This text is a guest essay by Christian Onof, in which he provides a précis of his forthcoming book The Problem of Free Will and Naturalism. Paradoxes and Kantian Solutions, soon to be published by Bloomsbury Academic. The essay is subdivided in two parts. Part I can be read here below. Part II will follow shortly.
PART I
Methodology and Structure of the Book
The title ‘The Problem of Free Will and Naturalism’ serves a dual purpose. First, the reference to the problem of free will announces that this is a book about the perennial philosophical conundrum of the compatibility of free will and the causal determinism of nature. Second, by choosing to include ‘naturalism’ in the title, I sought to underscore that it is the naturalistic framework in which the problem is typically set that is in fact the source of the problem. I devote the first three chapters to making this point.
Furthermore, I claim that the paradox which arises from trying to make sense of free will in a deterministic naturalist framework has the same structure as a paradox characterising the metaphysics of naturalism (Chapter Seven). These paradoxes, I argue in Chapters Four to Seven, can be dealt with only by endorsing the paradigm shift underpinning Kant’s critical philosophy; hence the subtitle of the book.
This shift, which involves the adoption of the metaphysics of transcendental idealism, may seem too high a price to pay for many contemporary philosophers. But, if the only alternative available is a position, hard incompatiblism, which requires a substantial revision of our self-conception, this must be a price worth paying, I think.
In Chapters One and Three, the book features a critical review of existing contemporary proposals to solve the problem of free will and determinism (hereafter ‘the Problem’). I address the problems encountered by both main approaches to the Problem, i.e. compatibilist and libertarian theories. This examination issues in the outline of a possible solution (Chapter Three). Kant’s own solution is first introduced, together with the relevant Kantian background, in Chapter Two; the main text where Kant addresses the Problem, namely the Resolution of the Third Antinomy in his Critique of Pure Reason (hereafter CPR) is then shown, in Chapter Four, to define a solution of the type described in outline in Chapter Three. This, together with a presentation of further features of Kant’s theory of freedom in Chapters Five and Six, constitutes the exegetical portion of the work.
Both in the review of contemporary theories and Kant’s own proposed solution, three types of account are presented as required for a complete solution to the Problem: I argue that certain issues have arisen in the contemporary debate and in Kantian exegesis as a result of not properly distinguishing between these types. To understand these distinctions, consider that the Problem is primarily that of an apparent incompatibility between the sense that I am in control of my action and the objective fact that it is completely determined by natural causality. A volitional account of free will is a proposal to solve this first-person issue of control.
A related third-person problem can also be identified insofar as determinism undermines moral responsibility. Here, I assume, for the sake of simplicity, that the determinism in question is psychological. Therefore, what is needed at the third-person level, is what I call a psychological account which is tasked with explaining how the agent can be responsible for causally determined actions. The compatibility of volitional and psychological accounts must also be ensured: this enables a satisfactory explanation of second-person attitudes of praise and blame to agents who are morally responsible and whom I treat as sources of control.
Note, further, that this control has two characteristics that typify what I consider to be fairly widespread libertarian intuitions: I am the source of the action and I could have done otherwise had I so decided. These define respectively sourcehood and leeway requirements, a distinction that is useful to understand the evolution of the contemporary debate.
The need for a third type of account of free will arises from the metaphysical issue of understanding the nature of the controlling agent in relation to the causality of nature. This defines a metaphysical account which provides a metaphysical grounding for the volitional and psychological accounts with which it must of course therefore be compatible. I now turn to the main articulations of the argument that I develop throughout the book’s seven chapters.
An Examination of Compatibilism (Chapter One)
Van Inwagen’s (1983) Consequence Argument sets out the problem of free will as a first-person one that is expressed in the form of a syllogism. The assumptions are: (i) that the past and the laws of nature jointly determine our acts, and (ii) that neither events predating our birth nor these laws are up to us. It follows that our acts, as determinate consequences of these laws and events, are also not up to us.
There are two main types of response to this argument:
(1) Compatibilists argue that a proper understanding of free will would invalidate the syllogism;
(2) Incompatibilists accept the validity of the argument. They are either hard incompatibilists who deny the possibility of free will, or libertarians who deny the premise of determinism.
Peter Strawson’s compatibilism involves demarking free will talk as a feature of a ‘participant’ as opposed to an ‘objective’ standpoint. Any such boundary however seems much more permeable than Strawson would have it as it evolves over time with the progress of the neurosciences and objective assessments of an agent’s mental capacities (or disabilities) impact his perception of his own free agency.
A long tradition of compatibilism from Hobbes to Ayer opted rather for a demarcation of the concept of free will in terms of a conditional analysis: I could have done otherwise had I willed otherwise, which provides me with a conditional type of leeway. But the case of phobias requires placing restrictions upon what can count as the agent’s willings, and exactly where they must be placed is contentious: the incompatibilist contends that all cases of causal determinism are to be excluded.
