With determinism rejected, Dupré considers the problem of free will, but could not get beyond the standard argument against free will until his most recent work examining the right role for determinism.

In his 1993 book The Disorder of Things, an attack on the "Unity of Science" movement, he proposed a refutation of reductionism and determinism.

The best way to introduce the position against which this book will argue is to consider the continuing influence of one particularly notorious founding metaphor of modern science, the idea that the universe should be considered as a gigantic machine. Traditionally the favored machine has been a clock... Anyone who thinks that such mechanical metaphors have faded in significance might reflect on the amount of scientific effort that has been devoted in recent years to investigation of the hypothesis that the human brain — and generally also the human mind — is really a kind of calculating machine, or computer.

The philosophical thesis most intimately connected with this mechanistic metaphor is determinism. The cosmic clock, we must assume given its provenance, has always told the correct time and always will. To achieve such precision its components must exhibit the same unvarying reliability as the whole. Once the clock is wound up and set in motion, its behavior and that of all its parts are determined for all eternity.

...the current vogue for chaos theory suggests that certain aspects of this complexity may be in principle beyond the reach of certain kinds of comprehension, notably prediction. Nevertheless, the metaphysical conceptions of order that originated in the picture of the mechanical universe seem to have been little threatened, if sometimes modified, by such developments. The case of chaos theory nicely illustrates this resilience. Prediction, though long conceived as a very central excellence of scientific understanding, is a goal that has tended to recede rather than approach as various scientific disciplines have increased their understanding of the complexity of the phenomena within their domains. Chaos theory appears to confront prediction not merely with an insuperable practical difficulty, but with a logically impassable obstacle. Yet paradoxically determinism, the metaphysical underlay of :he possibility of prediction, is strengthened rather than threatened by this development. For the central mathematical functions of chaos theory are quite deterministic...Even the increasing prevalence of probabilistic rather than deterministic hypotheses and methods from quantum mechanics up through the scientific hierarchy has not, I shall argue, led to the rejection of some of the most fundamental features of the deterministic world view.

(The Disorder of Things, pp.2-3)

To begin with, it will help to introduce what is perhaps both the clearest and the strongest version, which I shall call classical reductionism. Assume, first, a hierarchical classification of objects in which the objects at each level are complex structures of the objects com- prising the next-lower level. Paul Oppenheim and Hilary Putnam propose the following levels: elementary particles, atoms, molecules, living cells, multicellular organisms, and social groups. The investigation of each level is the task of a particular domain of science, which aims to discern the laws governing the behavior of the objects at that level. Reduction consists in deriving the laws at each higher (reduced) level from the laws governing the objects at the next-lower (reducing) level. Such reduction, in addition to knowledge of the laws at both the reducing and reduced levels, will also require so-called bridge principles (or bridge laws) identifying the kinds of objects at the reduced level with particular structures of the objects at the reducing level. Given the transitivity of such deductive derivation, the endpoint of this program will reveal the whole of science to have been derived from nothing but the laws of the lowest level and the bridge principles. The lowest level will be the physics of elementary particles. Thus, finally, truly basic science need concern itself only with the objects described by particle physics.

(The Disorder of Things, p.88)

The argument that must be addressed here turns on the requirement that explanations at different structural levels must at least be consistent with one another. Certain views about the nature of causality suggest that only some kind of reductive relation between higher and lower levels can achieve such consistency. Although it is not easy to find direct statements of such an argument, I believe that it underlies a good deal of continued insistence on at least the weaker versions of reductionism. It also provides a link between two of the pillars of the mechanistic picture of the world that it is the aim of this book to oppose. Although I am here considering how an assumption about causality entails reductionism, it can equally well serve as an argument from the falsity of reductionism to the falsity of the view of causality in question.

The crucial assumption about causality is what I shall refer to as the assumption of causal completeness. This is the assumption that for every event there is a complete causal story to account for its occurrence. Obviously enough, this is a view of causality the roots of which are to be found in the soil of determinism. The paradigm of a complete causal story is the sufficient (and perhaps even necessary) antecedent condition provided by a deterministic causal explanation. However, since the issue of causal completeness at the microphysical level is central to the present discussion, and since microphysics is generally agreed, in the light of contemporary interpretations of quantum mechanics, to be irreducibly indeterministic, it is important to consider the indeterministic analogue of deterministic causal completeness. It is not hard to see what this should be. The basic idea is that there should be some set of antecedent conditions that together determine some precise probability of the event in question.

