(Article originally published in Libet, B., Freeman, A., and Sutherland, K. (eds) (1999), The Volitional Brain (Thorverton: Imprint Academic)
and 1999 Journal of Consciousness Studies 6(8-9), 201-24: journal web page: http://www.imprint.co.uk/jcs.html)
Abstract. Hume claimed that anything that happens must either be causally determined or a matter of chance, and that a person is responsible only for choices caused by the person’s character; so that if any sense is to made of free will and responsibility, it must be on the basis that they are compatible with determinism. In this paper I argue that Hume’s claim depends on a covert assumption that whatever happens to any system in the world must be either the only development of the system which is consistent with causal laws, or else a development which is random. I argue that it is a serious mistake to make such an assumption covertly; and that without this assumption, good sense can be made of a concept of free will and responsibility as being indeterministic, thereby providing a viable alternative to compatibilist views
It is philosophical and scientific orthodoxy that anything that happens in the world must either be causally determined or random. And many philosophers and scientists draw from this that, if any sense is to be made of human free will and responsibility, it must be on the basis that human choices and actions are cases of deterministic causation - because, it is said, random or chance occurrences cannot reasonably be regarded as exercises of free will for which a person may be responsible.
A contrary view, which I have supported in my book The Mind Matters (henceforth TMM) and other writings, is that the causation we observe in the physical world, apparently progressing conformably with some combination of deterministic laws and randomness, is just one kind or mode or aspect of causation; and that there is in addition another kind or mode or aspect of causation operating in the conscious decisions and actions of human beings, and perhaps also of non-human animals. I call the former ‘physical causation’, the latter ‘volitional causation’ or ‘choice’.1 Then, a concept of free will as something neither deterministic nor random can be developed on the basis of the notion of volitional causation or choice.
The orthodox position on this kind of approach is that any causation involved in our conscious decisions and voluntary actions is nothing other than a type of, or a working out of, the ordinary physical causation of rules and randomness; or to put it another way, that the exercise of choice is just one of a number of levels of operation of physical causation.
And many philosophers and scientists go further, and say that a notion of volitional causation or choice as being distinct from ordinary physical causation is inconceivable, or at least logically impossible; and that in any event such a notion could have no bearing on questions of free will and responsibility. There are two strands to the argument in support of this view, both of them exemplified in the writings of David Hume: (1) whatever happens must either be causally determined by its antecedents or else be a matter of chance; and (2) for a person to be responsible for an action which that person chooses, the choice must be caused by the person’s character.
In this paper, I set out to show that this view is a serious mistake. Because of its close association with David Hume, I call it Hume’s mistake.
1 Hume’s argument
Hume’s argument (which I take from An Enquiry Concerning Human Understanding, Section VIII) begins with the observation, based on Newton’s physics, that when physical forces are applied to matter, the outcome is necessary in the sense that no other outcome is possible. However, he says, we know this only because of repeated observations of such occurrences, not because of any logically necessary connection between cause and effect.
In the same way, we repeatedly observe human beings behaving in similar ways in similar circumstances: pursuing happiness, avoiding pain, and so on. In so far as people behave differently in similar circumstances, this is because they have different characters; and so this is by no means inconsistent with regular conjunction of like causes and effects. Even when particular actions seem ‘out of character’, this does not mean that they are not the outcome of ordinary regular causation: rather, our difficulty in explaining them must be because of the ‘minuteness and remoteness’ of the causes, or ‘the secret operation of contrary causes’ - and Hume gives the analogy of a tiny piece of dust interfering with the proper functioning of a watch.
To Hume, there was no question of human reason interfering with the regular operation of causes in our motivation. Hume distinguished sharply between beliefs and desires, and argued that only desires could motivate us: reason could not prevent us acting in accordance with our desires, except by giving rise to a contrary desire; and we always act in accordance with the preponderance of our desires.
Our failure to acknowledge that our actions are determined in the same way as the processes of physical matter, Hume argued, is due to two factors.
(1) We tend to think there is a necessary connection between cause and effect in the physical world, and we see our own actions as not having any necessary connection with its antecedents in our character and circumstances. However, Hume claimed, there is no logically necessary connection between cause and effect in the physical world, just the regular conjunction of causes and effects - and the same is true of our actions.
(2) We sometimes seem to have a ‘liberty of indifference’ between alternative courses of action open to us, and still to be able to select between them. However, this feeling is unreliable and misleading; and when we reflect on what we and others have done, we can see that it was in fact the causal product of character and circumstances.
None of this, Hume argued, is inconsistent with personal liberty (by which Hume meant what we would call free will), because liberty means our power of acting according to the determinations of our will, that is, as we choose - even though our choice is in fact determined by causes. It is constraint, not determination by causes, which can deprive us of this liberty. If our actions were not determined by causes, then they could only be a matter of chance, so that they could not be an exercise of personal liberty. (Hume in fact asserted that chance has no existence; but even if, consistently with orthodox quantum mechanics, one accepts that there is objective chance in the world, in the sense of random events, one may agree that chance in that sense would not be conducive to what we ordinarily regard as choice.)
Hume went on to argue that his version of determinism was not inimical to our common-sense ideas of responsibility, but on the contrary was necessary to make sense of responsibility. This, he said, was for two main reasons.
(1) A person is not responsible for actions unless they proceed from a cause in the character or disposition of the person. The more premeditated an action is, the more we regard it as caused by the person’s character and the more responsible we consider the person to be. On the other hand, we consider people less responsible for actions performed hastily.
(2) The imposition of rewards and punishments for actions for which we suppose people are responsible makes sense only if we believe that rewards and punishments have consistent and regular effects on their behaviour; that is, if we believe that the prospect of rewards for certain behaviour and the prospect of punishment for other behaviour will consistently influence people to engage in the former and to refrain from the latter.
2 More recent statements
Hume’s approach has remained fashionable right up to the present, despite some persuasive opposition (e.g. Anscombe 1971, Nozick 1981, Flew and Vesey 1987). I will give some examples of more recent restatements by philosophers of essentially the same line of argument.
The most colourful of these is R. E. Hobart’s 1934 article ‘Free will as involving determinism and inconceivable without it’. Hobart writes:
Indeterminism maintains that we need not be impelled to action by our wishes, that our active will need not be determined by them. Motives ‘incline without necessitating’. We choose amongst the ideas of action before us, but need not choose solely according to the attraction of desire, in however wide a sense that word is used. Our inmost self may rise up in its autonomy and moral dignity, independently of motives and register its sovereign decree.
Now, in so far as this ‘interposition of the self’ is undetermined, the act is not its act, it does not issue from any concrete continuing self; it is born at the moment, of nothing, hence it expresses no quality; it bursts into being from no source. The self does not register its decree, for the decree is not the product of just that ‘it’. The self does not rise up in its moral dignity, for dignity is the quality of an enduring being influencing its actions, and therefore expressed by them, and that would be determination. In proportion as an act of volition starts of itself without cause it is exactly, so far as the freedom of the individual is concerned, as if it had been thrown into his mind from without - ‘suggested’ to him - by a freakish demon ... In proportion as it is undetermined, it is just as if his legs should suddenly spring up and carry him off where he did not prefer to go. Far from constituting freedom, that would mean, in the exact measure in which it took place, the loss of freedom. It would be an interference, and an utterly uncontrollable interference, with his power of acting as he prefers. In fine, then, just so far as the volition is undetermined, the self can neither be praised nor blamed for it, since it is not the act of the self.
More prosaically, in 1956 Alfred Ayer wrote:
But now we must ask how it is that I come to make my choice. Either it is an accident that I choose to act as I do or it is not. If it is an accident, then it is merely a matter of chance that I did not choose otherwise; and if it is merely a matter of chance that I did not choose otherwise, it is surely irrational to hold me morally responsible for choosing as I did. But if it is not an accident that I choose to do one thing rather than another, then presumably there is some causal explanation of my choice: and in that case we are led back to determinism.
