1. Ultimate Responsibility and Free Will
The first question about free will that I believe requires a thorough rethinking is the Compatibility Question: “Is free will compatible or incompatible with determinism?”
Most historical and modern debates about this question have focused on the requirement that free agents must have “open” alternatives or alternative possibilities for action (as in the garden of forking paths). Free agents, it is said, must have the ability or power to act and the ability or power to act otherwise. So when they do act, we can say that they “could have done otherwise.” This “could have done otherwise” condition is now often referred to as the “condition of alternative possibilities,” or AP, for short.
Arguments about whether determinism is compatible with this condition of alternative possibilities or AP have led to contentious debates in modern philosophy that have tended to stalemate over differing interpretations of what it means to say that persons have the ability or power to act and act otherwise and about the meaning of the expression “could have done otherwise.” I have argued that these persistent stalemates are symptoms of a deeper problem—namely that focusing on alternative possibilities alone is too thin a basis on which to rest the case for the incompatibility of free will and determinism.
It’s not that alternative possibilities and forking paths are unimportant for free will. Far from it. They are very important; and I’ll return shortly to explain why. It’s just that other considerations must also be brought into the picture in arguing for the incompatibility of free will and determinism: The Compatibility Question cannot be resolved by focusing on alternative possibilities or the power to do otherwise alone.
Fortunately, there is another place to look. In the long history of free will debate one can find another criterion fueling incompatibilist intuitions about free will, a criterion that I argue is even more important than AP.
I call it the condition of ultimate responsibility or UR. The basic idea is this: To be ultimately responsible for an action, an agent must be responsible for anything that is a sufficient reason (condition, cause or motive) for the action’s occurring. If, for example, a choice issues from, and can be sufficiently explained by, an agent’s character and motives (together with background conditions), then to be ultimately responsible for the choice, the agent must be at least in part responsible by virtue of choices or actions voluntarily performed in the past for having the character and motives he or she now has.
Compare Aristotle’s claim in his Nichomachean Ethics that if a man is responsible for wicked acts that flow from his character, he must at some time in the past have been responsible for forming the wicked character from which these acts flow.
This condition UR is no merely esoteric principle. It is deeply woven into the fabric of our ordinary reasoning about responsibility in everyday moral and legal contexts. If a drunk driver kills a pedestrian, and it could be shown that given the circumstances, including the visibility on the road, the condition of his nervous system, given the alcohol, he could not possibly have avoided hitting the pedestrian, that alone is not exonerating. One wants to know whether he is responsible for any of the prior circumstances that placed him in a position where he could not possibly have avoided hitting the pedestrian, such as his prior decisions to drink and drive, etc. He may still be responsible for putting himself in this position by virtue of earlier choices or voluntary acts, even if he could not have done otherwise at the time. This is in fact how we treat such cases in courts of law and in ordinary reasoning about responsibility.
This condition of Ultimate Responsibility or UR fits this pattern. It does not require that we could have done otherwise (AP) for every act done of our own free wills. But it does require that we could have done otherwise with respect to some acts in our past life histories by which we formed our present characters or wills. I call these earlier acts by which we formed our present characters or wills “self-forming actions,” or SFAs.
2. Self-forming Actions, Self-formation and the Will
To see why such self-forming acts are important for free will, consider a well-known example about Martin Luther offered by Daniel Dennett. When Luther finally broke with the Church in Rome, initiating the Protestant Reformation, he said “Here I stand, I can do no other.” Dennett asks us to suppose that at the moment Luther made this statement, he was literally right. Given his character and motives at the time, he could not then and there have done otherwise.
Does this mean Luther was not morally responsible, not subject to praise or blame, for his act, or that he was not acting of his own free will? Dennett says “not at all.” In saying “I can do no other,” Luther was not disowning responsibility for his act, according to Dennett, but taking full responsibility for acting of his own free will. So the ability to do otherwise (“could have done otherwise”) or AP, says Dennett, is not required for moral responsibility or free will; and so there is no conflict between free will and determinism. They are compatible.