Events and their causal explanations are conceived at particular levels of organization. The mutual annihilation of an electron and a positron is an event at the subatomic level; my thinking of the previous example is an event at the mental level; a supernova is an event at the astronomic level; and so on. However, events at lower levels may very well be constituents of events at higher levels, and it is this that leads readily from the commitment to causal completeness to an argument for some kind of reductionism. The argument is most easily seen in terms of a simple example. Consider an electron in my index finger. As I move my finger to type the letter b on my keyboard, the electron must move. Causal completeness at the level of elementary particles implies that there is some condition of events at that level sufficient to explain the movement of the electron. Suppose, then, that I offer some causal story at the macroscopic level about the movement of my hand. For example, suppose my hand moved because I intended to type the letter b, or, for those who deny that this is a causal explanation, suppose that my finger was dragged down by my typing instructor.

This is Roger Sperry's insight that a turning wheel exerts downward causal control over its atoms.

Either story, if it explains the movement of my finger, also entails the movement of the electron. At the microlevel, meanwhile, there are presumably sufficient causes operating on every elementary particle in my finger. The causal efficacy of events at the macrolevel had better at least be consistent with all these billions of microlevel causal facts. Such consistency would appear to require either divinely preordained harmony, or reducibility at least in principle. This point can be made clearer by noting that the causal completeness of the microphysical directly contradicts the supposition that my intention (or coercion) could be necessary for the movement of the electron, except insofar as it is itself a necessary consequence of events at the microlevel. For the events at the microlevel are, ex hypothesi, themselves sufficient for the movement of the electron, or, in the indeterministic case, sufficient to determine the precise probability of its movement.

This is the basis of Jaegwon Kim's attack on non-reductive physicalism, the idea that mental events have causal powers

So it appears that events at the macrolevel, except insofar as they are understood as aggregates of events at the microlevel — that is, as reducible to the microlevel at least in principle — are causally inert. This, of course, is the classical picture of Laplacean determinism, except that it does not depend on determinism, only the causal completeness, or causal closure, of the microlevel. And metaphysically, though perhaps not epistemologically, this picture entails the causal impotence of events at the macrolevel.

A connection should be noted between this argument and compositional materialism. The negative claims of compositional materialism are surely necessary conditions for the credibility of causal completeness at the microlevel. Leaving aside the possibility of epiphenomenalism (and with even fewer qualms, psychophysical parallelism), the existence of nonphysical minds with any causal efficacy in the physical world makes causal completeness at the microlevel impossible. This is simply because the appeal to reducibility in principle is precluded when the higher-level causal agent is not, even in a merely compositional sense, identifiable with any set of microphysical entities. Similar consequences follow from interventionist deities, ghosts, and so on. Without reducibility in principle, the same goes for interventionist people, animals, or even rocks.

My reaction to this should be clear from what has gone before. Although I want no truck with the nonphysical substances just mentioned, a central purpose of the ontological pluralism I have been defending is to imply that there are genuinely causal entities at many different levels of organization. And this is enough to show that causal completeness at one particular level is wholly incredible. By contrast with even the weakest versions of reductionism, the pluralism I have in mind precludes the privileging of any particular level. My response to the argument just described has two main parts. The next two chapters will explain in some detail why reductionism fails in major parts of biology and psychology. This can be seen, among other things, as preparing the inversion of the reductionist modus ponens (causal completeness requires reductionism) into my antireductionist modus tollens (the failure of reductionism implies the falsity of causal completeness).

Finally I want to introduce in a general way what I take to be the problems with global reductionism, and also to say a little about the possible virtues of more local reductionistic strategies. A first difficulty, already mentioned, is the oversimplification involved in the hierarchy of levels of organization. To begin with, a very striking exclusion is the whole of macrophysics. Atoms or molecules go to make up not only living cells but also many complex nonliving things. At the very least, science must be depicted as a branched tree to accommodate gravitation, electromagnetism, cosmology, geology, and so on, although it is in fact not clear how even these branches should be rooted in microphysics. Nor can the objects with which the macroscopic life sciences are concerned generally be treated as structures solely of objects from the next-lower level. A sponge, perhaps, is a multicellular organism that is nothing but an assembly of cells. But an aardvark would more naturally be treated as an, assembly of organs and other complex systems. Moreover, many of its essential constituents, such as hormones, are not cells, or parts of cells, but molecules, whereas ions, also vital for its well-being, are more nearly atoms. So we find not only an intermediate level of organization, but also items from lower levels. Nor can these lower- level entities be treated merely as exogenous environmental variables. Their presence may often be determined by very complex events at the level of organs or organ systems. Consider, for example, the increase of blood adrenalin of an animal in danger, a chemical consequence of perhaps a complex perceptual event. This illustrates what I shall emphasize in the next chapter, the possibility that lower-level events may perfectly well be determined by what is happening at a higher level.

