Hard determinists (such as I)—maintaining that genuine free will is illusory—typically concede that genuine volition exists but claim the sense of free will concerns thoroughgoing agent autonomy: origination. (See T. Honderich, A theory of determinism (1998).) Previous essays defended hard determinism by reducing origination to the illusory sense of free will, consisting of the misperception that experienced deciding causes action. In this essay, I attack the illusion of free will at its home base: volition. It is not even the case, I contend, that our behavior is under voluntary control.
Newcomb’s paradox is incoherent
My starting point is the well-known problem in
philosophical decision theory called Newcomb’s Paradox. I don’t claim to be a
serious student of decision theory; my purpose is to turn the paradox to a
different purpose—different but not entirely novel. Others have argued that the
denial of free will is inherent in the paradox’s assumptions, but I take the
argument a step further by contending that in relevant part the assumptions are
true.
In the form I’ll use, Newcomb’s Paradox concerns a game
organized by a faultless Predictor. The game’s set-up consists of two boxes,
one transparent and the other opaque. Predictor places a thousand dollars in
the transparent box but places an amount in the opaque box that’s contingent on
the prediction. Chooser is to pick either the opaque box or both boxes. When
Predictor knows Chooser will pick the opaque box, it places a million dollars
in it, but if Predictor knows Chooser will pick both boxes, it leaves the
opaque box empty. Chooser’s best outcome would be to pick both boxes when it
contains a million dollars, but the infallible Predictor then will have left
the opaque box empty, since being infallible, it has predicted the choice. But
the Predictor goes exclusively with its prediction and makes no changes after Chooser's actual choice, and following this line of thinking, Chooser should pick both
boxes. If Predictor put the million in the opaque box, then Chooser gets $1,000,000
+ $1,000 = $1,001,000. But since Predictor is always right, it would have
predicted that Chooser would choose both boxes, so Chooser is stuck with only a
thousand.
Both sides have advocates because valid
principles seem to support either decision. Supporting the both-box choice is
the principle that we should choose based on the consequences of our choices,
that is, based on what they cause. The maxim holding that correlation isn’t
necessarily causation expresses this principle. We should choose both boxes because what's done is done. The other principle is that we
can rely on lawful regularities when they—as opposed to relations of mere
correlation—underpin our choice. Since the Predictor always rewards the choice
of the opaque box with $1,000,000 and punishes the choice of both boxes by
awarding only $1,000, we should choose the opaque box. The two valid principles
lead to contradictory solutions because the puzzle itself contains contradictory
assumptions: if you can choose something besides what the Predictor prophesies,
the Predictor isn’t infallible, since Predictor then can be wrong. If you focus
on the assumption that Predictor is infallible, you pick the opaque box. If you
focus on the causal consequences of your ability to choose, you choose both
boxes.
Re-evaluating volition
Commentators haven’t widely recognized the incoherence
of Newcomb’s paradox because (as tacit compatibilists) they don’t see how a
precise prediction derived without considering Chooser’s volition could
be perfect: why can’t Chooser contravene a determinate prediction in willful
pursuit of its self-interest? The problem is evident if Predictor were to announce beforehand that it predicts Chooser will pick the opaque box: what’s to stop Chooser from refuting Predictor
by picking both boxes—netting a $1,000 increase in payoff? Predictor’s infallibility might be said to
stop Chooser, but the question is how could Predictor be infallible (or
how could the universe be suitably deterministic) when Chooser may want to thwart the prediction, if informed. What’s the outcome when Chooser—as seems entirely possible—wants to
pick contrary to prediction? The answer must be that Chooser fails to pick what
it wants. If Predictor forecasts the opaque box and Chooser wills the both-box choice,
Chooser can’t act according to its will. It will have to act as if in a hypnotic
trance when hypnotic suggestion conflicts with volition. (Post-hypnotic
suggestion may be a useful model for hard determinism.)
The weakness of this solution to Newcomb’s paradox is abruptness. Why should choice suddenly become impossible when prediction is complete yet occur when prediction is almost complete? The paradox should force us to reconsider the general proposition that genuine volition exists, and Libet’s findings provide a basis for re-evaluation. If we will our behavior, how do we know we do, given that we find out that we carried out a willful act only after our brains transmit the effective impulse? I’ve been calling that impulse the “decision,” but Chooser frames the decision after the act. Our predecision intentions comport with our actions only because of the formers' abstractness and vagueness, and strict volition is an illusion caused by our flawed perception of mental temporality. The conclusion holding that perfect-prediction’s logical possibility precludes choice is a dispositive claim against compatibilism, which—to deny any contradiction between free will and determinism—analyzes free will as volition. Our acts aren’t truly voluntary.
