FREEDOM AND THE POWER OF PREFERENCE
-- the Determinism and Freedom Philosophy Website --
Professor Lehrer, for the best of reasons, turns up in all the good surveys of philosophical opinions on the large and controverted subject of this website. He has been a clear-headed and forward-looking defender of the view given to the world by such great antecedents as Thomas Hobbes and David Hume. This paper is the latest stage in his defense of Compatibilism. Determinism goes with and does not conflict with freedom. The paper refers to earlier work having to do with competing analyses of that seemingly crucial sentence about freedom 'He could have done otherwise'. The paper goes well beyond this by way of an original idea of preference. It demonstrates that the last word has not been said either for or against Compatibilism -- or for sceptical views of both traditions of thought.
It is the power of preference that leads us to think and aver that things are up to us or in our power. Our preferences are the source of our freedom. But we must have our preferences because we prefer to have them and not because they are imposed upon us. We have reasons for our preferences. Our reasons for our preferences must be our reasons because we prefer them to be. We must be agent and author of our preferences for them to be free. Our preferences, however, seem to be part of the causal order. So how can we have our preferences because we prefer to have them? How can things be reasons for our preferences because we prefer that they are? How can we be the agent and author of our preferences, when our preferences are part of the causal order? How, on the contrary, can we have preferences because we prefer to have them if they are not part of the causal order? When we reflect on how preferences empower us, we understand that we must empower them. But how? These are the questions I shall seek to answer. The keystone to the answer is a kind of special preference that loops back onto our preferences including itself.
There is a minimal notion of freedom expressed by “can” and “could have”. It goes this way. I do something freely if and only if I could have done otherwise. Extending the minimal notion of freedom to preference, I prefer something freely if and only if I could have preferred otherwise. The appropriate meaning of “could have” required to articulate our freedom cannot be assumed to be contained in antecedent usage, however. On the contrary, we must construct the requisite meaning needed for an adequate theory of freedom.
Let me return to a simple analysis, recognizing that this is theory construction, not semantics. We are trying to construct, then, an analysis of “can” and “could have” that will enable us to build a theory of freedom. To accomplish this, I shall need a notion of preference that I have advocated. You may wish to substitute some notion of preference you consider more satisfactory. I distinguish between desire and preference. Desires I think of as something that arises in us without our bidding and often against our will. I suddenly desire to eat some chocolate as I am writing. Should I stop writing and get some chocolate? I have Ghirandelli semi-sweet chocolate in the kitchen. I have also eaten quite bit of this chocolate today. Then I have a desire to look at my paintings. I am a painter, and I like to look at my paintings. But once I do, I absorb myself in them for a while. So, here I am trying to write distracted by desires to eat chocolate and look at my paintings. What a nuisance these desires are! They suddenly appear, these uninvited distractions. Go away, I want to write! But they do not go away. I prefer not to satisfy them, and, in fact, I continue to write. What I prefer is up to me. What I desire is not.
I evaluate my desires and prefer to satisfy some and not others. So preference is typically the positive evaluation of something desired. But not always. Sometimes I consider some course of action and prefer to undertake that course of action without any desire to do so. I prefer to prepare the materials to pay my taxes, but I have no desire to do so. None whatever. It is rather, as I consider the consequences of preparing or not preparing the materials to pay my taxes, I evaluate preparing the materials as preferable to not doing so, and I prefer to prepare them without any desire to do so. There are many cases of odious or burdensome obligations. Though I prefer to do what I am obliged to do, I have no desire for it. What this reveals is that preference is a different kind of state than desire, though both may influence what I do. I have proposed that preferences are states of a different level of mentality. So, I think of desire as arising without my bidding or approval, the states of a first level mind that are not up to me, as contrasted with preference which are up to me. What I prefer as I evaluate what I desire or consider, either reflectively or spontaneously, at the level of the metamind, is where optionality and freedom enter the picture. Preference like choice, which is the immediate expression of preference, is a condition of our freedom.
This discourse could become fully and completely a discussion of preference and what it is like. I have, however, discussed preference elsewhere and so, having noted the distinction between desire and preference, it is the application of the notion in a theory of freedom that will be my focus.
Some have suggested that “I can do X” might be analyzed conditionally as “I will do X if I prefer to do X” and “I could have done X” as “I would have done X if I had preferred to do X.” Now, in the normal case, my preferences will be up to me. Consequently, analyzing “I can” and “I could have” in this way will yield the result that when I can do something in the way analyzed, then it is in my power to do it. Further, if I both can do something and can do otherwise instead, then it is up to me whether I do it. But that result is only a material consequence, for it is possible that what I prefer is not in my power, not up to me. Consequently, we have not conceptually insured even a minimal doctrine of freedom with the satisfaction of the conditionally analyzed notion of “can” and “could have”.
Before rushing too quickly to the conclusion that the analysis is unsatisfactory, it is important to notice the importance and insight of it. It is the kind of freedom that we lack when in chains or imprisoned. It is the kind of freedom that people have fought and died for. If it is not all that there is to freedom, it is a great deal. Moreover, since preference reflects our evaluation of options, it is typically and normally an expression of choice. Whether I choose X or not-X typically depends on what I prefer. So, in the typical case, the conditional analysis of “can” and “could have” will capture a notion of freedom in a way that a conditional analysis in terms of desires would not. A person driven hither and thither by desire may appear more passive than active, more controlled than empowered. But preference is different. What we prefer reflects our evaluations of what is preferable, desirable, good for ourselves, and, ideally, good for others. Preference amalgamates these evaluations in a choice disposing state.