And in 1961, J. J. C. Smart advanced a similar thesis in his article ‘Free will, praise and blame’. He argued that every event is either causally determined or, if not, it is due to chance: there is no third possibility other than ‘this event happened as a result of unbroken causal continuity’, and ‘this event happened by pure chance’ (p296). That is, he contended, there is no logical room between determinism and chance.
Galen Strawson’s 1986 book Freedom and Belief contains very full discussions of the first strand of Hume’s argument. At pp52-4 he expressly considers a ‘Leibnizian’ view that ‘reasons for action affect agents’ decisions, but in so doing only incline them towards, and do not necessitate them in, particular decisions to perform particular actions’. He asks: ‘upon what, exactly, are the agent’s decisions about actions now supposed to be based, other than upon its reasons?’, and goes on:
The trouble with the picture is familiar. If the agent is to be truly self-determining in action this cannot be because it has any further desires or principles of choice governing the decisions about how to act that it makes in the light of its initial desires or principles of choice. For it could not be truly self-determining with respect to these further desires or principles of choice either, any more than it could be self- determining with respect to its initial desires or principles of choice. But if it does not have any such further desires or principles of choice, then the claim that it exercises some special power of decision or choice becomes useless in the attempt to establish its freedom. For if it has no such desires or principles of choice governing what decisions it makes in the light of its initial reasons for action, then the decisions it makes are rationally speaking random: they are made by an agent-self that is, in its role as decision-maker, entirely non-rational in the present vital sense of ‘rational’ - it is reasonless, lacking any principles of choice or decision. The agent-self with its putative, freedom-creating power of partially reason-independent decision becomes a some entirely non-rational (reasons-independent) flip-flop of the soul.
Finally, as we will see, there are reflections of this same argument in recent criticisms of suggestions made by me and others that the indeterminism of quantum mechanics is relevant to freedom of the will.
3 Hume’s mistake
The fundamental mistake underlying Hume’s position and that of his twentieth-century followers is that they assume without discussion or disclosure (indeed sometimes assume as if it were a logical necessity) something very close to what they claim to prove: they assume that whatever happens to any system must be either (1) the only development of the system which is consistent with universal and impersonal causal laws, or (2) a development which is random, albeit perhaps complying with probability parameters established by such laws. It is hardly surprising that, if you accept that, then choice, as something other than the physical causation of rules and randomness, will be excluded; but the most significant step in the argument is made covertly by an assumption which is given no justification.
The history of science since the time of Newton provides some explanation of why many philosophers have made this assumption, and why they apparently have not even realised it was an assumption that needed justification. In a word, this assumption seems to arise from (1) the progressive success of science in explaining the processes of the physical world in terms of laws of nature and randomness; (2) the fact that our brains are part of the physical world; and (3) the apparent dependence of our choices on processes in our brains of which we are unaware. Now these three factors provide arguments supporting the assumption underlying Hume’s position, but they are far from conclusive - and there are contrary arguments which could be stronger. So the assumption should not have been made covertly: rather, it should have been presented as a hypothesis to be confirmed or refuted.
Associated with this fundamental mistake are a number of other mistakes apparent in Hume’s discussion, and at least implicit in the work of his followers.
First, there is the assimilation of motives for action to physical forces in Newtonian physics. Forces in Newtonian physics are measurable, commensurable, and conclusive: any piece of matter will necessarily be accelerated precisely as required by the vector sum of all forces acting on it. Hume claimed that we are motivated only by desires; and he assumed that desires like forces are commensurable, and that we always act in accordance with the preponderance of our desires. Both of these assumptions are questionable, indeed I would say false; and it is a serious mistake not to recognise the need to justify them. Alternative views, namely that our motives are diverse and incommensurable, that motives are characteristically inconclusive, and that a preponderance of motivation is established only by a choice, are ignored by Hume and his followers, or at most given lip service only.
Second, there is the associated assumption that, if reasons alone do not determine a choice in this quasi-Newtonian way, then reasons plus something else must determine it in this way; and that otherwise the outcome has to be random. This ignores the possibility that outcomes may be selected by agents on the basis of non-conclusive reasons - the selection requiring a qualitative judgement, rather than either a quantitative computation or a mere factual preponderance of some quantity of ‘desire’.
The hold that this assumption has on philosophers today is illustrated by the lengths to which even philosophers who are sympathetic to the idea of free will have gone, in their efforts to explain how choice could work. In order to account for a choice to do one thing rather than another, which gives effect to some reasons and not others, these writers have felt it necessary to invoke something else, other than the reasons and the choice - a clincher, as it were - to explain why one set of non-conclusive reasons prevailed over another set of non-conclusive reasons.
For example, Harry Frankfurt (1971) and Charles Taylor (1976) suggest that there is a higher-level evaluation of our conflicting reasons; although this of course gives rise to the question of what determines the result of this higher-level evaluation. Thomas Nagel (1986, pp116-7) suggests that the only way to complete an explanation of why an agent acted for certain reasons, rather than refraining from acting for other reasons, is to trace the cause or explanation into the formative causes of the agent’s character (cf. TMM pp391-2). And Robert Kane (1985, p85) seems to suggest that the choice must be clinched by something akin to coin-tossing or dice-rolling; although in a more recent book (1996), Kane has moved to a position closer to my own: I will specifically consider this book shortly (in Section 5).
A recent example of this quest for a clincher is McCall (1994, p277), where we find the following:
As we saw above, an indeterministic chess-playing machine could employ a randomiser to create options, but if so the selection of one of these options to be acted upon was by sheer chance. This is not good enough. The brain not only needs to create options, using something like the indeterministic intermediate-level mechanism just described, but it also needs to evaluate its options and arrive at an order of preference. It then selects its best option and acts on it. Now the process of evaluation and the selection of the best option need not be indeterministic. One would hope in fact that they were not: that the establishment of a ranked list of options was in accordance with set aims and goals, and that the brain could use a rational procedure like the hedonistic or utilitarian calculus, or the principle of universalizability and the categorical imperative, or for that matter a rigid and complete set of rules which in casuistic fashion covered every conceivable alternative, to determine its best course of action. The upshot would be that the initial indeterminism that created the choice-set was subordinated to a higher selection process which could be entirely deterministic.
To me, it seems clear that, if the supposedly open alternatives are set up only for a selection to take place by a deterministic process, then there is no reason why the alternatives cannot be set up by a pseudo-random (deterministic) process, as suggested by Dennett (1984) - and we are no closer to anything like genuine choice.
Third, there is the contention that responsibility requires that actions be caused (along with the operation of desires in this quasi-Newtonian way) by character, and that the only alternative is chance. This is at the heart of Hobart’s contentions, as well as Hume’s. What it ignores is that an action could perhaps be caused by a person’s character in the sense that it is chosen by the person from alternatives limited by the person’s character, on the basis of non-conclusive reasons which appear as they do and carry the persuasive force that they do because of the person’s character. The choice could then be not predetermined yet not random; and the person’s character could be important to the choice because it limits the available alternatives and affects how they are presented and thus inter alia how easy or how hard it is for the person to make one choice rather than another.
Fourth, there is the suggestion that without determinism, there could not be the broad predictability of human behaviour, both in general and in relation to a particular person, or the susceptibility of persons to encouragement by reward or deterrence by punishment. This ignores the obvious fact that, even if motives never necessitate actions by normal persons, they can very strongly incline them to act in certain ways; so that it may be very hard for them to act otherwise, and very unlikely that they will do so.
