Hodgson’s Black Box
In seeking to establish the existence of what he calls a ‘plain person’s free will’, David Hodgson adduces 8 conditions, the joint satisfaction of which would, he claims, result in our having such free will (proposition 9 asserts this conclusion). The plain persons’ conception of free will, Hodgson says, is the libertarian conception, in which it is incompatible with determinism. Although what ordinary people actually believe about free will is an empirical matter in need of research, it’s likely that many people (but not all) have at least a vague notion that to be free, we can’t be completely determined in our acts, that the self has to be some sort of first cause to be ‘ultimately’ responsible and therefore justifiably deserving of praise and blame. If asked how individuals get to be first causes, most honest non-philosophers would have to concede that to them it’s a mystery.
Although our plain person might subscribe to such a notion of libertarian free will, Hodgson’s account of it is not, as he describes it, ‘a simple and straightforward outline’ that supplies easily understood justifications for libertarian freedom and so might capture the popular imagination. Rather, it’s a highly speculative and scientifically radical account, involving contentious claims about quantum mechanics, evolution, consciousness, rationality, and the self, as well as conjectures about new types of natural laws (Hodgson’s ‘G-laws’). If we buy his arguments, Hodgson claims we’ll be led to the conclusion that in addition to the determinism and randomness that science finds in nature, there also exists a heretofore undiscovered ‘volitional causation’, something special to human conscious agents. To establish this conclusion would amount to a scientific revolution of the first order. But because Hodgson’s story is so speculative and his claims so empirically tenuous, it’s not surprising that what he gives us, finally, is not a preview of a scientifically well-motivated paradigm shift, but an obscure black box of free will – a still-mysterious capacity of the self to somehow clinch a choice without being determined to clinch it in a particular direction.
We must evaluate Hodgson’s account in the light of
competing scientific accounts of human choice-making behavior, using criteria of
simplicity, completeness, empirical support, and agreement with the rest of
science. How well does Hodgson’s account explain choice-making compared
with deterministic accounts? Although Hodgson may think that indeterministic
free will is necessary to justify socially necessary attributions of credit,
blame, reward and punishment (I don’t think it is) it’s quite possible that
nothing in the natural world corresponds to his ‘plain person’s’ conception of
free will. I will first comment on Hodgson’s claims about quantum mechanics,
consciousness, rationality, and evolution, and then analyze in more detail his
Proposition 7 concerning the subject and its capacity to select choices.
In order to give the selecting subject two
alternatives, neither of which is uniquely necessitated, Hodgson proposes
quantum mechanics (QM) as the basis for the requisite indeterminism. But
immediately he is forced to acknowledge that QM might supply only ‘mere
randomness’, and that in any case it’s not clear how QM indeterminacy relates,
if at all, to the macro level of human behavior. The problem here (and
elsewhere) is that in order to make the self clearly, transparently responsible
for a choice, Hodgson has to delineate some mechanism or process by which
it determines the choice. But of course QM resists any mechanistic, causally
transparent account precisely because it’s essentially probabilistic in its
predictions of behavior in the micro realm (putting aside the difficulties of
macro level effects). Later (p. 6), Hodgson says that the subject’s selections
‘are likely to approximate to QM statistics’, but then says it’s an ‘open
question’ how QM applies to conscious systems, and finally he declares that the
causal role of experiences in determining choices ‘can’t be fully accounted for
by any system of physical laws of general application, even those of
QM’ (emphasis added). Ultimately, the role of QM, ‘perhaps together with
chaos theory’, in supplying indeterministic alternatives is left obscure on
Hodgson’s account, and so adds nothing to a scientific explanation of how people
choose. This is a recurring difficulty in Hodgson’s paper: he proposes that
certain generally accepted processes or theories (QM, rationality, evolution)
contribute to an understanding of choice-making behavior, but in each case stops
short of supplying a definitive causal story about that contribution, leaving
the actual choice an explanatory black box. But of course such obscurity is
what’s necessary for ‘volitional causation’ – that which is neither random nor
determined – to survive as a (pseudo) explanation of choice. The details of
such causation are necessarily mysterious, for any transparent account of
them would likely reveal that the choosing subject is, contrary to Hodgson’s
thesis, no exception to standard macro-level causality.