However, even if the conditional analysis gives us most of what we value in freedom and perhaps most of what makes us uncomfortable as we deal with our freedom, a problem remains. Our preferences might controlled by another and not be in power or up to us. We must say something about freedom of preference as well as freedom of action based on the conditional analysis of freedom.
I suggested in earlier work, influenced by Frankfurt, that we iterate the conditional analysis to provide an account of freedom of preference. Putting the matter in terms of “can”, we might add to the analysis of “I can do X” as “I will do X if I prefer to do X” the further conditional that “I will prefer to do X if I prefer to prefer to do X”, or to introduce a notation reflecting levels of concatenation of preference “I will prefer (1) to do X if I prefer (2) to do X”. Thus, “I prefer to do X”, expresses first level preference, and in our level specified notion becomes, “I prefer (1) to X”, while “I prefer to prefer to do X”, expresses second level preference, and in our notation becomes, “I prefer (2) to do X.” More generally, “I prefer (n+1) to do X” is equivalent to, “I prefer to prefer(n) to do X”. To avoid a problem arising about freedom at level (2) or beyond, we might add the perfectly general conditional, “I will prefer (n) to do X, if I prefer (n+1) to do X” requiring that it hold for all levels of n. There is no vicious regress in the condition, since I do not have to do anything for the set of conditionals to be true. They can all be true. Moreover, there is no problem in principle of processing such higher order conditionals since, as the general conditional reveals, to process up a level I would only need to process whether I prefer what I preferred at the previous level, and the answer might be obviously, “Yes.” We are capable of processing things computationally, a step at a time, without grasping the whole sequence in a flash of intuition.
However, as some have correctly pointed out, Van Inwagen and Segerberg, it might be that someone controls what I prefer at all levels and what I would prefer at all levels if my preferences were different so that my preference at all levels are in his power and not in mine. Our detractor who will introduce a demon or manipulator imagined to control my preference, perhaps by inserting a device, call it, as I have, a braino, in my brain so that when he types “pX(1)” on his computer, I prefer (1) to do X and the same for any other level. I have two comments. First, there are no such manipulators. Once brainos started getting produced and installed, we shall have a problem about our freedom which we do not have at the moment. So, the analysis may have the virtue of being materially adequate and illuminating in our current state of technology. It could reveal the way in which our preferences can be in our power or up to us. They can be what we prefer them to be.
However, it is important to go beyond the actual case to obtain further illumination and security for our analysis against preference control. Moreover, the way to accomplish this is both clear and avoids the necessity of postulating an infinite hierarchy of preferences, even hypothetically. What is needed for a satisfactory analysis is the satisfaction of two conditions. The first is that I prefer to have the preferences I do have concerning X however far up the preference hierarchy they might extend or not extend. The second is that this preference to have the preferences I have is one that I have, not because it is manipulated by another, but just because I prefer to have it. Let us call the preferences we have at various levels concerning doing X my preferences structure for X. Now consider my preference to have my preference structure for X. Let us call that my power preference for X. I propose as one condition of freedom of preference that I prefer X and I have my preference structure for X because I prefer to have it.
There are special features of the power preference and of my having that preference because I prefer to have it. First of all, my power preference for having my preference structure for X is itself a preference in that preference structure because it is itself a preference concerning X. Thus, it is a preference for having a preference structure which includes that preference. The preference refers to itself as it is a preference for itself as well as other preferences in the structure. It is, therefore, an unstratified or level ambiguous preference. On one standard account of level ambiguous states, they are semantically ungrounded. Ungroundedness may be considered as a defect in the truth conditions for a sentence, but it is not a defect of a state intended to analyze freedom of preference.
Secondly, the preference ends a possible regress of preferences with a loop. Theories of higher order preference that appear to require a complete hierarchy of preferences are often criticized on the ground that the requirement is psychologically unrealistic. We have noted that if preferences over preferences are considered computationally or functionally, it is not clear that any difficulty in principle results from such requirements. However, the addition of preferences past the second or third level, though it can be processed computationally, might lack efficacy. So, there is an advantage in not requiring a hierarchy of preferences, and use of the power preference allows us to dispense with that requirement. The power preference insures that any preference that one has at any level is one that one prefers to have, for that is what the power preference is a preference for. It allows, in principle, that freedom of preference could result from a first order preference for X and the unstratified power preference for X which would, in that case, be a preference for the first order preference and for itself.
Moreover, the present account allows for the possibility that preferences which insure freedom might contain conflict between levels of preference. Frankfurt’s account and my own earlier account indebted to him require that the preferences insuring freedom be homogeneous in the sense that every preference in the structure be a preference for the preference of the preceding level. Such accounts require a conflict free preference structure concerning X, while the present proposal allows for the possibility of freedom in spite of conflict. I might prefer at the second level not to have the preference at the first level that I have. As I consider the resulting preference structure with the preference for X and the preference for not having that preference, however, I might find that I prefer to have that preference structure, including the conflict, and thus have a power preference for having that preference structure for X. The conflict might strike me as representing moral or just prudential complexity. I might, for example, prefer to spend my resources on a friend who is hardly in need rather than on a stranger who is in acute need. After some struggle within, I might still prefer to spend my money on my friend, but admiring the nobility of people having broader humanitarian concerns prefer not to have that preference. As I consider that preference structure, I might prefer to have it, both because I prefer to spend money on my friend and also prefer to have different preferences expressing a nobler perspective. I prefer to have the preference structure that I have including the conflict at the second level rather than repudiating the conflict for the felicity of a homogeneous preference structure. Freedom of preference is, on the account I am proposing, compatible with conflict within the structure.