Mistakes of this kind are made even in Galen Strawson’s careful discussion. He assumes (1986, p54) that a decision, on the basis of reasons which do not necessitate the decision, must either be governed by ‘further desires or principles of choice’, or else be random as ‘lacking any principles of choice or decision’. This is a false dichotomy, because it assumes what may not be the case, namely that the only possible connection between reasons and decision is one of determination (in which reasons determine decisions) rather than choice (in which decisions are made by an agent on the basis of non-conclusive reasons). That is, like Hobart, Ayer, and Smart, Strawson ignores the possibility that choice is a mode of causation in which an agent resolves non-conclusive reasons; not for any other reason, but by means of a capacity, which all normal human beings have - and can’t help having - to make such choices.2 And in suggesting that any undetermined part of a choice would have to be nonrational, Strawson ignores the possibility that human rationality itself may not be fully captured by rules and may thus be partly indeterministic - a possibility that I have strongly supported in Chapter 5 of TMM and in Hodgson (1995).
4 How it could be
Given that our brains are physical systems which, like other physical systems, appear to change over time in accordance with universal and impersonal laws, how could choice be anything other than the working out of such laws?
4.1 The general picture
It would seem that choice must either be (1) inconsistent, or (2) consistent, with the physical causation of rules and randomness; and if consistent must either (a) select the only result consistent with physical causation or else (b) select between two or more results all consistent with physical causation. Alternative (1) seems unlikely: although our understanding of the brain is far from complete, I think we can be fairly confident that no actual violations of physical laws are involved in volitional causation or choice. Alternative (2)(a) is in effect Hume’s position, and that of mainstream science and philosophy today. Alternative (2)(b) is the alternative I wish to explore.
The general picture that I suggest is as follows. What I propose is that, while the human brain-and-mind is a single system, the way this system changes over time in normal persons is best understood in terms of two overlapping causal histories, each of which is intellectually respectable and valid; but which are complementary, and are mutually irreducible in the sense that neither can be fully explained in terms of the other. One is the objective history of the brain, to which the concept of physical causation is applicable; while the other is the subjective history of the mind, to which the concept of volitional causation or choice is applicable. I suggest that each of these two histories is incomplete on its own, yet not inconsistent with the other. On the one hand, while the objective features of the brain which correlate with conscious experiences and actions will undoubtedly come to be specified with increasing accuracy, a person’s conscious actions may not be uniquely determined by the development over time of those ‘neural correlates’ in accordance with universal laws; and on the other hand, while a person’s conscious actions are uniquely determined by the person’s choice, made for reasons, those reasons may not capture all the unconscious or non-conscious factors which contribute to the actions.
4.2 An example
To put this in more concrete terms. Suppose that I am trying to decide what to do with $500 which I have. It occurs to me that I could spend it on a new amplifier for my sound system, or send it for famine relief; or buy a cheaper amplifier for $250 and send $250 for famine relief; or save it for later. Suppose that, in addition to thinking about such things as the unsatisfactory performance of my present amplifier and how much better music would sound from a new one, and the needs of the people affected by the famine and how I should help satisfy them, I also think such things as ‘famine relief will be provided whether I contribute or not’, ‘very little of what I contribute would get to where it is needed’, and ‘aid is counter-productive because it leads to a vicious circle of dependency’. Let’s say I decide to spend the $500 on an amplifier.
4.3 Resolving inconclusive reasons
The first point to make is that however fully and honestly I try to express my reasons, no verbal expression of my reasons will ever be conclusive of one result or another. So far as my subjective motivation can be expressed, before decision there are only inconclusive reasons, and it is my decision which makes some of them prevail over others. Yet, assuming I have expressed my reasons fully, there are no other reasons for my decision. My decision was a choice which resolved the issue, not a matter of chance: apparently, what was conclusive was not a further reason which arbitrated between the inconclusive conflicting reasons, nor was it a chance occurrence - rather, it was a choice based on these very reasons and no others, notwithstanding that they were, prior to the choice, inconclusive. Seemingly, that is the nature of reasons and that is the nature of choice: there is no clincher, other than the choice itself.
It might be argued that, although I cannot express my subjective reasons in a way that makes them conclusive, they must in fact come with ‘weights’ which, although I am unable to measure them, do in fact determine the issue. Certainly, my reasons do feel to me to be more or less persuasive, to press strongly or not so strongly. Even if the respective strengths or weights of reasons cannot be measured by me and are not commensurable to me, it can be argued that my reasons must have neural correlates with properties which, like any physical properties, are quantitative and sufficiently commensurable to be subject to the quantitative laws that govern all physical systems: this is in substance the mainstream view, and I cannot say it is not a possible view.
What I do say however is that another view is also possible - and in fact I have argued in TMM and elsewhere that this other view is the more reasonable and probable. In short, if a choice is in fact just the working out of quantitative physical laws acting on physical systems with quantitative physical properties, why are there any subjective feelings at all?
4.4 Character and choice
It might also be argued that my account leaves out the role of my character in the process of choice: the reasons affect me as they do because of my character, and my character in turn is embodied in the physical system which is my brain. It is my character which makes the outcome of decisions like this reasonably predictable: if I am a selfish person, the amplifier will win hands down; while if I am altruistic, the famine relief will win.
I accept that my brain is the immediate physical manifestation of my character, and that my brain has physical properties relevant to my choice which are expressive of my character. These physical properties will limit the alternatives open to me: it is because I am as I am that the alternatives which occur to me are to spend the money on an amplifier or on famine relief. These properties will affect the way I see the alternatives, and the appeal which the conflicting reasons have: the more selfish I am, the stronger and more persuasive will seem the reasons favouring the amplifier; the more altruistic I am, the stronger and more persuasive will seem the reasons favouring famine relief. This will mean in turn that the likelihood of one or the other outcome is affected by my character - but it does not mean that any outcome must be inevitable or predetermined.
Rather, a choice is made by me, a person having a certain character and also a capacity to choose - and since I have this capacity to choose, it is I, and not just my character, that makes the choice. I suggest that people are no different from each other in respect of their capacity to choose, except in so far as there are differences in their characters, manifested in their physical brains and thus in the alternatives available to them and the way these alternatives appear. And if this is the case, then there is no unfairness in attributing responsibility for choices to persons, considered as having both character and capacity to choose - with degrees of responsibility affected by those advantages and disadvantages of character which are beyond the person’s control. I say a little more about this in Section 4.8.
Furthermore, there is a possible and plausible physical correlate to this picture. If orthodox quantum mechanics is broadly correct, it is conceivable that, immediately prior to my decision, my brain is in a superposition of states (cf. Nozick 1981, pp298-9): if the decision is ultimately between just the two alternatives of spending the whole of the $500 on the amplifier or giving it all to famine relief, the superposition could be of just two states, one for each alternative. If I am a selfish person, the quantum mechanical probabilities of the two states might be 0.95 for the amplifier and 0.05 for famine relief; if I am altruistic, they could be the other way round. When one result occurs, according to quantum mechanics this will be a random event within those probability parameters; but from the subjective point of view, it could be my choice, with the actual choice having been made likely (but not inevitable) or unlikely (but not impossible) by the felt strength or persuasiveness of the reasons. On this approach, the objective probabilities determined by the physical aspect of the causal process could be broadly reflected in the felt strength of the reasons operative in the mental-volitional aspect of the causal process - although of course, whereas objective probabilities are commensurable, I am proposing that the felt strengths of different reasons are not commensurable, and that the result requires a choice by me.
Next, it could be argued that this account presupposes that I have a clear understanding of my reasons or motives - whereas in fact there will be reasons or motives operating of which I am unaware, or not fully or accurately aware. For example, the three reasons given above in quotes, although seen by me as genuine reasons, may more fairly be considered rationalisations produced unconsciously to justify a selfish decision. If I am a selfish person, given to rationalisation and self-justification, then this may be a more accurate account of how I made the choice than one which presents the choice as a wholly conscious resolution of genuine well-understood reasons.