Hodgson claims that the choosing self is separate from
contents of consciousness, something ‘active’ that is the ‘bearer or experiencer
of the contents of consciousness’. But recent theories of phenomenal
experience, some of them well-elaborated and consistent with neuroscientific
findings at many levels, suggest that the conscious self is actually one of the
contents of consciousness (Damasio, 1999, 2000; Metzinger, 2000a, 2000b)
– a stable phenomenological construction within experience, not an
irreducible agent/force that ‘bears’ or ‘experiences’ experience. The conscious
subject, by virtue of being a construction, is not an entity or agent which
could exert control over choices above and beyond those control mechanisms
already instantiated by neural processes, whether conscious or unconscious, some
of which constitute the sense of self. Despite the fact that Hodgson disavows a
distinct mental substance or soul, he suggests an underlying dualism in claiming
that the causal powers of the conscious self aren’t simply those of the physical
processes that correlate with consciousness. He needs this residual dualism, of
course, in order to make the conscious self rule over the deterministic
mechanisms of the brain. But as neuroscientific understanding of the sense of
self grows, it’s becoming more and more plausible that a certain sub-set of
neural processes are the self as modeled by neural processes, e.g., those
that instantiate internal representations of homeostatic functions that preserve
bodily integrity (Damasio, 1999, 2000).
On page 6, Hodgson argues that since pain and other phenomenal states play a causal role in mediating behavior, and since such subjective feelings can’t just be a matter of algorithmic computation performed by a physical substrate, there’s something non-computational and non-algorithmic (namely, consciousness) playing a causal role in selecting choices. But on increasingly well-documented functionalist-representationalist accounts of phenomenal consciousness, the second premise in this argument is questionable. Hodgson overlooks the possibility that (e.g.) pain has a particular subjective feel because of the functional and representational roles of the neural states that constitute it – that function and feel are perhaps identical, but differently ‘presented’ depending on whether one occupies a first versus a third-person perspective regarding the token phenomenal pain in question (Metzinger, 2000a, 2003). So consciousness might well be computational at the right level of description.
Hodgson says (p. 5) that we should accept the
first-person deliverances of consciousness as good, face-value evidence for his
claims about the existence of non-algorithmic, non-rule-governed volitional
causation: ‘All this is confirmed by the powerful and ineradicable feeling we
have that we are consciously making choices and making things happen by doing
them.’ But of course the feeling that something is the case can’t confirm that
it is the case, and good scientific explanations of behavior are
notorious for undermining commonsense folk-psychological convictions based in
untutored experience. In this instance, Daniel Wegner (2002) cautions us that
we should most definitely not accept the feeling that ‘we are consciously
making choices’ as evidence of anything except that we have such a feeling. The
actual status of the ‘we’, the ‘making choices’, and the feeling of
choice-making are matters for science, not introspection, to determine.
On page 3, Hodgson says that consciously experienced reasons are essential to choice, but that they only influence, not determine, our selections. One wants to know then, what does determine the choice? Hodgson’s thesis is that the self determines the choice via volitional causation, but of course such causation requires that the self can’t be fully determined itself or determined in its choice-making, which is why on Hodgson’s theory even very powerful, persuasive reasons must be ‘non-conclusive’ and undetermining. But, if reasons are indeed non-conclusive, then it follows there is literally no reason why the self reaches the choice it does – the choice is made, finally, on some other, non-reasons basis. And this basis is left obscure. As with quantum mechanics (section 1 above) and evolution (section 4 below), the role of rationality in explaining choices reaches a certain, ‘non-conclusive’ limit, and then hits the wall of the black box of freely willed choice, which remains frustratingly unexplained.