My response to Dennett is to grant that Luther could have been responsible for this act, even if he could not have done otherwise then and there and even if his act was determined. But this would be so, I argue, only to the extent that Luther was responsible for his present motives and character by virtue of many earlier struggles and self-forming choices (SFAs) that brought him to this point in his life where he could do no other. Luther’s biography gives ample evidence of the inner struggles and turmoil he endured getting to that point in his life.
And so it frequently is with self-formation and free will generally. Often we act from a will already formed, but it is “our own free will” by virtue of the fact that we formed it by other choices or actions in the past (SFAs) for which we could have done otherwise. If this were not so, there is nothing we could have ever done differently in our entire lifetimes to make ourselves different than we are—a consequence, I believe, that is incompatible with our being (at least to some degree) ultimately responsible (UR) for being the kinds of persons we are. So SFAs are only a subset of those acts in life for which we are ultimately responsible and which are done “of our own free will.” But if none of the acts in our lifetimes were self-forming in this way, we would not be ultimately responsible for forming the wills from which we act.
Focusing on this condition UR also tells us something else of importance about the traditional problem of free will. It tells us why that problem is about the freedom of the will and not just about the freedom of action. There has been a tendency in the modern era, since the 17th century, and coming to fruition in the 20th, to reduce the problem of free will to a problem of free action. I argue that such a reduction oversimplifies the problem. Free will is not just about free action, though it involves free action. It is about self-formation, about the formation of our “wills” or how we got to be the kinds of persons we are, with the characters, motives and purposes we now have.
Were we ultimately responsible to some degree for having the wills we do have, or can the sources of our wills be completely traced backwards to something over which we had no control, such as Fate or the decrees of God, or heredity and environment or social conditioning or hidden controllers, and so on? Therein, I believe, lies the core of the traditional problem of “free will.”
Finally, if the case for incompatibility of free will and determinism cannot be made by appeal to AP alone, it can be made if UR is added. If agents must be responsible to some degree for anything that is a sufficient condition, cause or motive for their actions (as UR requires), then an impossible infinite regress of past actions would be required unless some actions in the agent’s life history (SFAs) did not have either sufficient conditions, causes or motives in their past (and hence were undetermined).
Therein lies the connection between UR and determinism. Events are determined when there are circumstances in their past (whether the decrees of Fate, or God, heredity or environment, social circumstances, etc.) that are sufficient conditions or causes for the occurrence of the events. Given these past circumstances, the determined events cannot but occur.
If our present wills (characters, motives and purposes) were determined in this way, then UR would require that we must have played some role in forming them by earlier voluntary choices or action. But if these earlier choices or actions were also determined by sufficient conditions or causes in their past, then UR would require that we must have also been responsible for some of those earlier sufficient conditions or causes by virtue of forming them by still earlier voluntary choices or actions, and so on backwards indefinitely into our past. Eventually we would come to infancy or to a time before our birth when we could not have formed our own wills.
The only way to stop this regress is to suppose that some responsible choices or actions in our life histories must not have been determined by sufficient conditions or causes in their past, and hence must has been undetermined, if we are to be the ultimate sources of, and hence ultimately responsible for, our own wills. These undetermined regress-stopping acts would be the “self-forming acts” or SFAs that are required by UR sometime in our lives, if we are to have free will in the sense of a will “of our own free making.”
3. Alternative Possibilities (AP) and UR
But if one can arrive at the incompatibility of free will and determinism in this manner by reflecting on the condition of ultimate responsibility or UR alone, does this mean that alternative possibilities or AP have nothing to do with free will?
Some contemporary incompatibilists and libertarians about free will have been tempted to think that one could dispense with the requirement of alternative possibilities for free will altogether and go with an ultimate “source” condition like UR alone. But I think this is a mistake. For it turns out that there is a deep connection between AP and UR; and understanding that connection is crucial, I believe, to fully understanding the nature of free will.
To see the connection between AP and UR, we must first explore another reason why focusing on alternative possibilities alone is too thin a basis for resolving the Compatibility Question.