Of course, it is essential that I not only have the power preference for X but that I have it because I prefer to have it. Otherwise, it may be that I have the preference structure for X because I am manipulated by the braino rather than because I prefer to have it. This requirement, which I shall call the primacy condition, should be understood to insure that the primary cause or explanation of the preference structure is the preference for the preference structure. There may, of course, be secondary causes or explanations. For example, suppose that you are the friend and prefer that I spend my money on you. If that is the primary cause or explanation of why I prefer to spend money on you, there is no reason to conclude that I have freedom of preference. I may be in your power, and my preferences as well, and so it is in your power what I prefer, it is up to you what I prefer, and I lack freedom. I am in your control.
However, there is another possible manner in which your preference might influence my preferences which is compatible with my freedom of preference. I might prefer to do what you prefer in this instance because I prefer to have that preference structure. The primary cause or explanation of why I prefer what I do is that I prefer to have the preference structure I have, in this case, including a preference for what you prefer. It is not in your power or up to you what I prefer. It is in my power or up to me what I prefer, though, in fact, my preferences include a preference for what you prefer. Who has the power in the matter depends on whose preferences are the primary cause or explanation of my preferences. I am free in my preference for what you prefer if I prefer what you prefer because I prefer to have that preference. I am in bondage to you if I prefer what you prefer just because you prefer what you do. In short, if your preference is effective because I prefer that I have the preference structure I do and prefer what you prefer, then I am free and my preferences are in my power, while if your preference is effective just because you prefer what you do, then I am bound by your preferences and in your power.
Once we have in mind the distinction between the case in which the power preference is the primary cause or explanation of the preference and the case in which it is not, we are in a position to deal with a central issue, that of the compatibility of causation and freedom. Moore noted a long time ago that, though there is a sense of “could” which yields the result that under the assumption of determinism nothing ever could happen other than it did, it remains an open question whether this sense of “could” is the one that is involved when, claiming that we are free, we avow that we could have done or preferred otherwise. The question of the compatibility of freedom and determinism remains open, and a sufficient condition for freedom that is compatible with determinism can be based on what has been said above.
To clarify the issue of compatibility, it is important to be clear about the logic of compatibility. It will suffice for the purposes of the compatiblist to provide a set of sufficient conditions for freedom that are compatible with the thesis of determinism. The principle is that if c entails f and c is consistent with d, then f is consistent with d. This shows that we do not need to concern ourselves with the question of whether a condition is necessary for freedom in order to employ the condition in an argument for the compatibility of freedom and determinism. That is important for two reasons. In the first place, the necessary condition for freedom of action or preference might be some rather heterogeneous disjunction of conditions which is not very perspicuous. Secondly, and more important, it appears from the discussion above that we can formulate a sufficient condition for freedom of action and preference in terms of the power preference and the primacy condition, even if such a condition should prove not to be necessary.
Here, then, is a sufficient condition. Let us begin with freedom of preference. Suppose I prefer to do something, to write this paper, for example. We are seeking a sufficient condition for this freedom without worrying about whether it is necessary. If we add some bells and whistles that are not really needed for freedom of preference, that will not matter. So let us help ourselves to a set of conditions that are clearly sufficient, that clearly insure freedom of preference. Here are the conditions I propose for our consideration. My preference for doing X is free if
1. I have a power preference for X, that is, I would prefer to have the preference structure for X that I have.
2. The power preference for X satisfies the primacy condition, that is, I would have the preference structure for X because I would prefer to have it.
These conditions, though they insure that my preference is not manipulated by another might not provide adequate insurance that I could have preferred otherwise. So let us add a third condition to insure that.
3. If I had preferred to have a preference structure with the preference to do otherwise, I would have had that preference structure with the preference structure to do otherwise because I would have preferred to have it.
Conditions 1 and 2 insure that I have the preference structure that I do because I prefer to have it, while condition 3 insures that if I had had the preference structure to do otherwise, I would have had that preference structure because I preferred to have it instead. I claim that this is sufficient for freedom of preference. I prefer to do X because I prefer to have the preference structure for X including that preference for doing X, and if I had preferred to have a preference structure with a preference to do otherwise, I would have had that preference because I preferred to have it. It is up to me what preference structure I have, both actually and counterfactually, and that suffices for my freedom of preference. Whatever preference structure I have or would have concerning doing X, I would have it because I prefer to have it.
It should be noted that the subjunctive formulation of conditions (1) and (2) in terms of “would” requires not only that the person has the preference to do X because he prefers to have that the preference structure containing this but also would have it. This treats the very influential Frankfurt counterexamples to the principle of alternative possibilities by requiring alternative possibilities for freedom of preference. In these counterexamples, though a person does what he prefers, namely, to do X, and prefers to prefer to do X and so forth, there is an inactive counterfactual intervener with knowledge of what the person is about to prefer and the power, which he would have exercised if necessary, to make the person prefer to do X if the person were about to not prefer to do X. Such a person as described in the Frankfurt examples who prefers to do X would not have the preference to do X because he preferred to have that preference when he was about to not prefer to do X and the intervener caused him to have the preference to do X. The person would have the preference because the counterfactual intervener would intervene.