I see this as a matter of degree. Undoubtedly, we are not altogether clear-sighted about our true motivation, but this does not mean that our choices are not genuine choices based on the reasons as we see them to be. In my example, if I do have sufficient insight into my own character and motivation, I may question my reliance on the quoted reasons, and ask whether I am not just rationalising a selfish choice. To undertake this questioning would itself be a choice between alternatives thrown up by my character, made on the basis of further non-conclusive reasons; and the pursuit and outcome of this questioning may involve further such choices. In general, the better my understanding of myself and my motivation, the more complete is my authorship of and my responsibility for my choice: this is the place I see for the higher level of evaluation of conflicting reasons, considered in Frankfurt (1971) and Taylor (1976). However, I would argue that all normal adult human beings have sufficient insight into their motivation to be regarded as responsible for their voluntary choices and actions.
More broadly, I believe that there is choice wherever there is consciousness: the very function of consciousness is to allow for choice from available alternatives on the basis of consciously-felt reasons. Thus, if as I believe at least some non-human animals are conscious, it would follow that these animals make choices. But I suggest that their rationality and their insight into their own motivation is insufficient for them to be considered as having free will or responsibility for their actions; whereas the rationality and insight of normal adult human beings, even though far from complete or perfect, is generally sufficient for them to be considered as having free will and responsibility.
4.6 Longer-term choice processes
I noted a moment ago that in the course of choosing what to do with my $500, I might pursue subsidiary questions which themselves involve choices. I suggest that this in fact exemplifies a very usual feature of our decision-making. In any extended process of making a choice, it would be surprising if the alternatives which were possible at the beginning of the process, and which then had certain objective probabilities, all remained available with the same probabilities at the instant before the final decision; and if only at the final decision was there suddenly just one elimination of all but the chosen alternative. A far more plausible scenario would be that throughout the process, volitional causation was adjusting the probabilities of alternatives, eliminating some and introducing or varying the probabilities of others; so that immediately before the choice, physical causation could give a very high objective probability for an alternative which had a much lower probability at the outset of the process, and the ultimate selection of this alternative could then follow almost as of course.
Indeed, in my example it could be that I have over many years been thinking about the conflicting claims of self-interest and altruism; and my decision about the $500 could be an outcome of many long processes of volitional causation during which very many intermediate choices have been made. It could also be that, if objective probabilities for my decision about the $500 could have been calculated at the outset of these processes, they would have favoured one result (say, famine relief), whereas immediately before the decision, because of the intermediate choices, they very strongly favoured the other result (the amplifier). Then, my choice of the amplifier could fairly be regarded as a free choice for which I am responsible, even though by the time it came it was almost inevitable.
This can be applied to an example discussed elsewhere in this context (see Dennett 1984, p133). Martin Luther’s decision to defy the Church of Rome can be considered a free choice, for which he was responsible, even if at the time of his final decision he spoke substantially the truth when he said ‘Here I stand. I can do no other’ - because it was the outcome of a long process of choice-making in the course of which many intermediate choices had been made, and it was these intermediate choices which made the ultimate choice virtually inevitable (cf. Kane 1996, pp38-40).
And this line allows a further response to Hume’s point about responsibility and character. I have argued that character could be important as both limiting alternatives for choice and affecting how they are presented, rather than by simply pre-determining what is done or chosen, leaving the person (not just his or her character) responsible to a greater or lesser degree for the action or choice; and this means in turn that, in so far as the person’s character has been affected by prior choices of this kind, the person has some responsibility for character as well. A similar point made by Aristotle is dismissed by philosopher Bernard Williams (1995, p27) as ‘hopeless’, on the grounds that it ‘shifts the question back’ and ‘offers very rough justice’. However, rough justice is better than no justice; and so long as there is some responsibility for all choices, the justice is in fact not all that rough. So I think Aristotle was on the right track, and that Williams is led astray by Hume’s mistake.
4.7 Kinds of motives
My $500 example was one where there were two main motives operating, one being a non-moral motive and the other being a moral or altruistic motive; and indeed I formulated the example as one where there were two conflicting motives plainly of different types, so that their incommensurability was obvious. However, I want to make it clear that I am not saying (as have some philosophers, such as C. A. Campbell 1951) that there are just two types of motives, namely non-moral motives or ‘desires’, and moral motives or ‘duties’: rather I say that there is no sharp distinction between non-moral and moral motives, that there are many different kinds of motives, and that it is a very general characteristic of motives that they are incommensurable and inconclusive.
To give some examples. A runner in a marathon is in pain after 38k and wants to stop running so the pain will stop, but is determined to finish: there are two conflicting motives here, both basically selfish, but not commensurable. I am at a restaurant deciding between the fish and the beef, I generally prefer beef but I had beef yesterday, and I choose the fish: again, I do not believe the conflicting motives are commensurable. Nor do I believe that the choice in either case is random, although the opposing reasons are inconclusive and there are no other reasons. Of course, sometimes we do something akin to mentally tossing a coin because it is more important that a decision be made than that it go one way or the other (as it was for Buridan’s ass, which supposedly died of hunger and thirst because it could not choose between two sources of food and water located at points equally distant from it) - but I do say that is not the way we generally resolve conflicting inconclusive reasons.
We also sometimes have a choice as to what to believe, as well as what to do; and in these cases the conflicting reasons may have little to do with either self-interest or altruism, but may still be inconclusive and incommensurable. This is particularly clear in the case of a judge deciding the facts in a court case: she has to reach a belief, at least on the balance of probabilities, as to whether the facts are such as to entitle the plaintiff to a court order of the kind that the plaintiff is seeking. The conflicting evidence and arguments are generally inconclusive: the clincher is not some further reason nor the tossing of a coin, but the conscious decision based on these very reasons. This matter is considered in Chapter 5 of TMM and Hodgson (1995): 3 the discussion there strongly supports the view that rationality is not fully captured by rules, and so undermines Strawson’s assumption that indeterminism implies nonrationality.
My examples do not show that the orthodox view is wrong: what they do show, I contend, is that it is wrong to suggest that no other view is even possible, on the basis that anything that happens must as a matter of logic be predetermined or random. It is that suggestion, exemplified in Hume’s writing, which is a fundamental mistake that infects and distorts much current thinking about the mind in general, and about freedom and responsibility in particular.
An important feature of this account is that it has an answer to the basic dilemma about responsibility, which is strongly expressed by Galen Strawson (1986, esp. at pp311-2):
(1) There is a clear and fundamental sense in which no being can be truly self-determining in respect of its character and motivation in such a way as to be truly responsible for how it is in respect of character and motivation.
(2) When we act, at a given time, the way we act is, in some quite straightforward sense, a function of the way we then are, in respect of character and motivation. We act as we act because of how we then are, in respect of character and motivation.
(3) It follows that there is a fundamental sense in which we cannot possibly be truly responsible for our actions. For we cannot be truly responsible for the way we are, and we act as we act because of the way we are.
My account accepts that nothing an agent does or can do at the time of any choice or action can make the agent responsible for what alternatives are then available to the agent, for the way those alternatives then appeal, or indeed for having the capacity to choose between them - but claims that, leaving aside any question of responsibility for these matters, the agent can still be responsible for the way the agent exercises the capacity to choose. The way the agent is, in respect of character and motivation, does not determine what the agent does: it only determines what the alternatives are and how they appeal. The agent is not responsible for having the capacity to choose between these alternatives, but is not constrained in how that capacity is exercised, even by the way the agent then is in respect of character and motivation; and so the agent (and no-one and nothing else) is responsible for the way that capacity is exercised. That responsibility may be greater or less, by reason of the nature of the choice posed by the way agent now is - the harder it is, by reason of the agent’s inclinations, for the agent to make the ‘right’ choice, the less blameworthy will be the ‘wrong’ choice - but on this view, normal adult human beings always have some responsibility for their choices and voluntary actions (see Hodgson 1996, 1998).
And this means in turn that the agent may have some responsibility through prior choices for the way the agent now is in respect of character and motivation, and thus for presently-operating reasons and the way they appeal - thereby vindicating Aristotle against Williams.