Derk Pereboom’s Argument from Manipulation lends credence to such incompatibilist claims. By presenting three cases of decreasing levels of external manipulation of an agent’s brain and comparing them with the actual case of physical determinism, he makes a strong case for there being a continuity between these cases so that free will should be ascribed to none of them.
The problems of traditional compatibilism motivated contemporary compatibilists to focus rather upon the other key feature of control, sourcehood. This should be defined so that cases of brain manipulation can be distinguished from the actual case in the Argument from Manipulation. There are essentially two possible approaches here. ‘Real-self’ or ‘mesh’ proposals are theories of how the self as source of agency is to be identified. This will involve a mesh connecting the action to higher-order desires, as proposed by Frankfurt, to Bratman’s policies of practical reasoning or to Watson’s alternative of a valuation system. The ‘rationality’ or ‘reasons-responsive’ approach exemplified by Fischer and Ravizza’s theory of moral responsibility opts rather for characterising sourcehood in terms of the responsiveness to reasons of the mechanism leading to the action.
Pereboom’s Argument from Manipulation can however be shown to apply to such source compatibilist conceptions of free agency.
I argue that this results from a focus of source compatibilism upon the third-person problem which leaves us without a plausible volitional account. The source compatibilist would argue that it can be constructed from the psychological account defined by a mesh and/or a reasons-responsive mechanism. But this does not explain why I grasp reasons for action as mine: I call this absence of a satisfactory place for me as agent the problem of the Homeless Agent.
Kant’s own pre-critical conception of freedom is a type of compatibilism following in the wake of Leibniz’s and Wolff’s compatibilist concept of freedom, and therefore, it similarly fails to deliver a satisfactory volitional account. The critical Kant grasped that, only by placing the agent’s volitions outside the domain of natural causality, and therefore outside time, could this problem be addressed. This no doubt was a major contributing factor in his adoption of transcendental idealism. Kant’s critical theory of free will thus introduces the need for a metaphysical account aside from the volitional and psychological accounts of free will. Among its tasks, it has to define the locus of freedom. Chapter Two presents an outline of how Kant approaches this question.
An Examination of Libertarianism (Chapter Three)
The requirement of locating the agent’s free will outside the domain of natural causality will seem unappealing to many contemporary broadly naturalist philosophers. Contemporary libertarianism offers alternative approaches which keep the agent’s willings within this temporal domain. The libertarian addresses the first-person problem by assigning a key function to the leeway condition: leeway is understood as either simply required for leeway control or also for sourcehood. The libertarian therefore typically rejects the validity of Frankfurt-style examples that claim to show that we can dispense with the leeway requirement. I concur with this stance on the grounds that these examples work only because of a problematic isolation of individual decisions leading to a given action.
On the assumption of a naturalistic metaphysics, indeterministic causality is required to meet the leeway requirement. While I have strong reservations vis-à-vis the idea that this should be the default conception of causality, I temporarily shelve my concerns for the sake of an examination of libertarian proposals in Chapter Three.
I review the three main classes of libertarian approaches: event causal theories, agent causal proposals and non-causal libertarian accounts. The first address the Homeless Agent problem by loosening the deterministic grip of a psychological account that is broadly a Davidsonian belief-desire type of model of action. Although this creates a gap in what was a deterministic causal flux, it would seem that what fills this gap is nothing more than mere randomness according to the proponents of the Luck Argument (e.g. Haji and Mele).
This is certainly true of what Clarke calls the probability pool model of indeterministic causality, a model assumed in the representation of uncertainty in quantum physics, in which the determination of a decision on the basis of the available reasons seems random. But Clarke makes a good case for an alternative model, the probability fork model in which there is no gap between decisions and their reasons. Rather, which of the reasons will actually be causally effective in bringing about a decision is indeterminate. While this counters the Luck Argument, the choice made is not one that the agent can influence (stronger Luck Argument). It is also not one for which any reason can be provided (Rational Luck Argument). This means that the volitional account is not compatible with a convincing psychological account.
These shortcomings of event causal attempts to solve the problem of the Homeless Agent suggest the obvious move of introducing the agent’s will as a type of causality into the account. Clarke thus addresses the deficiencies event-causal libertarianism by ‘adding’ agent causality to the event-causal story. The problem then lies in how agent causality and the causality of reasons might be related. In particular, Clarke’s 2003 proposal attempts to avoid the Rational Luck objection by arguing that agent causality have the function of being required for the causal effectiveness of reasons. Since there is no account of why certain reasons are selected as those acted upon, the Rational Luck Argument still has some purchase here. But this relation of dependence between agent-causality and event-causality will constitute a key feature of my solution to the Problem.