For it turns out that having alternative possibilities for action is not enough for free will, even if the alternative possibilities should also be undetermined. One can see this by noting that there are examples, many of them, in which agents may have alternative possibilities that are also undetermined, and yet the agents lack free will. Such examples were first put forward by philosophers such as J. L. Austin, Phillippa Foot and Elizabeth Anscombe in the 1950s and 60s.
Here are three such examples. The first is one of Austin’s. He imagined that he had to hole a three-foot putt to win a golf match, but owing to a nervous twitch in his arm, he misses the putt. The other two examples are mine. An assassin is trying to kill the prime minister with a high-powered rifle when, owing to a nervous twitch, he misses and kills the minister’s aide instead. I am standing in front of a coffee machine intending to press the button for black coffee when, owing to a brain cross, I mistakenly press the button for coffee with cream.
In each of these examples, we can suppose, as Austin suggests, that an element of genuine chance or indeterminism is involved. Perhaps the nervous twitches or brain crosses were brought about by actual undetermined quantum events in our nervous systems. We can thus imagine that Austin’s holing the putt is a genuinely undetermined event. He might miss the putt by chance and, in the example, does miss it by chance. Likewise, the assassin might hit the wrong target by chance and I might press the wrong button by chance.
Now Austin asked the following question about his example: Can we say in the circumstances that he (Austin) “could have done otherwise” than miss the putt? Did he have alternative possibilities? Austin’s answer is that we can indeed say he could have done otherwise than miss it. For he was a good putter and had made many similar putts of this short length in the past. And even more important, since the outcome of this putt was genuinely undetermined, he might well have succeeded in holing the putt, as he was trying to do.
But this means we have an action (missing the putt) that is undetermined and such that the agent could have done otherwise. (He might have holed it.) Yet missing the putt is not something that we regard as freely done in any normal sense of the term because it was not within his voluntary control. It was not what he wanted to do nor what he was trying to do. The same is true of the assassin’s missing his intended target and killing the aide and my accidentally pressing the wrong button on the coffee machine.
One might be tempted to think that these three occurrences (missing the putt, killing the aide, pressing the wrong button) are not actions at all in such circumstances because they are undetermined and happen by accident. But Austin correctly warns against drawing such a conclusion. Missing the putt, he says, was clearly something he did, even though it was not what he wanted or chose to do. Similarly, killing the aide was something the assassin did, though unintentionally; and pressing the wrong button was something I did, even if only by accident or inadvertently.
The general point here is that many of the things we do by accident or mistake, unintentionally, inadvertently or unwillingly are nonetheless things we do. We may sometimes be absolved of responsibility for doing them (though not always, as in the case of the assassin). But it is for doing them that we are absolved of responsibility; and this can be true even if the accidents or mistakes are genuinely undetermined.
4. An Imagined World and Plurality Conditions
To see what this implies about free will, consider the following scenario. Suppose God created a world in which there was a considerable amount of indeterminism or chance in human affairs as well as in nature. In this world people set out to do things—kill prime ministers, hole putts, press buttons, punch computer keys, etc.—usually succeeding. But sometimes they fail by mistake or accident in ways that are undetermined, as in the preceding examples.
Now, further imagine that all actions in this world, whether the agents succeed in their purposes or not, are such that their reasons, motives and purposes for trying to act as they do are always predetermined or pre-set by God. Whether the assassin misses the prime minister or not, his intention to kill the prime minister in the first place is predetermined by God. Whether or not Austin misses his putt, his wanting and trying to make it rather than miss are preordained by God. Whether I press the button for black coffee, my preferring it to coffee with cream is predetermined by God; and so it is for all persons and all of their actions in this imagined world. Their motives and purposes for acting as they do are always predetermined by God.