5 Robert Kane
In Section 3, in discussing the search for a clincher, I suggested that Kane (1985) proposed that conflicts between competing inconclusive reasons were resolved at random. In his 1996 book The Significance of Free Will, Robert Kane has substantially developed and modified his views on free will, and has dealt carefully and at length with suggestions like this one of mine. In this Section, I will make a brief comparison of Kane’s approach with my own, which may be of interest to readers familiar with Kane’s book: others may prefer to go straight on to Section 6.
There is in fact little if anything I disagree with in Kane’s development of his account of free will in Chapters 8 to 10 of this book: indeed, that part of the book endorses four of the key contentions that I argued for in earlier writings - as well as in this paper.
(1) The contention that, prior to a choice being made, an agent’s reasons are characteristically inconclusive, inter alia because they are incommensurable; and that it is the agent’s choice or decision which resolves the issue (TMM pp133-5, Hodgson 1994). Kane endorses the idea of incommensurability at p167; and at p133, he postulates that in situations where an agent has to choose between alternative courses of action and has reasons or motives supporting each alternative, the agent makes one set of reasons or motives prevail over the others by deciding.
(2) The contention that what the physical perspective can only treat as a chance occurrence may correctly be seen from the mental or experiential perspective as an agent’s choice (TMM pp389-93, 444-7; Hodgson 1994). Kane says, at p147, that from the physical perspective, free will looks like chance - that from a physical perspective, there is just an indeterministic chaotic process with probabilistic outcomes; but experientially considered, the process is the agent’s effort of will and the single outcome is the agent’s choice.
(3) The contention that the problem of free will is closely interlinked with the problems of consciousness and of the indeterminism disclosed by quantum mechanics (TMM passim and esp. pp393-4; Hodgson 1994). Kane asks, at p148, ‘How can a physical process of the brain be at the same time a consciously experienced effort of will?’; and suggests that this is just part of the mystery of ‘how neural firings in the brain could be conscious mental events.’ And on pp150-1, he suggests it is also implicated with the general problem of indeterminacy-in-nature introduced by quantum physics.
(4) The contention that the objective probabilities for various outcomes are to some extent reflected in the subjectively-felt strength of reasons; and that rational decisions may nevertheless be made in favour of actions with lower antecedent probabilities (TMM pp392-3). At p177, Kane points out that antecedent probabilities of available alternatives do not necessarily indicate which of them are more or less rational for the agent to choose.
However, Kane’s position might seem at variance with mine in that, whereas he apparently eschews special forms of agency or causation, I treat the distinction between physical causation and volitional causation as central to my argument in this paper. But I believe that a closer consideration shows that his rejection of special forms of agency or causation is qualified in a way which makes it less than clear; and that the latter part of his book, where he comes close to my position, does involve an implicit introduction of a special form of agency or causation.
In order to develop this point, it is necessary first to say a word about terminology. Although Kane argues that free will, in the usual sense of involving available alternatives and substantial responsibility (Kane talks about ‘ultimate responsibility’) is incompatible with determinism, he nevertheless acknowledges that there are legitimate senses of what he calls ‘free agency’, not involving ultimate responsibility, that are compatible with determinism.
At p115, Kane notes that, in order to account for how an agent can choose rationally, voluntarily, and with voluntary control, advocates of free will have conjectured that there must be some explanatory factor over and above past circumstances and laws of nature; and that they have looked for this factor in special forms of agency or causation. At p116, Kane eschews any appeal to any special forms of agency or causation that are not also needed by deterministic accounts of free agency. But determinists who accept reductionism can simply claim that free agency just is a particular kind of working out of past circumstances, laws of nature, and chance, expressed in a different language; whereas, as noted above, Kane argues that agents make one set of reasons prevail over another, and that although this looks like chance from a physical perspective it is, experientially considered, the agent’s choice. So surely he is saying, as I do, that in the exercise of free will something more is happening over and above the working out of past circumstances in accordance with laws of nature, and chance - this ‘something more’ involving the conscious activity of the agent.
Now Kane may accept this, and still claim that it does not involve a special form of agency or causation not also needed by determinist accounts of free agency; but if so, he needs to explain how the causation involved in making something happen relates to the rules-and-randomness causation presupposed by the physical sciences. Perhaps Kane assumes, as I do, that there is a single ongoing causal process involved in the exercise of free will, and perhaps he is looking for a unitary concept for, and a unitary account of, this physical-and-mental causation. But I think it is preferable to recognise that we do not at present have an adequate language for such a unitary account, and to make the best use we can of the concepts which we actually do have. Kane accepts, as I do, that agents cause things by doing them, in a way not apparent to the physical viewpoint - and I say that what is involved here is best classified and developed with reference to the distinction I have drawn between physical causation and volitional causation.
One other possible difference between Kane’s position and mine is that he appears to distinguish sharply between theoretical reasoning or deciding what to believe, and practical reasoning or deciding what to do (pp22-3), and he considers free will only in relation to the practical reasoning; whereas I closely link free will with consciousness and plausible reasoning generally, and I suggest that volitional causation can be exercised in deciding what to believe, as well as what to do. This brings out what I see as the important link between volitional causation and rationality; and it makes possible an account of volitional causation and free will that makes them a natural and vital part of human activity, and understandable as a product of evolution.
6 Quantum mechanical randomness and choice
An echo of Hume’s mistake can be found in recent attacks on suggestions that quantum mechanics may be relevant to free will and responsibility. The idea that the indeterminism of quantum mechanics leaves room for the operation of free will was put forward as long ago as 1927 by Arthur Eddington (1929, Ch14); and more recently it has been supported by people such as John Eccles (1994) and myself (TMM; Hodgson 1994). One standard response to this suggestion has been that the exercise of choice is actually incompatible with the randomness which is the only indeterminism postulated by quantum mechanics.
In TMM , I argued that what, to the mental viewpoint, was a choice made between available alternatives for non-conclusive reasons, might appear to the physical viewpoint as the outcome of a combination of deterministic and random processes. I suggested (TMM p392) that ‘the quantum physical probabilities of the alternatives [for choice] may be related to the consciously-felt weight of competing reasons for the different actions’:
Where the alternative which occurs is an action which was objectively highly probable, one might have a case of an action with small ‘mental input’, an action done virtually by habit or as of course. Where the alternative which occurs is an action which was objectively improbable, perhaps one could have a case of concentration, effort, exercise of ‘will-power’. Where a decision has to be made between alternatives of objectively similar probability, perhaps the weighing, judgemental aspect of decision-making is especially exercised.
I noted the standard response of incompatibility (TMM pp392-3):
In similar vein, there is a reference in Gomes (1978: 450) to an argument by Schrodinger to the effect that quantum physics is in fact inconsistent with free choices being made between superposed alternatives of quantum states, because those free choices could violate the statistical predictions of quantum physics. Even if this were correct, it would not be conclusive: it could simply mean that, when choices are made, the ‘known’ as represented in the relevant state function is not all there is to know, because these is a mental element involved which cannot be so represented. In any event, the argument seems to overlook that every choice is a unique event: each person is unique, and a matter for choice cannot be precisely one which has previously faced the person (that would itself be a difference!). Accordingly, the quantum physical superposition of alternatives corresponding to each choice must itself be a single unique highly complicated quantum state; so that there could never be a series of like state reductions in which statistical effects could be disclosed.
In his review (Lockwood 1991) of my book, Oxford philosopher Michael Lockwood took up this objection:
the widespread occurrence of ‘improbable’ outcomes, even with respect to individually unrepeatable situations, would still, in aggregate, violate the quantum mechanical statistical algorithm. Thus, if the quantum statistics are to be preserved, exercisers of Hodgson’s brand of free will would have to be subject to some sort of non-local mutual constraint - rather like contestants in a race, each of whom may in some sense have the ability to win, even though none can do so save to the exclusion of the others. (Perhaps we have here the basis of a novel approach to the problem of evil!).