Despite conceding we are rational animals, Hodgson maintains there are no ‘clinchers’ or final determiners to be found in the operations of rationality. But of course there often are obvious clinchers in the various reasons that we normally cite in explaining morally responsible choices, and these stand as counter-examples to Hodgson’s thesis. Virtuous, law-abiding citizens often choose the way they do because they find the reasons for good behavior – e.g., maintaining their reputation among their peers, the ‘benselfish’ (Dennett, 2003, pp. 193-7) consequences of altruistic acts – compelling, and to be thus compelled is not to render good behavior any less admirable or the agent unworthy of praise. Nor would finding out that the acts of rational evil-doers were clinched by their nefarious, abhorrent reasons make such acts less wrong and the agent less accountable to moral norms and the law. Rationality is the criterion of responsible agenthood precisely because it permits the agent to anticipate the consequences of an act, and our practices of holding people responsible engage such rationality by making certain consequences reliably contingent on behavior (drive while drunk, lose your license). If Hodgson’s black box of free will were interposed between the motivated, reasons-responsive agent and the social contingencies of our responsibility practices, such practices would have no power to shape behavior, since after all, the agent would be ultimately free to choose a course of action that ignored whatever contingencies were in place (a point made by Pinker, 2002, p. 177). But it’s manifestly the case that such practices do have tremendous power, that rational agents do indeed respond rationally to social norms and laws. That is, they respond in a reasons-responsive manner, a manner that is best explained in terms of a high-level, agent-centered, but nevertheless causal story in which the self is not free to ignore, on some obscure, extra-rational basis, the demands of anticipated contingencies.
On page 4, Hodgson says that ‘What is often overlooked is that, apart from rules of reasoning such as those of mathematics, logic, and probability theory, there are no known rules (that is, strict rules as distinct from non-conclusive heuristics) governing good plausible reasoning.’ But these rules of reasoning are powerful rules indeed; they are at the core of formal argument, of deducing the implications of evidence, and of quantitative theory construction. Rational agents ignore them at peril of becoming irrational. Of course there is clearly more to embodied, real-world human rationality than formal entailment from axioms and premises, for instance those problem-solving intuitions that spontaneously arise from long-standing experience in a skill or knowledge domain. But it’s not necessarily the case that what’s unformalizable at the phenomenal, subject level is for that reason non-algorithmic at the machine or design level (Dennett, 1989).
Hodgson argues that ‘If we
cannot rely on our plausible reasoning as the conscious non-algorithmic process
that we instinctively take it to be, then any confidence that we could have in
it would have to depend on the circumstance that it comprises computation-like
processes whose reliability is assured by the evolutionary tests they have
passed; yet any belief in this circumstance and accordingly any justified
confidence would itself depend on extensive plausible reasoning, giving rise to
a vicious circle’. But this not-so-vicious circle can be broken simply by
dropping the not-so-instinctive assumption that our reliance on reason is only
warranted if it’s non-algorithmic. Confidence in our reasoning powers doesn’t
depend, as Hodgson seems to think, on some sort of non-mechanistic,
non-computable causal disconnection from the world’s influence upon us. Rather,
it’s warranted because, by and large, we get the facts about the world right via
normal perceptual and cognitive processes that reliably track the world via
consistent causal channels, and we reason successfully via other processes that
reliably distinguish perceived reality from our simulated internal worlds of
memory, imagination, and anticipation. Any choice-making ability that allowed
indeterministic slack between input and output would be less efficacious,
less rational, than a system that accurately reflected, via deterministic
processes, the relevant contingencies of the environment and its own
self-interested agenda. Because we are entirely physical, embodied creatures,
there is good reason to think that rationality and cognition are indeed
instantiated by deterministic, causal processes, albeit highly ramified and
perhaps only fully understood at the representational, not physical, level. But
again, having the causal story go through transparently, from world to
rational agent and back out again in cognitively adept behavior, would reveal
the agent as only the proximate, not ultimate cause of behavior, and therefore
not ultimately deserving of credit and blame. So rationality, for Hodgson, must
remain causally obscure within the black box, not transparently available to
science, which is why his account won’t win many converts among the empirically
Hodgson wants to keep his theory naturalistic, and to this end cites the evolutionary origins of our choice-making powers. But the problem he faces is that evolutionary explanations are causal explanations involving various mechanisms of selection that operate via discrete, more or less specifiable systems (genes, organisms, environments), and it’s unclear how such a causal, mechanism-employing, rule-governed process could produce something a-causal, non-mechanistic, and non-rule governed (volitional causation). He says:
‘Our rationality is well adapted to dealing with problems remote from the evolutionary tests that faced our evolutionary ancestors, and this makes it unlikely that it is no more than a matter of useful algorithmic processes selected through those tests’.