I would argue that persons in such a world lack free will, even though it is often the case that they can do otherwise (and thus have alternative possibilities) in a way that is undetermined. The reason is that while they can do otherwise, it is only in this limited Austin-style way—by mistake or accident, unwillingly or unintentionally. What they cannot do in any sense is will otherwise than they do; for all of their reasons, motives and purposes have been pre-set by God. Their wills are always already “set one way” before and when they act, so that if they do otherwise, it will not be “in accordance with their wills.”
The lesson here is that when we wonder about whether the wills of agents are free, it is not merely that they could have done otherwise that concerns us, even if the doing otherwise is undetermined. What interests us is whether they could have done otherwise voluntarily or willingly, intentionally and rationally. Or, more generally we are interested in whether they could have acted voluntarily, intentionally and rationally in more than one way, rather than in only one way, and in other ways merely by accident or mistake, unintentionally, inadvertently or irrationally.[1]
I call such conditions –of more-than-one-way, or plural, voluntariness, intentionality and rationality—“plurality conditions” for free will. They are deeply embedded in our intuitions about free choice and action. Most of us naturally assume that our freedom and responsibility would be deficient if it were always the case that we could only do otherwise by accident or mistake, inintentionally, involuntarily or irrationally. If free will involves more than alternative possibilities and indeterminism, these plurality conditions would appear to be among the additional requirements. But why is this so, and why are these plurality conditions so deeply embedded in our intuitions?
5. Will-setting
To answer these questions we have to consider another important, though often neglected, topic in free will debates that may be called “will-setting.” In the imagined world described earlier, all the motives and purposes of agents in every situation were already pre-set or “set one way” by God. Another way to put this is to say that all the will-setting in this world was done by God, rather than by the agents themselves—even though the agents could sometimes have done otherwise.
Actions are “will-setting” when the wills of agents (their motives and purposes) are not already “set one way” before they act (as the assassin’s will is set on killing the prime minister). Rather the agents themselves set their wills one way or the other in the performance of their actions.
Choices or decisions, which are self-forming actions or SFAs in the sense defined earlier, are “will-setting” in this sense. The agents’ wills are not already set one way before they choose, but they set their wills, one way or the other, voluntarily and intentionally, in the act of choosing. Will-setting choices are in this sense will-settling rather than being (already) will-settled.
We can now see the connection between will-setting and the plurality conditions. When the will is already set one way, as in the case of Austin or the assassin, or me at the coffee machine, the action is one-way voluntary. If the agents were to do otherwise than what they are intending or trying to do, it would not be voluntarily, but only by accident or mistake (i.e., unwillingly). In the case of will-setting actions, by contrast, the actions are voluntary either way the agents act—hence they are no plural voluntary. Will-setting actions are also plural rational since the agents make the reasons for preferring one of the options prevail by choosing it, whichever option is chosen. And if we assume for genuine cases of will-setting, as I think we should, that the agents know what they are doing and are doing it on purpose, then will-setting actions will be plural intentional as well.
We now have an answer to the question of why the plurality conditions are important for free will. They follow from the requirement that if we are ever to be creators of our own wills, some actions in our life histories must be will-setting and not already will-settled. At these moments, we must be able to go in different directions voluntarily, intentionally and rationally (i. e., willingly) either way we choose.
We now have a sequence of connected notions: (1) acting “of one’s own free will” (free will), to (2) being ultimately responsible for the will from which one acts (UR), to (3) will-setting (setting one’s will in one way or another in the process of acting itself) to (4) the plurality conditions (being able to act in more than one way, voluntarily, intentionally and rationally. Each of these notions in turn implicates the others; and all, I believe, are required to account for freedom of the will as something more than mere freedom of action.
Moreover, we can now add another notion to the sequence. For if (4) the plurality conditions obtain, it follows immediately that the agent (5) “could have done otherwise” (i.e., that the condition of alternative possibilities or AP is satisfied). For if you are able to do otherwise voluntarily, intentionally and rationally, then you are able to do otherwise. And if you are able to act in more than one way voluntarily, intentionally and rationally, then you are able to act in more than one way. You have alternative possibilities.