The incompatibility argument is also raised in the review of my book by Jeffrey Barrett (1994).
In Eccles (1994), John Eccles proposed that exocytosis (the discharge of neurotransmitters across synapses) could be influenced by quantum-scale events. In his review (Wilson 1995) of this book, biologist David Wilson again raised the incompatibility argument:
It is doubtful that the trigger mechanism for exocytosis requires so little in the way of physical movement, but there is an even greater difficulty with this model. There is no way it can avoid violating the requirement of randomness in such quantum mechanical events. Human intentions are quite nonrandom, and even a single intentional movement, say of an arm and hand to lift an object, would require the firing of a number of cortical neurons in a strictly nonrandom way. It would be like a nonrandom radioactive decay of atoms, a clear violation of physical law. It appears that the Eccles model includes the difficulty that all interactionist dualist models seem to share - violation of physical law where the nonphysical mind is supposed to influence the physical world.
The argument is also found in Chalmers (1996, Ch4).
I do not think that any of these writers has answered my points that (1) the felt strength of reasons, and thus the free choices based on them, could tend to follow quantum mechanical probabilities, (2) the choice of an open but improbable outcome could be due to some mental property which the quantum state vector can take no account of, and (3) the unrepeatability of situations of choice would preclude measurable departure from the quantum mechanical algorithm.
Only Lockwood really addresses these points; and even he deals only with the third of them, arguing that, because my suggestion could involve many unique events which were improbable, there could in aggregate be a measurable departure from the quantum mechanical algorithm.
On that point, I do accept that quantum mechanics gives probabilities for individual events, and that the quantum mechanical statistical algorithm could be violated by the combined (im)probabilities of many individual events of different kinds. But we are not considering just any individual events: we are considering unique events of extreme complexity; and I question whether quantum mechanics, with its holistic approach, even permits the assignment of definitive probabilities to such events. While in reductionist Newtonian physics there can always be a calculation to take account of any feature of a situation that has never occurred before, quantum mechanics says that the properties of parts of a system depend on the properties of the whole as well as vice versa; so there must surely be extreme difficulty, if not impossibility, in assigning definitive probabilities to unique events of the complexity involved in human choices.
And of course there are in addition my two other points. I need not elaborate here the first point - that significant conflict with quantum mechanics is unlikely if, as seems plausible, the felt strength of reasons in a general way tends to reflect quantum mechanical probabilities - except to say that these probabilities would, consistently with what I have just said, be quantum mechanical probabilities determined leaving out of account any unique unprecedented features of the whole situation of each individual choice.
However, there is more I should say about the second point - that there could be a contribution from a mental property - which relates this point to the other two points. It has been suggested by philosopher David Griffin (1997, p255) that the randomness of quantum mechanics may be a manifestation of a spontaneity which is a very general feature of the universe. Thus, spontaneity may be manifested as randomness in systems that are relatively simple, like individual particles of matter or radiation, or aggregates of such particles in which randomness is either preserved or cancelled out. However, in systems which are complex and organised appropriately, this spontaneity could plausibly be manifested as a kind of indeterministic self-organisation; and in systems with the complexity and organisation of certain living organisms, spontaneity could be manifested as volitional causation of the kind I am postulating. On this approach, quantum mechanics could not deal fully with spontaneity in systems having this much complexity and organisation, at least without being modified or supplemented; yet would not be violated in relation to any of the systems to which it has successfully been applied to date.
I believe the incompatibility argument assumes an interpretation of quantum mechanics which has the consequence that the question of choice and free will is begged. In considering the problem of free will, a central question is whether or not there might be such a thing as a choice between genuinely open alternatives, made for non-conclusive reasons. If it is assumed that quantum mechanics means that everything must always change over time randomly within rule-governed probability parameters (i.e. that quantum mechanics compels randomness), then this assumption immediately excludes genuine choice and free will: this is Hume’s mistake again. I suggest that, particularly where the interpretation of quantum mechanics is controversial, a more scientific approach would be to hypothesise the existence of choice and free will, and to see if a conflict with any reasonable interpretation of quantum mechanics could be demonstrated by experiment.
Now the standard interpretation of quantum mechanics postulates that events occur at random within the probability parameters given by the rules of quantum mechanics. Those rules may give such a high probability for some result (say, the decay of between one quarter and three quarters of one billion atoms of a radioactive substance during the half life of the substance in question) that any different result would be considered so improbable as to conflict decisively with the rules of quantum mechanics (even though, on the many-worlds interpretation of quantum mechanics favoured by many physicists and cosmologists - and Michael Lockwood - there must be numerous infinities of minds or points of view which are actually observing such things all the time!). Even if the rules of quantum mechanics indicate an equal probability of two outcomes of a particular experiment, so that no result of an individual experiment could conceivably conflict with the rules of quantum mechanics, conflict with quantum mechanics could emerge from many repetitions of this experiment. Thus, suppose one hundred such experiments are conducted, and the numbers of each result noted, and that this is repeated ten thousand times: quantum mechanics would predict that the results would be grouped in a bell-shaped curve centred on the 50/50 result. A result very different from this would be possible, but the probability of such a result could be so small that it would be considered as conflicting decisively with quantum mechanics.
Could the hypothesis of free will be shown thus to conflict with the rules of quantum mechanics? Having regard to my three points, the possibility of a demonstrated conflict between free will and quantum mechanics seems remote indeed. It should also be kept in mind that a quite improbable event may be the result of a number of probable events; so that if, for example, an outcome required a combination of four events, each having an independent probability of 0.7, then the probability of the outcome would be only about 0.24. So if a free choice of a particular result required a combination of possible events, then a quite improbable choice could result from a combination of quite probable events, a combination which was itself no less probable than any other single combination of such events and/or their alternatives.
Or, to take another approach, suppose that psychological experiments gave results which suggested consistent statistics concerning ‘choices’ made by persons in particular circumstances, but randomness within those statistics. This could not be taken as suggesting that each individual case was not a genuine choice of the person concerned. Although the psychological testing could do no more than give statistics, it could well be that in each case the person chose between genuinely open alternatives, with the choice resolving the non-conclusive reasons supporting each alternative - and that the choice in each case could, consistently with all applicable universal laws of nature, have been different.
A complaint could be made that I have dodged the issue of whether (1) the quantum mechanical probabilities are correct and in the long run a series of choices will always fit the probabilities perfectly, or (2) in some cases the quantum mechanical probabilities are overridden by choices and thus made incorrect. I have indeed not answered this question, because I think it is at this stage the wrong question, for reasons I have given. These reasons show that the real question at present is whether there can be plausible account of the operation of choice and free will which is compatible with the broad picture of physical causation suggested by quantum mechanics; and the task of doing this is far less daunting than that of reconciling free will with the determinism of classical physics. It is certainly not ruled out by the argument that quantum mechanical randomness is inconsistent with choice.
7 Chess, Life, and Superlife
The general shape of an account of choice which is compatible with physical causation as suggested by quantum mechanics can be approached by considering analogies provided by games.
One early use of that kind of analogy was by Gilbert Ryle in The Concept of Mind: he suggested (Ryle 1949, p77) that we should not be concerned about what he called ‘the bogy of mechanism’; and he likened the laws of nature to the rules of the game of chess, pointing out the scope which is left in that game for the exercise of intelligence and choice. However, contrary to the main thesis of Ryle’s book, that analogy in fact suggests a kind of dualism. If one took the constituents of the physical world to be like a chessboard and chess pieces, the laws of nature to be like the rules of chess, and the development over time of the world to be like the progress of a game of chess, then one would have to postulate something over and above the physical world and the laws of nature - something making choices between alternatives left open by the laws of nature - if one were to explain what goes on in this world; just as one has to postulate players making choices to explain what goes on in a game of chess.
