational content of I.

An agent could be set up such that her cognitive and action-production systems use indeterministic processes to maximize this type of control (think of exercising control in fleeing from a predator). But as Dennett argued in Elbow Room and elsewhere, the real issue here is predictability, not indeterminism.

So it seems this way to measure control does not straightforwardly apply to so-called regulative control.

This is a blog, so this will be quick and dirty. So hopefully Justin, Chris, Prof. Fischer and those on the side of ‘indeterminism=more control of some type (necessarily?)’ will cut me some slack. That said:

I’ve never understood why we should think of regulative control as a form of control. It is radically different from the forms of control studied by science (motor control, cognitive control, attentional control, self-control), and from the folk psychological form of control discussed in the above two paragraphs. And it makes little sense to think of agents exercising regulative control, as Prof. Fischer mentioned. So it seems we do have reason to favor Neil’s version of the free will/control split that emerges, rather than Chris’s.


I am inclined to agree with your claim that “if the question is simply about control and if the mechanism whereby control is exercised is explicated in terms of the causal role of some appropriate mental items, then whether the process of an agent's producing an action is indeterministic or deterministic seems to matter little.” But wouldn’t you agree that your antecedent is not CTA, but rather CTA plus some other (controversial) claims? If not, I’d definitely be interested to hear why you think your antecedent is simply CTA. Assuming you agree that your antecedent is more than CTA, then it is open to a libertarian to argue that control includes certain modal properties of the causal process (or broader situation). So what can a libertarian add: (specific) modal properties.

Alan and Neil,

I agree that opportunities may not always increase control. I meant to say what Justin said, though I was clear than him. Perhaps an opportunity only increases control if one already possesses the relevant ability. So giving Ann the opportunity to leave the room only increases her control if she has the ability to leave the room. Perhaps also there is a leveling-off point as well, where the addition of opportunities does not increase control. Also, the issue of indeterminism/determinism concerns the agent’s extrinsic, i.e. relational, features.


You claimed that Joan exercises “compatibilist control plus luck.” As I understood your argument, you contend that while Joan exercises compatibilist control in, say, jumping left, she does not exercise any control over whether she jumps left rather than right. Because of this it follows that she exercises compatibilist control plus luck. Have I understood your argument?

If so, then why not think that the “compatibilist control” Joan exercises over jumping left is sufficient for her exercising control over her jumping left rather than right?


Thanks for the question(s). First of all, regulative control is similar to, or identical to, the kind of control Bob Kane wishes to refer to by the term: "plural control" or "plural voluntary control". I'm not sure what's so mysterious about this! It is the kind of control we possess when we are said to have control "over" some outcome or action; it entails freedom to do otherwise. Insofar as control is a kind of freedom, it simply is the freedom one has when one has freedom to do one thing or another. What is so puzzling about this?

And I never mentioned "exercising regulative control", just "guidance control".

Finally, when "measuring" something, we can use different metrics, some of which are ordinal, some cardinal. When one has both regulative and guidance control, there is a clear sense in which (ordinally, at least) one has "more control". Again, I don't see what's agitating you so greatly about this--or about any of the above!

There are two kinds of freedom (or control). This idea is ancient--the Stoic Chrysippus talked of a dog that ran happily along with its master although it was on a leash which was such that, if it had tried to not run, it would have been dragged along anyway. And Locke talked of a man who is put in a room while asleep--Locke thought the man stayed in the room voluntarily, although the door was locked. And so on to the Frankfurt cases. The ancient tradition has it that there are two kinds of freedom or control. What's so puzzling about this? And why would it be mysterious to say that when an agent has both sorts of control, there is a sense in which she has more control?

As Andrei notes, Bob thinks that in his indeterministic picture control is diminished in one respect; Bob says: “indeterminism is functioning as a hindrance or obstacle to [the businesswoman’s] realizing one of her purposes – a hindrance or obstacle in the form of resistance within her will which has to be overcome by effort” (Four Views, p. 178). But at the same time, he thinks that adding in the indeterminism in the way he specifies enhances control in another respect, since it provides for plural voluntary control. The threat to this position as I see it is that the hindrance to control that indeterminism yields will preclude the sort of plural voluntary control intuitively required for moral responsibility in the basic desert sense (although this requirement is challenged by Frankfurt examples).

I think that the threat is made clear by the following argument. Consider a decision made in the kind of context Bob specifies in his businesswoman example -- in which moral reasons favor one action, prudential reasons favor a distinct and incompatible action, and the motivational strength of these sets of reasons are in close competition. On an event-causal libertarian picture, the relevant causal conditions antecedent to the decision – agent-involving events -- would leave it open whether the decision will occur, and the agent has no further causal role in determining whether it does. With the causal role of the antecedent events already given, whether the decision occurs is not settled by any causal factor involving the agent. In fact, given the causal role of all causally relevant antecedent events, nothing settles whether the decision occurs. I call this the disappearing agent objection, and to me it seems to show that while the indeterminism in the event-causal context allows different options for decision and action to be causally open, it precludes the plural voluntary control intuitively required for moral responsibility. This argument lends support to the views Andrei, Neil, and Robert express earlier in the thread.