Another key feature is the role of spontaneity which McCann introduces in his non-causal proposal. The problem this account faces, however, is that of making sense of spontaneity in a naturalistic context. What remains is then the possibility of an exclusively agent-causal approach, such as O’Connor’s elaborate proposal. On his account, action is partly triggered by the agent’s self-determining causal activity. This activity features an event component that brings about an intention to act aimed at satisfying some desire. This introduces a teleological account of willings which replaces the function of event-causality. It is however not clear that the desires an agent acts upon must be those she aims to satisfy: this introduces a problem of compatibility of psychological and volitional accounts. The metaphysical account, in turn, suffers from not determining the temporal location of the intervention of agent-causality, which amounts to a deficit of rational control in terms of the timing of the action. While O’Connor’s theory addresses the issue raised by the Rational Luck Argument through the specification of further reasons for the reasons the agent chooses to act upon, there is still a lack of explanation in relation to why the agent chose these further reasons.
All libertarian theories suffer from such an explanatory lack according to Galen Strawson. Strawson argues that, insofar as I have rational control over my action, the reasons for it should be attributed to how I am mentally speaking. This leads to an infinite regress insofar as I must have chosen these reasons, for which further reasons are required. Free agency would therefore require an infinite number of choices, a clear impossibility.
I propose that to address this Infinite Regress Argument, we need to draw upon an important insight from Thomas Pink. This insight in effect distinguishes the issues raised by the Rational Luck and the original Luck Argument. The first identifies a problem of action intelligibility, while the second concerns the randomness implicit in the choice of the particular action. Pink points out that these problems have historically (since Hobbes) been conflated. It follows that the justification of an action need not come from the reasons which cause it. Pink proposes that an action’s purpose could be adduced to justify it, which offers a way out of an infinite regress of reasons.
The Outline of a Solution
This rational self-sufficiency of an intention/action’s volitional account in terms of its purpose would address the justificatory dimension of Strawson’s argument if extended to the whole of anyone’s agency. All that is required is for the sum-total of anyone’s willings to define a purpose. I argue that this is the case if the latter is defined minimally as the coherence of my individual willings.
But now, Strawson would point out that, since reasons are causes, the infinite regress also applies to causes. So we still have to show how an infinite causal regress in the psychological account is to be avoided. To do so, I draw upon Kant’s metaphysics of transcendental idealism (TI). Indeed, this addresses the Homeless Agent problem by placing her outside the causal flux in time. It seems that this must be the only way for her action to escape from the infinite causal regress which, as the Antithesis of the third antinomy explains, characterises any event in time.
These proposals for addressing the explanatory and causal dimensions of Strawson’s argument must now be made compatible in a metaphysical account. That is, we must answer the question of how the totality of an agent’s actions can be both the effect of her reasons (psychological account) and brought about by willings directed at a purpose which is minimally their coherence (volitional account). Here, causal over-determinacy will be avoided as long as my willings are the source of the causal efficacy of these reasons, following Clarke’s proposal for reconciling agent and event-causality. In summary, there are three components to my response to the Infinite Regress Argument:
A. The removal of willings from the natural causal nexus governing through the adoption of the metaphysical framework of TI, as required to address the Homeless Agent Problem;
B. A metaphysical account drawing upon Clarke’s proposal that the causality of an agent’s reasons has its source in the agent’s willings;
C. Pink’s notion of the rational self-sufficiency of an action’s explanation in terms of its purpose in a volitional account which is applied to the whole of her agency.
Kant’s own critical solution to the Problem is examined in Chapters Four to Six to show to what extent it fills in this proposed outline of a solution.
Kant’s Metaphysical Account and the Resolution of the Third Antinomy
Kant’s own solution to the Problem is such a disputed topic that there is not even agreement as to whether it is to be classified as compatibilist or incompatibilist. As I show in Chapter Two, this is, in no small measure, a consequence of the overlooked implications of naturalism in the contemporary classification. The fundamental question which needs to be answered here is simply that of how Kant would respond to the Consequence Argument. I argue that there is no doubt he would reject it on the grounds that he would not accept its minor premise, namely that the past and the laws of nature are not ‘up to us’. As a result, he is to be classified as a compatibilist.
Further, I argue, contra Tobias Rosefeldt, that Kant proposes that some laws of nature could involve the agent’s contribution, i.e. that his solution is an altered-law one, which concurs with Eric Watkins’ and Benjamin Vilhauer’s interpretative stances. I show that, for Kant, the agent’s psychological causality is a natural one whose ground lies in the agent’s freedom. The first belongs to the empirical domain, which is that of appearances for TI; the second belongs to the merely intelligible domain of reality-in-itself (things in themselves). On my understanding of TI, empirical objects are the appearing aspect of some reality-in-itself: an agent thus belongs both to the domain of reality-in-itself and appearances.