Suppose that some human scientists, to whom the game of chess was unknown, somehow gained access to a universe, distinct from their own, which so far as they could ascertain consisted entirely of what we would understand be myriad games of chess played competently (with reasonable skill, and no illegal moves) to a finish (no resignations or agreed draws). The space of that universe would be vast numbers of chessboards, the fundamental particles would be various kinds of chess pieces (in two colours); there would be successive discrete time-like steps, and also discrete whole processes corresponding to whole chess games.
The scientists would work out rules governing the movement of the various particles, and rules determining when each process came to an end (and the board in question returned to what they would recognise as an initial configuration). They would work out (1) what moves were possible for particles that we know as pawns, bishops, knights, etc.; (2) that, from the beginning to the end of a process, one particle moves in each time step, and different coloured particles move in successive time steps; (3) how particles can be caused to disappear (can be ‘captured’); (4) that the particle we know as a king (of either colour), alone of all particles, cannot move to any position where it could be captured by a particle of the other colour; and (5) that a process ends when a king can be captured in the next move by a particle of the other colour and cannot itself move (‘checkmate’), or where none of the particles of the colour whose turn it is to move can move (‘stalemate’), or where the same configuration of particles occurs three times (‘draw by repetition’).
They would also detect some further regularities. For example, where it was possible to end a process by ‘checkmate’ in one time step, this would happen in almost all cases; and where it was possible for this to happen in three time steps, irrespective of what happens in the second of those time steps, it would happen in a great but lesser majority of cases. Where it was possible for a process to end in stalemate by one move of a particle of a certain colour, this would happen more often when there were fewer particles of that colour than when there were more. And so on. The scientists could work out quite extensive and elaborate statistics on such matters.
It may be that the scientists would come up with the hypothesis that there were two purposive systems operating in each process, one associated with each colour of the particles; with each system moving its particles with the purpose that the process end in ‘checkmate’ of the king of the other colour, or at worst, in ‘stalemate’ or ‘draw by repetition’. That hypothesis would have the disadvantage that there would need to be much more to the universe in question than the space, the particles, the time-like steps, the mandatory rules, and the statistics: it would require that there also be whatever was necessary to constitute systems having purposes, and having the capabilities necessary to pursue those purposes with some effectiveness. However, the hypothesis would not be excluded by the fact that all observations were substantially in accordance with the statistics which the scientists had worked out; and it would have the advantage that it would make sense of the statistics, which would otherwise be brute facts with no rhyme or reason. If the scientists happened to become aware of the game of chess in their own universe, I think it is likely that they would prefer this ‘dualistic’ hypothesis. (Of course, it could turn out that the purposive systems in fact operated wholly in accordance with deterministic rules other than those of the rules of chess, as do chess-playing computers in our world; but that is another matter - in a chess universe, such systems would still involve a dualism.)
Now suppose that these scientists gain access to a second universe, which so far as they could ascertain consists entirely of what we would understand to be the Game of Life, played on a vast scale. This game was devised in about 1970 by John Conway, a Cambridge mathematician. Its rules can be stated shortly:
Life occurs on a virtual [and potentially infinite] checkerboard. The squares are called cells. They are in one of two states: alive or dead. Each cell has eight possible neighbours, the cells of which touch its sides or corners.
If a cell on the checkerboard is alive, it will survive in the next time step (or generation) if there are either two or three neighbours also alive. It will die of overcrowding if there are more than three live neighbours, and it will die of exposure if there are fewer than two.
If a cell on the checkerboard is dead, it will remain dead unless exactly three of its eight neighbours are alive. In that case, the cell will be ‘born’ in the next generation. (Levy 1993, p52)
It is clear that, once an initial configuration of live cells is set up, everything that happens thereafter in this game is entirely determined by these rules. However, given an initial state with sufficient potential, the game unfolds in ways which have some similarities to life in our world. In particular: (1) large-scale events occur, which in many cases unfold in accordance with large-scale rules; (2) minute differences in initial configurations can produce huge differences in outcomes; and (3) accordingly, there is substantial unpredictability as to what will happen when previously unknown large-scale configurations arise.
The space of the Life universe to which our scientists gain access would be a limitless checkerboard, the fundamental particles would be the two possible states of each square or cell (‘alive’ and ‘dead’), and there would be successive discrete time-like steps. The scientists would no doubt work out Conway’s two rules: if a cell is alive in one time-step, it will be alive in the next time-step if and only if 2 or 3 of the 8 adjoining cells are alive; and if a cell is dead in one time-step, it will be alive in the next time-step if and only if 3 of the 8 adjoining cells are alive. They would hypothesise that everything that happens in this universe is determined by those two rules; and that hypothesis would not be falsified. They would observe interesting larger-scale patterns, and no doubt would work out rules relating to the development of these patterns; but these rules could only indicate the same developments as the two basic rules indicate. The scientists would be unlikely to hypothesise that any purposive systems were operating in this universe.
What about our own universe? Is it more like the chess universe, or the Life universe?
It seems to me fair to say that, after Descartes, a common view among educated Westerners would have been that the physical universe is like the Game of Life everywhere except in parts of the human brain, where (perhaps in the vicinity of the pineal gland) it is like chess. By the end of the nineteenth century, advances in the physical sciences, coupled with Darwin’s theory of evolution, had given rise to a strongly competing view among educated Westerners that there is nothing about the human brain to suggest that different rules apply there; and that our universe is like the Game of Life everywhere.
Twentieth-century science has told us a lot more about the details of the laws of nature governing the development of our universe. Obviously, the laws are vastly more complicated than the rules of the Game of Life; but that in itself does not suggest any efficacious purposive systems. In addition, however, it appears that the laws of nature are partly indeterministic - so that they allow for genuine alternatives in the way our universe develops over time. These alternatives, however, seem to occur randomly within statistical parameters; and they generally tend to cancel out at scales much above that of atoms and molecules, so that at those larger scales, we generally seem to have rules which are as deterministic as those of the Game of Life.
One way of putting the issues which I see as dividing the approach to human purpose of people like myself, who contemplate a substantive role for purposive choice and free will, from that of the majority of scientists, who don’t, is this: could the indeterminism, which appears to exist at the atomic level, (1) give rise to genuinely open alternatives at the level of human choice and action (like the alternatives open to a chess player); and (2) involve the actual exercise of choice and purpose in selection between those alternatives?
Let us now suppose our scientists gain access to a third universe. In this universe, there appears to be a vast and complex game - let us call it Superlife - with regularities which suggest some deterministic and some statistical rules. There also appear in this universe to be many broadly integrated and continuous systems of vast numbers of the particles of this universe, which the scientists call ‘agents’. Each of the states of these systems is unique, being different from all its own earlier states and from all states of all other systems; but contains traces of its own earlier states, which the scientists call ‘memories’. These systems appear to develop over time generally in accordance with the rules of the universe (deterministic and statistical). But unlike the Life universe, the rules of this universe cannot be shown to exclude genuine alternatives in the development of these systems, at a macroscopic scale which the scientists can readily observe; and developments which actually occur tend to suggest that each of the agents is a purposive system, just as was suggested in the case of the systems associated with the different colours in each process in the chess universe.
The scientists come up with two rival hypotheses about this third universe. One is that the development over time of this universe is entirely governed by deterministic and/or statistical rules (and is purely random within the probability parameters indicated by the statistical rules): insofar as purpose appears to be displayed by the systems they call agents, this does not involve genuine choices between available alternatives, and is explained entirely by the history of how these systems were produced, over eons of time, through countless generations of earlier systems which emerged and dissipated (‘evolution’). The other hypothesis is that, while the development over time of the universe does conform to rules, there are leeways left by these rules, and within those leeways the systems called agents really do pursue purposes and really do make choices, between alternatives which the rules really do leave open to them: each choice between such alternatives is a unique efficacious occurrence, determined not by any rules but by the system itself in its then unique unprecedented state, with only the alternatives and tendencies being determined by the rules.