How then, one wonders, did our purportedly non-algorithmic rationality get selected by evolution? On this question, Hodgson is notably silent. In fact, it’s questionable whether the problems we deal with are really so remote from those faced by our evolutionary ancestors, and therefore questionable to suppose there must be any radical discontinuity in our cognitive powers from those possessed by our forbearers. If we drop the requirement that consciousness and rationality must be non-algorithmic and non-rule-governed (see sections 2 and 3 above), then the problem of their natural origination becomes tractable. And if we agree (as Hodgson does) that evolution – a blind, deterministic, mechanistic, purely physical process – did indeed produce our choice-making capacities, this lends considerable support to the idea that consciousness and rationality themselves are nothing other than a complex elaboration of deterministic, mechanistic, fully physical processes, even though it may not feel that they are.
It’s also not clear how being non-ruled governed
creatures would confer any selective advantage. As noted above in section 3 on
rationality, any indeterministic slack between the influence of reasons and
behavioral output would undermine, not increase, the efficacy of goal-directed
action. So although Hodgson wants (understandably) to make his account
consistent with one of biology’s most fundamental and productive theories, the
central elements of his account – non-algorithmic consciousness and
reason – don’t seem to be implied or explained by that theory. This undercuts
the scientific plausibility of his thesis, especially when we compare it to
mainstream neuroscientific, cognitivist and evolutionary-psychological
explanations of rationality and consciousness that accept good old-fashioned
causal determinism as a working assumption, and that fill in the evolutionary
story behind human behavior (e.g., Dehaene, 2002; Dennett, 2003; Pinker, 1999;
Ridley, 2002). To compete against these explanations, Hodgson must explain
precisely how evolution installed the black box of free will, and what
indispensable function it serves, beyond supplying an ultimately responsible,
Of crucial importance to the plausibility of Hodgson’s theory of choice is the delineation of the subject that does the choosing and the process by which the choice is made, discussed in Proposition 7. On both counts, as on other topics, Hodgson is frustratingly vague. Moreover, he vacillates about what’s crucial in explaining choice: a person’s specific character versus a general capacity, shared by all persons, to select choices. He says:
‘The totality of “the way a person is”, prior to the selection being made, is inconclusive as between the available alternatives: it gives rise to reasons and tendencies to act in one or other of the ways that are open, but does not pre-determine the outcome.’
So we ask, what does determine the outcome? Presumably something other than the totality of the way the person is, namely the ‘capacity to select’. Regarding this Hodgson says:
‘In relation to this capacity, each person is entirely the same, unaffected by differences in pre-choice states, whether due to genes, environment, prior selections, or all three; and in relation to its exercise, to the extent that each person can notionally be considered apart from differences affecting alternatives, reasons and tendencies, each person is entirely the same’ (my emphasis).