(The Disorder of Things, pp.99-102)

Although the problem of freedom of the will is generally understood as involving a conflict between human freedom and determinism, the rejection of determinism has not generally been perceived as providing a solution to the problem of human freedom. There are good reasons for caution here. It has occasionally been suggested that quantum mechanics, by showing that certain events can occur without determining antecedent conditions, shows how human actions could be free.

Chance is not the direct cause of action. This is the standard argument against
free will

Perhaps a movement is somehow generated by an amplified quantum event. But a solution to the problem of freedom of the will is not to be obtained by replacing the picture of a person as mindless machinery with that of a random action generator. The idea that people act for reasons seems more reconcilable with an account under which those reasons turn out to be nothing but states of the machine, than one which seems to preclude either causes or reasons by placing the action wholly beyond explanation. Nor is it helpful to move from determinism to a probabilistic uniformitarianism that claims that human actions are not determined by antecedent physical conditions, but only made more or less probable. This sounds not so much like an account of a (metaphysically) free person, as of a somewhat unreliable one.

But the effect on our conception of human freedom of relaxing the assumption of determinism can be seen quite differently. One of the problems with the quantum amplifier account of freedom is that, just as with human freedom under traditional libertarian accounts, the quantum indeterminacy is treated as a source of anomaly in a world otherwise driven by inexorable laws. A more radical rejection of complete and pervasive causal order suggests a quite different but much more promising approach.

To the philosophically uncorrupted, what is striking about humans as compared with other natural phenomena is not their unpredictability or their role as sources of disorder in an otherwise reliable and well-ordered universe. Quite the contrary: it is people who appear as somewhat orderly and predictable entities in a generally chaotic world. Where we find order, it is generally because it has been intentionally created by other humans...The point is just that humans have all kinds of causal capacities that nothing else in our world has: capacities most notably to make and execute all manner of plans capable of determining their behavior in complexly organized ways for considerable distances into the future.

There is no good reason for projecting these uniquely human capacities in a reductionist style onto inanimate bits of matter. Nor is there anything ultimately mysterious about particular causal capacities' being exhibited uniquely by certain very complex entities; no more than it is insolubly mysterious that certain weather patterns can generate tornadoes capable of picking up houses or that leguminous plants have the capacity to fix nitrogen. Thus the solution I propose to the free will problem is simply to recognize that what is problematic about humans is not that they are exceptions to an otherwise seamless web of causal connection, but that they are extraordinarily dense concentrations of causal capacity in a world in which such order is in short supply. Scientific investigation of the sources of these capacities may help to explain how this is possible; but this is to begin to explain, not, as determinists often want to suggest, increasingly to problematize, this remarkable feature of humans.

It is natural, on the view I am presenting, to feel considerable sympathy for one aspect of Kant's account of freedom in terms of autonomy (1785/1948). It is the ability to do what one thinks one ought to do rather than what, at that moment, one would like to do that is, arguably, uniquely human, and that makes possible the execution of long complex schemes. Unlike Kant, I do not mean to restrict what one ought to do to a narrow sense of duty (although conceivably this may serve to distinguish a qualitatively distinct and higher form of conduct), but to include any prudential "ought" driven by any kind of human project. Thus planning an armed robbery or a racist political advertisement is, in the sense I am considering, as much a display of uniquely human capacities as building a hospital or dismantling a nuclear weapons system.

The most relevant differentiation from Kant's position that I want to insist on is that without the commitment to determinism or even causal completeness, I am under no pressure to engage in the kind of metaphysical excesses in the noumenal world that have made Kant's position so implausible.

This is downward causation

I deny that there is any problem in reconciling the possibility of directed, intentional human action with the surrounding causal order...there is no reason why changes at one level may not be explained in terms of causal processes at a higher, that is, more complex, level. In the case of human action, the physical changes involved in and resulting from a particular action may perfectly well be explained in terms of the capacity of the agent to perform an action of that kind. Unless there is some reason to think that some part of the physical surroundings is undergoing some distinct and contrary causal process, there need be no conflict.

(The Disorder of Things, pp.214-217)

The Solution to the Problem of the Freedom of the Will, Philosophical Perspectives 10, 385-402 (1995)
How Much of the Free Will Problem Does (the Right Kind of) Indeterminism Solve?, Philosophical Inquiries, in press, 2013