The former hypothesis has the advantage of simplicity: there is no need to postulate anything beyond the particles, space, time, and the rules (deterministic and statistical). The latter does require the additional postulate of purpose or choice; but in this universe, unlike the chess universe, there is no necessity to postulate that there is anything more to this universe than the game. Conceivably, it could just be a fact about Superlife that these systems, by virtue of their own properties coupled with the rules and the existence of alternatives, can detect the existence of alternatives and non-conclusive reasons supporting these alternatives, and can make a choice between the alternatives on the basis of those reasons.
Assuming that the scientists can experiment with this Superlife universe, they could try to refute one or other of these hypotheses - but it could be difficult. On the one hand, the scientists could try to show that what appear to be genuine macroscopic alternatives are not really open, or are not relevant to the apparent purposes of the systems. On the other hand, they could try to show that the systems appear to have and give effect to purposes in ways that can’t be fully explained by rules and randomness and evolution.
But suppose now that, in the absence of any conclusive or near-conclusive experiment, the scientists notice that the Superlife universe is indistinguishable from their own universe, that the apparently purposive systems are indistinguishable from human beings in their own universe, and that these systems in fact report having experiences and purposes which seem to the scientists to be similar to those which the scientists themselves have. To me, that would make it reasonable for the scientists to treat very seriously indeed the hypothesis that there are real choices and purposes in the Superlife universe - as indeed in their own.
So: I have discussed three types of universe - a chess universe, a Life universe, and a Superlife universe. The chess universe involves outright dualism, with purpose provided entirely from outside the particles and rules of the physical universe. The Life universe is a monistic physicalist universe, in which everything that happens is just a working out of the behaviour of the particles as required by the rules of the universe. The Superlife universe, our own, appears to have within it both rules and purpose: the mainstream scientific view at present would have it that it is just a more elaborate version of a Life-type universe, with the apparent purposeful conduct being simply the working out of the rules, with or without randomness; whereas I would argue that a very plausible alternative is a strong dual aspect view, according to which genuine choice coexists with statistics within a unitary universe, and within certain unitary systems of that universe - with the statistics being apparent to an objective, third-person, physical viewpoint; and choice being apparent to a subjective, first-person, mental viewpoint, and involving real selection between alternatives which the physical viewpoint can only treat as statistical probabilities. In such a universe, there could be compatibility between (apparent) randomness and choice.
The widely-held view that determinism is compatible with free will, maybe even necessary for it, has been appealing just because the influence of Hume’s mistake has hindered the articulation of a viable alternative. I believe this paper shows there is a viable alternative to compatibilism, one that makes perfectly good sense of a notion of free will and responsibility which is incompatible with determinism.
And although it goes beyond the scope of this paper to argue at length for the probability, as distinct from the possibility, of free will of that kind, I suggest that it fits in very well not only with common-sense ideas of responsibility, but also with a theory of plausible reasoning as involving, not just algorithms, but also the ability of conscious subjects to make holistic judgements that (fallibly) resolve non-conclusive reasons. It is that ability which I say is not available to zombies, or to computers as presently understood. And this in turn can explain the advantage of consciousness, which has promoted its development in evolution and the mechanisms that ensure we use it in dealing with novel situations.
In a recent article, Galen Strawson (1998) suggested that everything to be said about the metaphysics of free will had been said before, and that there could not possibly be any answer to his dilemma of responsibility. Well, the contents of this paper are not entirely new: some of the ideas have their origin in Nozick (1981), and others are just an elaboration and systematisation of common-sense ideas. But the totality is I believe original and new; it does answer Strawson’s dilemma; and certainly it is not taken into account in Strawson’s writing.
1 I have previously used the term ‘agent causation’; but because of certain connotations of that term, I now think it best to use a different term.
2 Since writing this, I have come across two somewhat similar arguments against Strawson: see Abelson 1988, pp182-4; O’Connor 1995, pp188-9.
3 In this article, I argue in some detail that judicial fact-finding cannot be explained in terms of application of rules, such as Bayes’ theorem.
Abelson, R. (1988), Lawless Minds (Philadelphia: Temple University Press).
Anscombe, G. (1971), ‘Causality and determinism’, in (1981) Metaphysics and the Philosophy of Mind (Oxford: Oxford University Press).
Ayer, A. J. (1956), ‘Freedom and necessity’, in Philosophical Essays (London: Macmillan).
Barrett, J. A. (1994), review of Hodgson (1991), Philosophical Review 103, 350-2.
Buser, P. A. and Rougel-Buser, A. (eds) (1978), Cerebral Correlates of Conscious Experience (Amsterdam: North-Holland).
Campbell, C. A. (1951), ‘Is “freewill” a pseudo problem?’, Mind 60, 441-65.
Chalmers, D. (1996), The Conscious Mind: Toward a Theory of Consciousness (New York: Oxford University Press).
Dennett, D. (1984), Elbow Room (Oxford: Oxford University Press).
Eccles, J. C. (1994), How the Self Controls its Brain (Berlin: Springer).
Eddington, A. (1929), The Nature of the Physical World (London: Dent).
Flew, A. and Vesey, G. (1987), Agency and Necessity (Oxford: Blackwell).
Frankfurt, H. (1971), ‘Freedom of the will and the concept of a person’, Journal of Philosophy 68, 5-20.
Gomes, A. O. (1978), ‘The brain-consciousness problem in contemporary scientific research’, in Buser and Rougel-Buser (1978).
Griffin, D. (1997), ‘Panexperientialist physicalism and the mind-body problem’, Journal of Consciousness Studies 4, 248-68.
Hameroff, S., Scott, A., and Kaszniak, A. (eds) (1998), Toward a Science of Consciousness II (Cambridge MA: MIT).
Hobart, R. E. (1934), ‘Free will as involving determinism and inconceivable without it’, Mind 43, 1-27.
Hodgson, D. (1991), The Mind Matters (Oxford: Oxford University Press).
Hodgson, D. (1994), ‘Neuroscience and folk-psychology - an overview’, Journal of Consciousness Studies 1, 205-16.
Hodgson, D. (1995), ‘Probability: the logic of the law - a response’, Oxford Journal of Legal Studies 14, 51-68.
Hodgson, D. (1996, 1998), ‘Folk psychology, science, and the criminal law’, paper delivered April 1996 at Science and Consciousness Conference, Tucson, Arizona; in Hameroff, Scott, and Kaszniak (1998).
Hume, D. (1748), A Enquiry Concerning Human Understanding.
Kane, R. (1985), Free Will and Values (New York: SUNY).
Kane, R. (1996), The Significance of Free Will (New York: Oxford University Press).
Levy, S. (1995), Artificial Life (Harmondsworth: Penguin).
Lockwood, M. (1991), review of Hodgson (1991), Times Literary Supplement, 15 November, 31.
McCall, S. (1994), A Model of the Universe (Oxford: Oxford University Press).
Nagel, T. (1986), The View From Nowhere (New York: Oxford University Press).
Nozick, R. (1981), Philosophical Explanations (Oxford: Oxford University Press).
O’Connor, T. (1995), ‘Agent causation’, in O’Connor (ed) Agents, Causes, and Events (New York: Oxford University Press).
Ryle, G. (1949), The Concept of Mind (London: Hutchinson).
Smart, J. J. C. (1961), ‘Free-will, praise and blame’, Mind 70, 483-94.
Strawson, G. (1986), Freedom and Belief (Oxford: Oxford University Press).
Strawson, G. (1998), ‘Luck swallows everything’, Times Literary Supplement, 26 June, 8-10.
Taylor, C. (1976), ‘Responsibility for self’, in Rorty, A. O. (ed.) The Identities of Persons (Univ. of California Press).
Williams, B. (1995), Making Sense of Humanity (Cambridge: Cambridge University Press).
Wilson, D.L. (1995), review of Eccles (1994), American Scientist 83, May/June, 269.