So we have a sufficient condition for freedom of preference when whatever preference structure I have or would have concerning doing X is one I would have because I prefer to have it. Suppose that I do something because I freely prefer to do it and I would have done otherwise if I had preferred otherwise which I would also have preferred freely. That is a sufficient condition for freedom action. I could have done otherwise. So from the three conditions formulated above we obtain an account of freedom of preference and freedom of action.
Are the three conditions offered above compatible with determinism? They are. The reason is related to the remarks about the primacy of causation and explanation given above. You may cause me to prefer something without that destroying my freedom of preference if I prefer what you prefer because I prefer to have that preference structure and my preference for that preference structure is the primary cause or explanation. Similarly, some other reason or motive may cause me to prefer something without that destroying my freedom of preference. Suppose I prefer to respond to that reason or motive because I prefer to have a preference structure responding to such reasons or motives. Moreover, suppose my preference for that preference structure is the primary cause or explanation of it. That insures my freedom. Furthermore, the assumption of the primacy of preference in causation and explanation is compatible with causation and explanation.
These remarks introduce a further complication, and a necessary one in our theory of what is sufficient for freedom. The complication concerns the influence of reasons or motives. Any compatiblist account must include a theory of how reasons can influence our preferences, indeed, cause us to prefer one thing rather than another, without undermining our freedom of preference. Moreover, one advantage of a compatiblist account of freedom of preference is that it permits us to explain the function and efficacy of reasons and motives causally rather than leaving this unexplained. The problem is to explain the difference between causation that insures freedom and causation that creates bondage. The solution consists in part of a difference in how reasons or motives cause us to prefer what we do when we are free as opposed to when we are not.
There is a system of reasons or motives to which I am responsive in the formation of preference. What those reasons and motives are may not be up to me. Recall that what we desire is often not up to us, though we may freely prefer to satisfy the desire or freely prefer not to satisfy it. In other cases, desires may overwhelm us, and we may seek the satisfaction of them obsessively or compulsively without doing so freely. What is the difference? The answer should be clear from the preceding remarks. The difference is whether we respond to the reasons we have because we prefer to do so. We are here going beyond the power preference, a preference for preferences for X, to a more complicated structure of preference for responding with preference to reasons. There is a special preference, which I shall call an ultrapreference, for the more complicated structure of responding to reasons with preference. Again, however, I may have the ultrapreference I do have for X, including a preference for responding to reasons, because I prefer to have it.
Here, then, is the theory I am offering as sufficient to enable us to distinguish between causation by reasons that insures our freedom of preference and causation that does not. Suppose I have a system of reasons SR for X. This system may include reasons for doing X and replies to objections to that course of action. Now suppose I have a system of reasons SR and respond to that system by forming a preference for X. Notice, first of all, that I may prefer X under the satisfaction of the three conditions of freedom and, therefore, freely prefer X even though I form that preference in response to SR. That is so when I respond to the reasons because I prefer to respond to SR in that way, and I would respond otherwise if I preferred otherwise because I preferred to respond otherwise. I have an ultrapreference to respond to certain reasons for X by preferring X, and I have that ultrapreference because I prefer to have it. The primacy of causation and explanation obtained from my having the power preference and ultrapreference because I prefer to have them is what insures my freedom within the nexus of causation.
We can now say that I freely prefer X because I prefer to have the preference structure that I have for X and because I have the reasons that I do for preferring X. The explanation of how both can be true is given by the ultrapreference that I have for responding to certain reasons which I have because I prefer to have it. We do not need to appeal to unexplained or mysterious factors to explain this kind of freedom. We do not need to suppose that the influence of motives is something other than causal. It is causal. It is just that the causal efficacy of the reasons depends on my preference for responding in that way. That is what permits me to be rationally free in my preference at the same time that my preference is caused and, indeed, caused by the reasons that are part of the system of reasons. I respond to system of reasons because I prefer to do so.
At this point, a question naturally arises concerning the relationship between the ultrapreference and the responsiveness to the system of reasons. One might ask whether I have my ultrapreference because I respond to the system of reasons I have or whether I respond to the system of reasons I have because of my ultrapreference? The answer is that both are true. I have my ultrapreference because I respond to the system of reasons, and I respond to the system of reasons because of my ultrapreference. The relationship of the system of reasons to the ultrapreference is one of mutual causal support and dependency.
An analogy I have used before to explain the causal dependency is that of a keystone in an arch. The arch requires the keystone to stand, for otherwise it would collapse, but the other stones, constituting the system of reasons, are also essential to the arch. Moreover, the keystone, which is the ultrapreference, contributes to its own support by supporting the sides of the arch which support it. It is, in a way, self-supporting but only in relationship to the other stones. Finally, if there is a system of arches, corresponding to structures of reasoning for preferences, they may constitute a dome with a common circular keystone at the top. In that case, the circular character of the keystone is part of what makes it effective in supporting the dome of rational preference and the system of reasons, the arches, which, in turn, support it. All of this is only an analogy and a metaphor. The point of it is to illustrate that the relation of mutual support of the system of reasons and the ultrapreference can be interpreted causally without paradox. The notion of mutual causal support is a commonplace one.
Nevertheless, the feeling might remain that something must come first. Either we first have a system of reasons which causes us to form the ultrapreference or we first form the ultrapreference which creates the system of reasons. Sartre, who speaks of choice rather than preference, thinks there is a kind of ultrachoice that we must make without justification and without excuse and that choice is both free and fills us with dread as we realize there is nothing to support it. Though I am indebted to Sartre for my conception of an ultrapreference, I do not agree with him that it is without justification and without excuse. Preference, unlike mere desire, involves evaluation in terms of reasons we have for satisfying or not satisfying what we desire or consider. Thus, our preferences are formed with at least personal justification, however fallible or fallacious the justification might be.