So Hodgson claims that both the amount and exercise of the capacity to select is identical across persons, unaffected by any differences in pre-choice states. But if this capacity is really identical across all persons, then the only things that can explain differences in behavior are differences in character. Hodgson says, in fact, that
‘Different persons have different characters, and act differently because of these different characters. However, I am suggesting that this is because of the differences that pre-choice states make to alternatives, reasons, and tendencies, not because of any differences in the persons’ capacity to select’ (my emphasis).
So the capacity to select, since it’s exactly the same for all persons, really adds nothing to the explanation of how individuals make their choices, despite the fact that Hodgson says it alone does the selecting. It seems we need only appeal to the way the person is, which is what standard deterministic explanations of choices do. But for Hodgson all this useful, explanatory, predictive information about character, alternatives, motives, desires, etc. is somehow still ‘inconclusive’ – there’s still the ‘clincher’ needed, that provided by the capacity to select:
‘…our particular characteristics do not otherwise affect the way we exercise our capacity to select. We do this [that is, we select] by choosing which alternative occurs, thus providing the clincher; and there is an element (by which I do not mean a distinct or severable element) of this process that is entirely up to us, unaffected by any differences between different persons. And as asserted by the fifth proposition, this does not mean that the selection, or any part of it, is random or otherwise not rational’ (original emphasis).
According to this passage, which seems to contradict his claim that ‘[persons] act differently because of these different characters’, our selection of a choice is not determined by our character, rather our selection is influenced, but not decided by, the alternatives that our character brought into being prior to the selection. What finally decides or clinches the choice, is ‘entirely up to us’, but for Hodgson the ‘us’ doing the selecting is something other than our character, since it’s something ‘unaffected by any differences between different persons’. So, although Hodgson claims I am wrong to say (in Clark, 1999, p. 286) that he ‘makes the choosing subject an abstract entity devoid of character and motives’, it seems to me that I’m correct in this assessment, given that Hodgson’s clincher has nothing to do with character, but instead derives from a capacity possessed and exercised identically across all agents.
He counters my assessment by saying: ‘The subject is the unique totality of all its properties, and it is precisely because this unique totality together with particular experiences enters into the causal process that outcomes are not predetermined by constituent properties which it may share in varying degrees with other entities and with which general laws can engage.’ But the unique subject that Hodgson cites here is exactly that which he claims plays no role in the final determination of choice, since he says the ‘us’ that provides the clincher is ‘unaffected by differences between different persons’. The unique subject, if it is to be credited with clinching a choice, has to determine the choice, but according to Hodgson, the clincher isn’t determined by the alternatives given by the unique subject, only influenced by them. The clincher is determined by a capacity identically shared and exercised by all subjects. Thus it seems my original description of Hodgson’s chooser, as an ‘abstract entity devoid of character and motives’, is on the mark. What has free will and can be held ultimately responsible, on Hodgson’s theory, has nothing to do with the real, distinguishable people that show up in court.
Note also that Hodgson wants the clincher to be rational, not random; after all, if it were random we couldn’t credit or blame the agent. But rationality necessarily requires the situated subject to take into account its unique circumstances and motives in deciding a course of action. Any aspect of a subject causally disconnected from such motives and circumstances, i.e., something identical across all subjects which is merely influenced but not determined by such specific, situational motives and circumstances, would have, finally, no reason to clinch the choice in a particular direction advantageous to the agent. So again, Hodgson’s abstract capacity for choice adds nothing to our understanding or explanation of how we choose; it’s a disconnected, causally irrelevant cog in the choice-making mechanism, put there only to rescue the ultimately autonomous self. We do, of course, share with other agents general rational capacities for deliberation, anticipation, and recollection, but there’s no reason to think that these capacities are in any way undetermined, instantiated as they are by our physical brains. And these capacities can’t, of course, be Hodgson’s ‘element …of this process that is entirely up to us, unaffected by any differences between different persons’, since people vary considerably in cognitive styles, self-knowledge, foresight, and other traits and abilities that constitute the exercise of rationality.