I also show how such an altered-law option deals with the Historical Agency problem identified by Ralph Walker, according to which it would seem that the agent has to be the causal ground of the whole of the causal history leading to her present action: it is the agent’s psychological causality, not her past which is grounded in her freedom. I also address the controversial issue of whether Kant is merely showing the logical possibility of free will in a deterministic world or whether the modality of real possibility is at stake: I conclude that what is at stake is the compatibility of the real possibility of transcendental freedom and natural causal determinism, i.e. what Kant refers to as its permissibility.
What this initial interpretation of Kant’s solution leaves us with however is the problem of making sense of what it means to place our agency outside time (Temporal Agency problem). In particular, even if the agent’s psychological causality is grounded in this free agency, it has, qua natural causality, an unchanging law. This seems incompatible with our having leeway in time (Temporal Leeway problem).
In Chapter Four, I carry out a detailed analysis of Kant’s solution as he spells it out in the Resolution of the Third Antinomy which focuses primarily on the metaphysical account. I address the oft-overlooked issue of the Resolution’s focus upon the possibility of agential freedom while the antinomy is primarily a cosmological problem about the grounds of empirical causal series. While the thesis of this antinomial problem argues that a first cause must be found, the antithesis excludes the possibility of any such cause insofar as it would be uncaused. I argue that while the resolution starts in a very general fashion which primarily concerns this cosmological problem, it then focuses upon the Problem, in which the possible first cause in question is an agent’s transcendental freedom. This shift is understandable given the primacy of Kant’s ethical interests.
I show that the argument has a well-defined structure. Initially, Kant proves that a solution to this Third Antinomy is possible only within the framework of TI, confirming the above conclusion A from the analysis of contemporary proposals: it is necessary for the causality of freedom to be outside time. In the three sections of the resolution, Kant then shows, respectively:
• The compatibility of the real possibility of transcendental freedom together with the causal determinism of appearances in general.
• The compatibility of this real possibility and causal determinism within the same agent. This involves two tasks:
First, the compatibility of a really possible intelligible causality (transcendental freedom) and an actual empirical causality must be shown.
Second, it is necessary to specify the relation between these two causalities.
Note that for Kant, real possibility, while it is defined in terms of compatibility with the formal conditions of experience (space, time, the categories), can in practice only be attributed to that which is found to accord with the laws governing the objective domain of appearances, i.e. nature.
The examination of Kant’s Resolution of the Third Antinomy shows that Kant is a libertarian, a label which would clash with his compatibilism only if he were a naturalist. It also confirms that Kant’s solution is an altered-law one in which the real possibility of transcendental freedom involves distinguishing an agent’s intelligible character, which is the law of his causality of freedom, from his empirical character, which must be understood as the law governing his psychological causality, and which is the appearance or (transcendental) effect of the intelligible character. This means that Kant’s solution conforms to conclusion B from the analysis of contemporary proposals.
While the Resolution of the Third Antinomy is focused upon the metaphysical account, Kant does provide some elements for the psychological and volitional accounts. In terms of the first, Kant’s understanding of the empirical character is in terms of subjective principles of action, although I show that this can easily be translated into the terminology of a belief-desire model. Kant makes statements about the possibility of fully predicting our action conditional upon the knowledge of this character, which have caused puzzlement among some commentators. Clearly, such statements only make the need to address the Temporal Leeway problem all the more pressing.
As far as the volitional account is concerned, what many commentators have taken to be a straightforward identification of the agent’s causality of freedom with the causality of reason leads to a further problem of Moral Leeway rearing its head: with this identification, free agency is equated with moral agency because the law of the causality of reason is the moral law.
Kant also introduces what is seemingly another conception of freedom, namely practical freedom, which is defined in terms of independence of any causal necessitation. While in the resolution, he argues that practical freedom would be ‘abolished’ if we were not transcendentally free, in the Canon of Pure Reason he claims that we can be said to be practically free whatever ultimately turns out to be the case as to the reality of transcendental freedom. These statements are apparently contradictory and, while many interpreters have attempted to reconcile much if not all of the statements Kant makes about practical and transcendental freedom (e.g. Beck, Allison, Kohl), this has led many to reject the conception of practical freedom, supporting their claim by pointing out that the terminology ‘practical freedom’ is hardly used by Kant elsewhere in his published writings.
I argue that practical freedom must be understood as a practical concept picking out the same freedom as the theoretical concept of transcendental freedom: it does not therefore refer to a weaker conception of freedom, hence Kant’s claim about its dependence upon transcendental freedom: I understand the abolition of the latter to refer to a proof of its real impossibility. With that interpretation of what the Resolution claims about practical freedom, together with what the Resolution achieves, namely a proof that the real possibility of freedom is compatible with natural causal determinism, there is no conflict with the Canon’s (a later chapter in Kant’s CPR) claim that we are practically free.
to be continued
© Christian Onof, 2024.