On the other hand, a hard determinist affirming the incompatibility of freedom of preference and causation might argue that what reasons we have and what preferences we form based on those, including the ultrapreference itself, are caused by our having those reasons, or in some other way. So the reasons, the justifiers and excusers, cause our preferences, including our ultrapreference, and there is no room for freedom in causal sequences that lead to the formation of the reasons and from the formation of reasons to the preferences. Thus, the traditional dilemma. Either the reasons to which we respond cause the preferences including the ultrapreferences, and, therefore, are not free, or we respond to the reasons we do because of an uncaused free preference to respond to such reasons, an uncaused version of the ultrapreference, and our preferences are not caused. We can, as I have suggested, exit from this dilemma.
The exit is supported by the keystone ultrapreference. We respond to the system of reasons we have because we prefer to respond in that way, that is, because of the ultrapreference for responding in that way. If the ultrapreference contains a power preference satisfying the three conditions given above, the ultrapreference is free. I may have it because I prefer to have it and, if I preferred otherwise, I would prefer otherwise because I would prefer to have that preference. Does that entail that my ultrapreference is itself uncaused and also without justification? Does inserting causation and justification destroy the freedom of the ultrapreference? Not at all! The ultrapreference, like other preferences, is justified by my evaluation in terms of my reasons to which I respond with preference.
But which comes first, the system of reasons supporting the ultrapreference or the ultrapreference for preferring the system of reasons? Neither. The system of reasons supporting the ultrapreference and the ultrapreference for the system of reasons stand or fall together composing an arch of reason capped with the keystone ultrapreference. The causal structure of the arch explains the possibility of compatiblism. The ultrapreference is a free preference causally supported by the reasons that justify it.
There remains a question about the temporal order of the causal process, about the sequence of causality, and what occurs within it. Any answer to such a question is highly speculative. My proposal is that when we choose and decide we often simultaneously choose to make the reasons that support the choice effective as we make the choice. Of course, we may reflect on the reasons for and against performing an action. But when we choose, that choice may constitute, at the same time, a choice to respond to the reasons for it. The idea that we first choose to make some reasons the effective ones in choice and then, after that, respond to the reasons by making the choice, though it may sometimes occur, does not seem the typical structure in choice. In effect, choosing a course of action, to write this paper, for example, and choosing to make the reasons in favor of the choice effective, that I committed myself to do so to a friend and former student, for example, seem to me to occur together and to depend on each other. I choose at the same time to make the reasons effective and to write the paper.
Since preference to do something is a disposition to immediately choose to do it when the situation is propitious, the preference to write the paper and the preference to make the reasons for doing that effective are made at the same time. This may be obscured by earlier intentions and preferences which influence me. It remains up to me, however, to prefer to fulfill those intentions and to sustain the preferences. So the preference to act and the preference to make the reasons for the preference effective are simultaneous. The two preferences may, in fact, be included in one, the one may be the same, in a sense, as the other, but that claim takes me beyond what is required for my purposes. I leave it before us as a conjecture for evaluation.
There remains a synchronic question. How does this structure of a system of reasons and an ultrapreference come into existence in the first place? The simple answer is that it is caused. I am not sure how. However, I think that the rough picture is that we at a certain stage have no preferences. Only impulses and desires, perhaps informed by unreflective beliefs and impressions, which drive our behavior. We form patterns of action, patterns of responding to desires and beliefs, at the level of the first order mind unencumbered by the burden of ratiocination and higher order evaluation. Then we acquire the higher order mind, the metamind, that allows us to represent and think about our desires and beliefs, to prefer the satisfaction of some desires and not others, to accept some beliefs but not others, as we sort out the first order states. At this point, at the point of the intervention of the metamind, we go through a brief, somewhat confused period, of vacillation between mind and metamind, and, finally, we arrive at a state of preference and acceptance, of systems of reasons supporting preferences and preferences supporting reasons. That is my conjecture concerning the transition from mental causation to metamental causation and the freedom of preference. It is not more than a conjecture.
The argument from what I have said to the conclusion of compatiblism is brief. Combine the account of freedom of preference in terms of the three conditions concerning the power preference stated earlier with the account of the structure of a system of reasons and the ultrapreference containing a power preference satisfying those conditions, and you have sufficient conditions for freedom of preference. A sufficient condition of freedom of action is easily obtain therefrom. Suppose I do something because I prefer to do it when my preference is free. Suppose in addition, I would have done otherwise if I had preferred to and that preference would have been free. On those suppositions, my action was free. Suppose my action and preference satisfy these sufficient conditions for freedom. Is there anything in these conditions that entails that these structures of preference which suffice for my freedom are uncaused? Nothing at all. Is there anything in this account which entails that something must be undetermined, that there must be some point at which there is no causal explanation for why I prefer one thing and not another? There need be no such point.
If I were completely rational in my preferences, which I admit without regret I am not, all my preferences might be responses to reasons which causally support them. There might be causation all the way around, before and after my preferences. But some of them would be free. When I prefer what I do because I prefer to have the preference structure I do, including my ultrapreference, I have all the freedom of preference that interests me. Why should I wish, in addition, that the causal order should be undetermined? It may be undetermined, as many aver, but I am indifferent in consideration of my freedom, though curious for theoretical reasons, about whether they are right.