A plausible scientific account of human behavior and choice-making must be transparent in its connections to other well-established scientific laws, principles, processes, and entities. The prospects for Hodgson’s account achieving such connections are remote, given that it depends on a host of empirically tenuous claims and tentative suggestions concerning quantum mechanics, consciousness, rationality, evolution, and special laws of nature concerning choice and morality which to my knowledge have little precedent in the scientific literature (space limitations preclude my taking up the prospects for his ‘G-laws’).
For Hodgson, there can’t be a ‘conclusive’ causal story explaining choices on the basis of an algorithmic rationality, personality traits, motives, external circumstances, and other determinate factors, for that would be to succumb to a determinism that shows the agent to be part of a larger causal chain, and thus not ‘ultimately’ responsible. To get the buck to stop at the agent, Hodgson must insert within each of us the black box of an undetermined, radically free, yet rational decision-making capacity. But it’s unclear how being undetermined in any respect is compatible with being optimally rational; it’s unclear how human agents are exceptions to ordinary natural causality, even in their highest deliberative capacities; and moreover we don’t need to possess such black boxes – that is, to be ultimate originators – to be held compassionately responsible and accountable. Jurists and legal scholars are, under increasing pressure from the sciences, finally coming to terms with determinism, and Hodgson is bucking the tide in supposing that justice requires libertarian free will (see, for instance, Moore, 1997 and Morse, 1998, 2000). The consequences of accepting that we are indeed fully caused in our choices, although beyond the scope of this reply, may well include humane reforms in the criminal justice system (Clark, 1996; Greene & Cohen, 2004), reforms which I suspect Hodgson would support.
© Thomas W. Clark, 2004
Blackburn, S. (1999), Think (Oxford: Oxford University Press).
Clark, T. (1996), ‘The freedom of Susan Smith’, Humanist, 56 (2).
Clark, T. (1999), ‘Fear of mechanism’, in Libet et al (1999).
Damasio, A. R. (1999), The Feeling of What Happens (New York: Harcourt Brace).
Damasio, A. R., Grabowski, T., Bechara, A., Damasio, H., Ponto, L.L.B., Parvizi, J., Hichwa, R.D. (2000), ‘Distinctive patterns of subcortical and cortical brain activation associated with self-generated emotions and feelings’, Nature Neuroscience, 3 (10), 1049-56.
Dehaene, S. ed. (2002), The Cognitive Neuroscience of Consciousness (Cambridge, MA: MIT Press).
Dennett D. (1989), ‘Murmurs in the cathedral’, review of Roger Penrose's The Emperor's New Mind in Times Literary Supplement, September 29-October 5, 1989, 4513:1055-1056.
Dennett, D. (2003), Freedom Evolves (New York: Viking Press).
Greene, J. & Cohen, J. (2004) “For the Law, Neuroscience Changes Nothing and Everything,” Philosophical Transactions of the Royal Society of London (Science B), Theme issue on “Law and the brain”, 359, pp.1775–1785, online at http://www.csbmb.princeton.edu/~jdgreene/NewGreene-WebPage_files/GreeneCohenPhilTrans-04.pdf
Libet, B., Gleason, C., Wright, W., and Pearl, D. (1983), ‘Time of conscious intention to act in relation to onset of cerebral activities (readiness potential): the unconscious initiation of a freely voluntary act’, Brain, 106, 623-42.
Libet, B., Freeman, A., and Sutherland, K. (eds) (1999), The Volitional Brain (Thorverton: Imprint Academic).
Metzinger, T. (ed) (2000a), Neural Correlates of Consciousness: Empirical and Conceptual Questions (Cambridge, MA: MIT Press).
Metzinger, T. (2000b), ‘The subjectivity of subjective experience: a representationalist analysis of the first-person perspective’, in Metzinger, 2000a, pp. 285-306.
Metzinger, T. (2003), Being No One: The Self-Model Theory of Subjectivity (Cambridge, MA: MIT Press).
Moore, M. (1997), Placing Blame: A general theory of criminal law (Oxford: Clarendon Press).