Therein lies the connection between free will, UR and AP we were looking for. Free will requires both ultimate responsibility, UR, and alternative possibilities, AP, at least for some choices or actions in our lifetimes (SFAs) by which we form our own wills. But note that the connection between (1) free will and (5) AP or the power to do otherwise, on this account, is not direct. It goes through other notions, such as (2) ultimate responsibility, (3) will-setting and (4) plurality conditions.
That is one reason why discussions of free will and alternative possibilities often lead to unresolved debates about whether “could have done otherwise” really is incompatible with determinism or whether free will really requires “could have done otherwise.” Intuitions will tend to conflict on these matters as long as these other notions are not taken into account.
6. The Dual Regress of Free Will
If this reasoning is correct, it would appear that alternative possibilities or AP our needed for free will after all, at least sometimes in our lives. Yet we saw earlier that one could argue for the incompatibility of free will and determinism from UR alone without appealing to AP. How are we to make sense of this?
The answer lies in UR. Both AP and indeterminism follow from UR, but by different argumentative routes. I call this “the dual regress of free will.” Two separate regresses are associated with UR. The first begins with the requirement that agents be responsible to some degree by virtue of past voluntary actions for anything that is a sufficient ground or reason for their actions in the sense of a sufficient cause; and it leads to the conclusion that some actions in the life histories of agents must be undetermined (lack sufficient causes).
The second regress begins with the requirement of UR that agents be responsible for anything that is a sufficient ground or reason for their actions in the sense of a sufficient motive; and it leads (by way of will-setting and plurality conditions) to the conclusion that some actions in the life histories of agents must be such that the agents could have done otherwise—i.e., to AP.
Agents have sufficient motives for their actions when their wills are already “set one way” on acting as they do before and when they act—as the assassin’s will is set on killing the prime minister. UR requires that if the assassin’s will is set on killing the prime minister, then the assassin himself must be to some degree responsible for his will’s being set that way by virtue of earlier choices or actions which were such that his will was not already set one way when he performed them. It cannot be that his will was entirely set that way by Fate or God, genes and environment, social conditioning or other factors over which, if you go back far enough, the assassin himself had no control.[2]
The first of these regresses results from the requirement that we be ultimate sources of our actions, the second from the requirement that we be ultimate sources of our wills (motives and purposes) to perform those actions. In this manner, the requirements of indeterminism and alternative possibilities have a common origin in the idea that we must be to some degree ultimate sources of our willed actions.
It can further be shown that the two regresses converge.[3] The undetermined actions needed to stop one regress are the same as the will-setting and plural voluntary actions needed to stop the other. The convergent actions that result are the “self-forming” actions, SFAs, needed for us to be to some degree creators of our own wills, or to have free will.
[1] “Voluntarily” or “willingly” mean here acting “in accordance with one’s will” (motives and purposes), “intentionally” means “knowingly” (as opposed to “inadvertently”) and “on purpose” (rather than “accidentally” or “by mistake”) and “rationally” means “having reasons for so acting and acting for those reasons.”
[2] It is important to recognize here that sufficient motives for action need not be sufficient causes of those actions. (See my 1996, 2000.) One can see this in the cases of the assassin or me at the coffee machine. Agents have sufficient motives for their actions when their wills are already “set one way” on acting as they do before and when they act—as the assassin’s will is set on killing the prime minister and mine on pressing the button for black coffee. But having such motives does not guarantee or determine that the actions will be performed. Indeed in the imagined examples, the actions themselves, killing the prime minister and pressing the button for black coffee, were undetermined and hence lacked sufficient causes. Yet the agents had sufficient motives for performing them.
[3] I show this in more detail in my book The Significance of Free Will (1996) and in the article “The dual regress of free will and the role of alternative possibilities” (in Philosophical Perspectives (2000): 57-80)), among other places. One can get a glimpse of this convergence, however, by noting that self-forming actions or SFAs lack sufficient prior causes because they are not determined. And they lack sufficient prior motives as well because the agents’ wills are not already “set one way” before they perform them (but rather the agents “set their wills” one way or the other in the performance of the self-forming actions themselves).