Having the positive argument before us, let us consider the advantages and objections to it. First let us consider the advantages, and then we shall turn to the sticking points. The account accommodates many of the claims of libertarians without the assumption of indeterminism. Some, Reid, Chisholm and Taylor, for example, have said that freedom requires agent causality. I must be the author or agent of my preferences if they are free. But that claim leaves open the question of what is involved in an agent causing her own preferences. My claim is that an agent is the cause of her preferences when she has those preferences because she prefers to have them.
Now, one might add, if one wishes, that the agent is a substance, and the substance causes her preference. My suggestion is that this causation amounts to nothing more than her having her preferences because she prefers to have them. What does she do to cause the preferences other than prefer to have them? Perhaps nothing. Perhaps she simply responds to the reasons to which she prefers to respond. Moreover, if she has the preferences because she prefers to have them, then she, and not another, is the agent or author of them. If another were the agent or author of them, then she would not have them because she prefers to have them but because the other prefers that she does.
Many have argued that agency is to be explained in terms of rationality, like Kant, or identifying with the desire to act in terms of reasons, like Velleman. Though acting from reasons may contribute to my sense of agency, it also may not, and so this identification of agency with rationality and the desire to respond to reasons seems to me to reflect a rationalistic bias. I may feel alienated from the life of reason when reasons lead me to a course of action. Personally, I find this familiar enough. The life of reason sometimes leaves me feeling that reasons are ruling my life, and I am bereft of any sense of agency in responding to them. I go where the reasons lead me feeling alienated from the process and with a diminished sense of agency. The problem in such cases is that my preference for acting according to the reasons I do is not a preference that I have because I prefer to have it. I have a sense of agency when I prefer what I do because I prefer to have that preference and when the effective reasons for the preference, if any, are effective because I prefer that they be effective. I am in this view close to the view that Frankfurt advances when he emphasized the importance of identifying with a preference for a sense of agency, but I suggest that my identification with the preference is the result of having that preference because I prefer to have it and responding to my reasons for having the preference because I prefer to respond to them.
There is also the issue of control and power. If I am free in my preferences, then they must be in my control, in my power. Suppose, then, that my preferences and the system of reasons that supports them are ones I have because I prefer to have them, and, moreover, are such that if I preferred not to have them, I would have that preference because I preferred to have it. Are my preferences in my control? I control them all right. I have them because I prefer to have them. If I preferred not to have them, I would have that preference because I preferred to have it. What I prefer is in my control in the sense that my preferences, whatever they may be, are ones that I have because I prefer to have them. I am in control. Do I have power over my preferences? They are in my power. I have the preferences I do because I prefer to have them. To avoid tedium, I proceed no further in this line of argument.
Is the theory a theory of volitions? Preferences are like volitions. One problem with trying to analyze freedom in terms of volitions is that the theory appears to lead to an infinite regress, when we try to analyzed the freedom of volitions themselves, or, instead, a kind of surd when a philosopher claims that volitions make us free without need to explain what makes the volitions free. But the power preference avoids the regress and the surd. What makes the preference free is that we have the preference structure we do because we prefer to have it. What makes that preference, the power preference, free, is that we have it because we prefer to have it as a preference included in the preference structure. The power preference, since it is a preference we have because we prefer to have it, ties freedom in, down and together in an explanatory loop.
Finally, the theory avoids a kind of paradox of libertarianism. The paradox is that the libertarian insists on the importance of indeterminism and then faces the objection that the mere lack of causal determination does not seem to contribute in any positive way to the account of freedom. Ekstrom, who provides us with a profound and articulate theory of libertarianism, tries to find a useful place for indeterminism at the point of forming a preference in consideration of reasons in the process of deliberation. That is a plausible point for a libertarian to insert indeterminism. But what purpose does it serve? It is supposed to bolster the idea that it is up to me at this point what I prefer. But it is up to me only if I have some positive power to resolve the matter. The mere lack of determinism does not explain the positive power. The positive power, I propose, is captured by the primacy of the power preference, by my having the preference I do at that point, and responding as I do to the reasons I have, because I prefer to have those preferences and to respond in that way. In short, I propose that where the libertarian thinks she gains some advantage by introducing indeterminism into the picture at some point, what is needed is some account of positive empowerment. That is what the power preference gives us without appeal to indeterminism provided we have it because we prefer to have it.
The advantages before us, let us look at the problems that confront compatiblism and explain how the sufficient account for freedom solves them. First of all, incompatiblists will ask how it can be up to us or in our power what we prefer or do, if our actions and preferences are caused in a way that extends backward in time to conditions that are beyond our control.. As Spinoza noted, if a projectile were conscious, it might reflect, “I could swerve to the left or to the right, but, I prefer to continue on this agreeable parabolic path.” Don’t we seem like the projectile if determinism is true?
No doubt, some will find that image so compelling that there is no convincing them otherwise. But notice that we do not believe that the projectile would swerve off the path if it so preferred. So, there is a difference at the level of action often noted by compatiblists from Moore on. Moreover, even if the projectile were conscious and felt inclined to continue on its parabolic path, it would not have a preference to move on that path because it preferred to have that preference any more than it would continue on that path because it preferred to do so. The causal role of preference in action and in the empowerment of preference is missing. That is why the example is both convincing and misleading.