Morse, S. (1998), ‘Excusing and the new excuse defenses: a legal and conceptual review’, Crime and Justice, 23, pp. 329-406 (University of Chicago Press).
Morse, S. (2000), ‘Deprivation and desert’, in From Social Justice to Criminal Justice: Poverty and the Administration of Criminal Law, Heffernan, W.C. and Kleinig, J., eds., (Oxford: Oxford University Press).
Pinker, S. (1999), How the Mind Works (New York: W.W. Norton & Company).
Pinker, S. (2002), The Blank Slate (New York: Viking Press).
Ridley, M. (2003). Nature Via Nurture: Genes, Experience, and What Makes Us Human (New York: Harper Collins).
Strawson, G. (1998), ‘Luck swallows everything’, Times Literary Supplement, June, 1998, at http://www.naturalism.org/strawson.htm.
Wegner, D. (2002), The Illusion of Conscious Will
(Cambridge, MA: MIT Press).
by David Hodgson
I am grateful to Tom Clark for providing a response to my paper that forcefully expresses a version of the mechanistic viewpoint favoured by many, perhaps most, scientists and philosophers.
A central disagreement I have with him concerns his use of the expression ‘black box’ as applying to my account of free will. It suggests that I am proposing that, in the production of choices, there operates a discrete system, as to which no explanation is or can be given of what happens inside it, whereas a complete explanation can be given of the inputs to this system and of what happens to the outputs of the system. I think this misapprehends my account in two important ways.
First, Clark’s black box terminology attributes to my account a separation, into discrete parts, of one element of choice-making that is mechanistic and another distinct element that is not mechanistic, a separation that my account in fact strongly rejects. I say that choice-making can be considered as an exercise of informal rationality, and thereby understood as a process in which a rational agent makes a choice on the basis of reasons that were, prior to the choice, inconclusive. Alternatively, choice-making can be considered as if it were a mechanistic process, either in terms of physical processes evolving in accordance with laws of nature, or in terms of computation-like algorithms. However, if my position is correct, any mechanistic approach to choice-making can at best explain what alternatives are available and give numerical probabilities, and must treat the occurrence of just one of the alternatives as a matter of chance within those probability parameters. I certainly do not suggest a ‘black box’ that determines which alternative occurs, after mechanistic processes run out. Rather, I say that the mechanistic approach gives an incomplete account of a whole process that can best be understood in terms of a choice made for reasons.
Second, Clark’s approach suggests I am proposing something wholly inexplicable and incapable of being understood, whereas I fact I say we are very familiar with and have a reasonable understanding of informal rationality. Clark himself suggests that our rational processes are ‘perhaps only fully understood at the representational, not physical, level’; and it is clear that, except in so far as our rational processes are algorithmic at the representational level, at that level those processes are indeterministic because premisses do not wholly constrain conclusions. The difference between his position and mine on this point is not that I propose some mysterious black box that takes over where algorithms run out, but rather that I suggest that our informal rationality, which (as Clark and I agree) we understand pretty well at the representational level, is truly explanatory and causally efficacious; whereas Clark seems to say that it merely supervenes on some underlying mechanistic process that we do not fully understand (Clark’s black box?).
Another important disagreement between us arises from Clark’s suggestion that the conscious self is but one of the contents of consciousness. Now, I accept that many of us may be fundamentally mistaken as to the nature of the subject of experience, for example in so far as we take this subject to be distinct from associated brain processes and/or to have continuity and stability and capacity to be active; and that some of our ideas about these matters may be no more than fallible contents of consciousness. But the idea that experiences are had by a subject (rather than being somehow ‘free-floating’) is so deeply presupposed and embedded in our language and in our ways of thinking that to deny this, without proposing a language concerning experiences that does not have this presupposition, is nonsense. In our language, ‘pain’ means ‘pain as experienced by some feeler of pain’ and cannot be understood in any other way; and the same goes for any other references to and descriptions of conscious experiences.