The event-causal libertarian might try to solve the problem by specifying that the event- or state- causal etiology of a decision for which an agent is morally responsible must feature agent-involving states such as values or standing preferences, thereby making it intuitive that it’s the agent who settles whether the decision occurs. For instance, in Laura Ekstrom’s version of event-causal libertarianism (2000 book, 2003 article), a decision for which an agent is morally responsible must result by a normal causal process from an undefeated authorized preference of the agent’s, where such preferences are noncoercively formed or maintained, and are caused but not causally determined by considerations brought to bear in deliberative processes. In her view, these conditions on the formation of preferences intuitively tie them to who the agent is, and also preclude causal determination of preferences by factors beyond the agent’s control. One might argue that on this conception the agent does settle which decision is made.

But suppose again that in our morally charged situation, the agent can decide either morally or out of self-interest, and considerations in favor of each option arise in her mind, and are in motivational equipoise. She, like most people (almost everyone, I would say), has both moral and self-interested undefeated authorized preferences. We can now ask: with the motivational equipoise in place, what is it that settles whether her self-interested or her moral preference will be causally efficacious, and thus whether the self-interested or moral decision subsequently occurs? It seems that Ekstrom can only say that when one of the preferences is efficacious, this happens without anything about the agent settling that it is, since the extent to which the agent is involved at this point is exhausted by the preceding considerations, which by hypothesis are in equipoise. Crucially, adding in the agential requirements on formation of preferences does not make it intuitive that the agent settles which decision occurs, as would seem to be required if she is to have the control required for morally responsibility for her decision. So it appears that this sort of move will not help the event-causal libertarian’s case for the right kind of plural voluntary control.

On Chris’s point about increasing opportunities enhancing control: if the agent can’t settle which of the opportunities is actually realized, then even if her control is enhanced, it isn’t enhanced enough to secure basic-desert moral responsibility. And while John may be right to say that indeterminism may open up the possibility of regulative control, given incompatibilist intuitions, in the event-causal context it won’t be enough regulative control for basic-desert moral responsibility. On Josh’s last comment, it’s my sense that it wouldn’t be unnatural to call an agent’s plural voluntary ability to settle which decision is made a kind of control, and when an agent settles which one of a number of possible decisions actually occurs, it seems right to say that she then exercises a kind of control.

Chris, unlike Derk (in his comment), I take opportunity availability in this sense to reduce control. Because Joan does not control whether she jumps left or right, she exercises less than compatibilist control in jumping left. She exercises compatibilist over the following fact: that it is open to her to jump left or right. But she has no further control. Jane exercises control in jumping left; Joan doesn't, except insofar as she exercised control over having these options.

John: so if the physicists from MIT deliver the news that determinism is false, we will learn that we are freer than we might have thought (though no more responsible)?


Thanks for the question. Well, it would depend on what form the indeterminism takes. As you know, mere indeterminism doesn't itself entail that the libertarian elements would be in place to secure "control", rather than mere randomness. But, yes, if the MIT physicists delivered that very interesting news, we would learn that we might have both regulative and guidance control. Of course, we already know we don't need regulative control for moral responsibility, so we haven't learned anything that would make us more sanguine about moral responsibility, at least in my opinion.

Derk’s example involving motivational equipoise and the questions that he raises nicely illustrates just how crucial the notion of agent causation is in this context. As I understand the way he puts it, if we are assuming that some event-causal account of action is correct and that motivational equipoise is in place, we cannot make sense of how an agent can be causally efficacious when she is making a decision or choice, even if we include her pre-existing preferences. The reason why we cannot make sense of this, it seems to me, is because we do not secure a causal role *for her* when she is making her own decision or choice, that is, when she is performing such mental actions.

To try to put the point slightly differently: we might want to articulate a notion of “control” that is MORE BASIC than plural voluntary control, regulative control, or guidance control, as the latter notions have been used above. This more fundamental notion of control might be related to those mentioned by Josh Shepherd, namely, the basic kind of control that an agent has when exercising her own cognitive capacities, such as when one directs the focal point of attention, say. Plausibly, this more basic notion of control makes explicit the causal role of the agent in her performance of such mental actions: in such cases, she is in control of her own cognitive capacities in the sense that she is exercising or employing them during the performance of such mental actions.

The comments to this entry are closed.

Books about Agency

3QD Prize 2014: Marcus Arvan