Let us, however, look at the harder arguments for incompatiblism. Both Ginet and Van Inwagen have argued that if we have no choice about the past, which we do not, and if we have no choice about the laws of nature, which are part of their definition of laws, then we have no choice about the outcome of the past conditions and the laws of nature. Or to put another way, there is nothing that we can do to render true claims about the past false or render true claims about the laws false, and, yet, if we can do otherwise than we do, then either the true claims about the past would be false or true claims about the laws would be false. Therefore, if we can do otherwise, we can render either the claims about the past false or the claims about the laws false. But we cannot do these things. So, if determinism is true, then we cannot do otherwise than what we do. Such arguments are very persuasive. How should we reply?
My first reply is that the word “can” is highly ambiguous in a way that makes it very difficult to argue in a cogent manner about the issue of compatiblism formulated in terms of “can”. As Moore noted, there is clearly a sense of “can” that makes the argument offered sound. I have no doubt that there is a version of the argument that Van Inwagen offers using “can” as he intends so that he can meet the objections of his critics and argue that if determinism is true, then no one could have done otherwise – in his use of “can”. Similarly, there are other uses, meanings, of “can” some of which I have proposed, from which compatiblism follows. Conditional analyses and variations thereof in terms of accessibility relations to possible worlds analyzing accessibility in ways that favor compatiblism are illustrations. Formulating the matter in terms of “can” is, I ruefully conclude, hazardous and the question is easily begged.
Nevertheless, I do think that the conditions that I have offered are sufficient conditions, not only for freedom, but for saying that a person could have done otherwise. So how do we avoid the argument that if determinism is true, then no could ever have done or preferred to do anything other than what they did or preferred to do? There is a principle that figures into the argument in one form or another which says that if a person can do something, and something is a necessary condition for his doing it, like the falsity of either the past or the laws of nature, then if the person can do the thing, then the person can bring about the necessary conditions for doing it as well. This is plausible. It says that if something is necessary to doing something, then if you can do the thing in question, you can bring about the necessary condition. Plausible, but false. One necessary condition of my writing this paper is the birth of my parents, but there is nothing I can do to bring about the birth of my parents. So the general principle that says that if some condition is necessary for a person doing something, then if the person can do the thing, the person can bring about the condition.
Now any incompatiblist worth his or her arguments will jump up and down and reply that it is only conditions that do not exist and which are necessary for doing something that a person must be able to bring about if they can perform the action. Again, however, this is false. If the conditions simply will occur if the person performs the action, that is all that is required.
Now when we turn to what could have been, what a person could have done, we must say that any condition that is necessary for the person to have done the action is one that would have occurred. So, as Lewis noted, when we say that a person could have done something that she did not do, then, if determinism is true, we must conclude that if the person had done what she did not do, then the necessary condition of her doing it, namely, the falsification of the statements about the past or statements of laws would have occurred. But we cannot conclude that the person doing otherwise would have brought about occurrence of that condition. That would be a mistake. All we can conclude is that one of those statements would have been false.
Now let us look at the power preference. Suppose I had preferred otherwise. Moreover, suppose that I had preferred otherwise because I preferred to have a preference structure containing that preference. Then a necessary condition for my preferring otherwise is that I not have preferred what I actually did and that all necessary conditions for that were satisfied. That includes that either the laws concerning my preference or the past would have been otherwise. I see no problem with this. It does not follow from the fact that I could have preferred otherwise that I could have thereby brought about the necessary conditions for so preferring, and, more specifically, it does not follow that I could have brought it about that the past or laws would have been otherwise. All that follows is that they would have been otherwise if I had preferred otherwise.
You might inquire further what would have been otherwise – the past or the laws? Suppose it is the laws of nature that would have been otherwise. What that means, of course, is not that something would have been both a law and false but, rather, that something that was law would not have been law. Is that strange? Some, Van Inwagen and Ginet , have suggested that whether something is a law of nature does not depend on our choices or preferences, though Van Inwagen has noted the possibility of voluntaristic laws. Most laws do not concern human choices and preferences. Such laws, the laws of chemistry, for example, do not depend on our choices and preferences. However, they are not laws concerning our choices and preferences, laws of the form, Lc, to the effect that under conditions C a person S will choose or prefer X, for example. Now consider laws like Lc, supposing there are some. What is the point of saying that they do not depend on human choices and preferences? They concern our choices and preferences, after all, and say that under conditions C people choose or prefer X. Such laws of the form of Lc clearly do depend on our choices and preferences. If we do not choose or prefer X under conditions C, then Lc would not be a law. So, there is nothing strange in the idea that if I had chosen or preferred other than X under C, then Lc would not have been a law, though, in fact, it is a law.
Now we come to the critical objection, however. If Lc is a law, shouldn’t we conclude from this that no one could have not chosen or preferred X when they were in conditions C? No doubt that there is a sense of “could have” that would warrant this conclusion, but, as I have noted before, we need the additional premise that when a person could not have chosen otherwise in this sense, then the person could not have chosen otherwise in the sense required for the person to be free to have chosen otherwise. And it is precisely this step I wish to question. Notice, that a law statement must be true, and there is a sense of “could have” in which it seems obvious that no one can make what is true false and, thus, that no could have made what is a true law statement false. But that sense of “could have” does not deprive us of freedom because it rests on the assumption that no can make something be both true and false, which is a tautology. That does not preclude us from saying what is required for freedom, namely, that somebody could have done something which would have made something false instead of being true, though, in fact, she did not, and it is true that she did not.