It could be argued that our language is in this respect not adequate to accurately reflect or describe reality; but Clark makes no suggestion as to why this might be so, much less any plausible proposal for a new language that does a better job. In fact the only faintly plausible strategy that has been proposed for talking and thinking about experiences in a way that does not presuppose a subject that has the experiences, is the suggestion that experiences are useful fictions. (This seems to be Dennett’s position.) This avoids the contradiction of asserting there are experiences but denying there is a subject; but does so at the cost of denying that there really are experiences – that is, denying that there really are contents of experiences.
Turning to more specific points, I do not suggest that volitional causation is special to human agents. Rather, I say that there is volitional causation wherever there is consciousness; although I also say that free will, as generally understood, requires rationality of the order of that possessed by human agents.
I do not overlook ‘the possibility that (eg) pain has a particular subjective feel because of the functional and representational roles of the neural states that constitute it’; but I do consider it highly implausible. I believe this possibility is the merest speculation, in the absence of any account of how or why algorithmic computations could involve subjective feelings such as pain, or of what would distinguish algorithmic computations that involved such feelings from algorithmic computations that did not involve such feelings. Further, it is close to being disproved by scientific work on such things as phantom limbs and synaesthesia, where feelings occur in the absence of their usual causal roles.
I do not say, as Clark suggests, that choice is determined by the self and not by the reasons. I say that neither the self nor the reasons (nor indeed both of them together) pre-determine the choice, but that the subject and the reasons together determine the choice (by the subject making the choice for the reasons). It is a self-refuting fallacy to say that, if reasons are non-conclusive, there is no reason why the self reaches the choice it does – self-refuting because to deny the rationality of all reasoning apart from reasoning in which the reasons are conclusive is to deny the rationality of all informal reasoning, and thus to deny the supportability of most of our beliefs. I accept that it is possible that our informal reasoning is supported by processes that are algorithmic/computational, but we certainly do not know that this is so; and in order to have any confidence in our beliefs, we must accept the rationality of our informal reasoning, whether or not it is supported by algorithmic/computational processes. And of course I do not suggest that the rules of formal reasoning are unimportant – only that they do not account for all human rationality.
I do not deny that very often, perhaps most of the time, people find reasons for acting well (or badly) to be compelling, so that action in accordance with those reasons follows almost inevitably. The circumstance that the likelihood of any different action may be near zero (whether considered in terms of a numerical probability derived from physical laws, or some more generally understood likelihood based on a high-level agent-centred account) does not mean these actions are not freely chosen.
Clark claims that any indeterministic slack between the influence of reasons and behavioural output would decrease the efficacy of goal-directed action. This of course presupposes that our informal rationality depends entirely on computational algorithms – which might be the case, but is not certainly so and is contrary to my contentions. In my paper I give a specific reason why ‘indeterministic slack’ may contribute to efficacy, a reason which Clark does not address – namely, that it leaves room for a conscious organism to respond appropriately to particular gestalt experiences, grasped as a whole, and not merely to general features that can engage with natural laws.
Clark says that to compete against mainstream evolutionary explanations, I ‘must explain precisely how evolution installed the black box of free will and what indispensable function it serves’. This is a bit rich, when the mainstream explanations have totally failed to explain what installed subjective consciousness or what function it serves, whereas my account does at least propose a function for consciousness.
Clark suggests I say that the totality of the way a person is does not determine the outcome, whereas in fact I say that it does not pre-determine the outcome: as noted above, I say that the outcome is determined by the subject choosing on the basis of reasons. And in focussing on the capacity to select as being the determinant, Clark is again misconstruing my position: I say as plainly as I can that the capacity to select, as to which we are al the same, is not a severable part of the totality of the way we are; and I certainly do not say that the ‘us’ that provides the clincher is unaffected by differences between different persons. The clincher is provided by the totality choosing on the basis of the reasons.