Philosophers have thought, however, that there is a more important modal notion of “can” connected with laws, a more significant sense of “can” and “could have”, that enables us to distinguish laws from mere accidental general truths. It is, for example, an accidental general truth that everyone in this room is older than four years old. The accidental character of the general truth might be captured by a modality, by saying that someone could have falsified this general truth by bringing her four year old to hear about freedom and the power of preference, but no one could have done anything to falsify any law of nature. I have two replies with which I shall conclude. First of all, it may be true of some but not all laws of nature, namely those concerned with matters other than human choice and preference, that they cannot be falsified by human choice or preference, but it does not follow from this that no law about human choice or preference can be falsified by human choice or preference.
Secondly, and more crucially, there is another and, I believe, better way of distinguishing between laws and accidentally true generalizations than by appeal to what can and could not be. Laws warrant counterfactual inference of a kind that is not warranted by accidental generalizations. The specification of the exact kind of counterfactual inference is a delicate matter requiring some subtle distinctions, but the general point is clear enough. The accidentally true generalization that all people in this room are older than four years does not warrant the counterfactual inference to the conclusion that if a four year old had been brought in to hear about freedom and the power preference she would been older than four years old. Laws by contrast do warrant counterfactual conclusions about what would have been. So, if Lc is a law and I, in fact, do not choose or prefer X, someone would be warranted in inferring that if I had been in conditions C, which I was not, then I would have chosen or preferred X. The person would not, however, be warranted in inferring that I could not have chosen or preferred otherwise than X when I am in C. They could only infer that I would not.
I conclude with the remark that this conception of a law is one that differs from what incompatiblists such as Ginet and Van Inwagen have assumed about laws in their arguments for incompatiblism. Am I, therefore, just missing the point of their argument? One motivation for concern about the problem of freedom and determinism is that it appears that a scientific conception of the world based on laws of nature and explanation is incompatible with the sort of freedom we believe ourselves to have. I have intended to allay this concern. The scientific conception of the world is, I suggest, one including statements of laws that warrant inference and counterfactual inference about the world. Some philosophers may go beyond this conception to one affirming a thesis of laws that contain a kind of necessity that is incompatible with human freedom. If one builds incompatibility with freedom into the definition of laws by defining the latter in such terms, then incompatibility will be the result. It would be a mistake, however, to conclude that the scientific conception of the world based on laws that warrant inference and counterfactual inference are incompatible with our conception of human freedom. One important problem is to explain how human freedom can fit into the causal nexus of inference and counterfactual inference warranted by laws. My explanation is that it is contained in the keystone loop of preference and ultrapreference which ties together our conception of ourselves as free with our conception of a causal order explaining our freedom. Causally explained freedom accounts for our place in nature.
1. Keith Lehrer, Self-Trust: A Study of Reason, Knowledge and Autonomy (Oxford, 1997), 11-12.
2. The classic statement of the conditional analysis was G. E. Moore, Ethics (London, 1912, Chap. 6.
3. Keith Lehrer, “Preferences, Conditionals and Freedom,” in Keith Lehrer, Metamind (Oxford, 1990), Chap. 3, reprinted from Peter Van Inwagen (ed.) Time and Cause (Dordrecht, 1980).
3.Harry Frankfurt, “Freedom of the Will and Concept of a Person,” Journal of Philosophy, 68 (1971), 5-20.
4. Peter van Inwagen and Krister Segerberg both communicated the objection to me in correspondence. For the published version, see Peter van Inwagen, An Essay on Free Will (Oxford, 1983), Chap. 4, amd Krister Segerberg, “Could have but did not,” Pacific Philosophical Quarterley, 64 (1893) 230-41.
5. Saul Kripke, “Outline of a Theory of Truth,” Journal of Philosophy 72 (1975), 53-81.
6. Moore, Ethics, Chap. 6.
7. Harry Frankfurt, “Alternate Possibilities and Moral Responsibility,” Journal of Philosophy, 68 (1969), 388-404.
8. I was not able to pin down this interpretation to a specific passage, for his prose has a charmingly diffuse character, but I am indebted to Jean-Paul Sartre, Being and Nothingness, Hazel Barnes (trans.) (New York, 1956).
9. Thomas Reid, Essays on the Active Powers of Man (Edinburgh, 1785), Essay 4.
10. Roderick M. Chisholm, “Freedom and Action,” in Keith Lehrer (ed.), Freedom and Determinism, (New York, 1966), 11-44.
11. Richard Taylor, Action and Purpose, (Englewood Cliffs, 1966).
12. Immanuel Kant, Critique of Practical Reason (1788).
13. J. David Velleman, “What Happens When Someone Acts,” Mind, 101 (1992), 461-481.
14. Harry Frankfurt, “Freedom of the Will and Concept of a Person,” op. cit.
15. Laura Waddell Ekstrom, Free Will: A Philosophical Study, (Boulder, 2000), Chap. 4.
16. Baruch Spinoza, Ethics, R. H. M. Elwes (trans), (London, 1919), Props xlvii-xlix.
17. Carl Ginet, “Might We Have No Choice?,” in Keith Lehrer (ed.), Freedom and Determinism, (New York, 1966), 87-104.
18. Originally argued in Peter van Inwagen, “The Incompatibility of Free Will and Determinism,” Philosophical Studies, 27 (1975), 185-199, and more fully in An Essay on Free Will (Oxford, 1983).
19. G. E. Moore, Ethics, op. cit.
20. David Lewis, “Are We Free to Break the Laws?,” Theoria 47 (1981), 113-121.
21. Carl Ginet, “Might we Have No Choice?,” op. cit., and Peter van Inwagen, “The Incompatibility of Free Will and Determinism,” op. cit.
22. Peter van Inwagen, An Essay on Free Will.
HOME to Det & Free front page
HOME to T.H. front page