Changed on July 2, 2012, 11 a.m.: Clarified that Predictor announces prediction in second-to-last paragraph. (See discussion with Joseph.)
I elaborate on the logic of the second-to-last paragraph in 10.1.2 Reflexive prediction, determinism, and the impossibility of free will
The weakness of this solution to Newcomb’s paradox is abruptness. Why should choice suddenly become impossible when prediction is complete yet occur when prediction is almost complete? The paradox should force us to reconsider the general proposition that genuine volition exists, and Libet’s findings provide a basis for re-evaluation. If we will our behavior, how do we know we do, given that we find out that we carried out a willful act only after our brains transmit the effective impulse? I’ve been calling that impulse the “decision,” but Chooser frames the decision after the act. Our predecision intentions comport with our actions only because of the formers' abstractness and vagueness, and strict volition is an illusion caused by our flawed perception of mental temporality. The conclusion holding that perfect-prediction’s logical possibility precludes choice is a dispositive claim against compatibilism, which—to deny any contradiction between free will and determinism—analyzes free will as volition. Our acts aren’t truly voluntary.
Changed on July 2, 2012, 11 a.m.: Clarified that Predictor announces prediction in second-to-last paragraph. (See discussion with Joseph.)
I elaborate on the logic of the second-to-last paragraph in 10.1.2 Reflexive prediction, determinism, and the impossibility of free will
I'd have to go with picking the opaque box, if it's a perfect prediction by definition the prediction will be the same as the actual.
ReplyDeleteAs ever I've yet to find a definitive answer as to whether Quantum Mechanics can support determinism, or rather thought through all the consequences there of.
A, perhaps, even more interesting question is whether Q.M. supports Realism. Read about Leggett's Inequalities if it grabs you, Markus Aspelmeyer and Anton Zeilinger published a paper in Nature, April 2007.
I am beginning to wonder if free will is merely a useful illusion.
Well, there's a seemingly perfect argument, as you say, for picking the opaque box. But there's also a perfect argument for picking both boxes. It's not "fair" to pick one perfect argument and ignore the other.
DeleteOn determinism, keep in mind the two separate senses. I don't think qm is relevant to the determinism I discuss here: absence of free will. It's relevant to the other sense: absence of fundamental natural laws that are probabilistic. I endorse determinism in both the senses, but I'm confident free will doesn't exist and not confident that fundamental stochastic laws don't; perhaps the difference is because I understand the psychology relevant to the first sense but not the physics relevant to the second.
While I lack any physicist expertise, I think I can see that quantum mechanics--if taken as a description of reality--is inconsistent with realism and materialism. So one must weigh the conceptual arguments that materialism is true against the empirical arguments that qm is true--rather than a partial model.
I'm convinced free will is an illusion, but I don't think it's useful. Inevitable perhaps. My analysis on this point is at 10.3 "Why free will?" (http://tinyurl.com/7lw2zjb)
I guess the point of contention lies here:
ReplyDelete"But the Predictor goes exclusively with its prediction and makes no changes after Chooser's actual choice"
When is the actual "choice"? If it's defined as the final will, or desire, of the agent, proceeding the action (vocalising, or otherwise indicating the choice in an irrevocable manner) then, by definition, the "choice" cannot be changed.
Another question I have is; is the agent made aware of the perfect predictor's prediction?
Choice *is* the *action*, not the intention. The distinction is important for my argument because I claim that under the experiment's conditions, action would necessarily diverge from intention, disproving a compatibilist interpretation of free will.
ReplyDeleteChooser knows Predictor is infallible but doesn't know the prediction.
Sorry if I'm taking a while with this one, I feel I'm circling around a point, trying to get to it from a few different angles, I'm a little frustrated with myself actually.
ReplyDeleteSo the choice is the action not the 'volition', I can see how that ties in with the least contrived interpretation of Libet findings. An itch I can't scratch is this:
Could Libet's findings be that both the awareness, and rationalisation of volition were antecedent to the action?
That is, say we have a set of conditions, A1-A100 , and another set B1-B100 (actually I wanted to use Aleph rather than 100, but lack the option). These conditions can be things like firing rates of neurones, presence and numbers of neurotransmitters within the brain etc.
Conditions A1 to A100 exclusively lead to action 1 (pick both boxes)
Conditions B1 to B100 exclusively lead to action 2 (pick the opaque box)
It seems that a compatibilist could label conditions A(1 to 100) as the volition to perform action 1, and conditions B(1 to 100) as the volition to perform action 2. Awareness, and rationalisation of conditions A , or conditions B, need not preceed the choice/action.
If we had a condition set C that did not exclusively lead to either action 1 (both) or 2 (opaque), then I think causation, determinism and compatibilism would all fail.
I am a little concerned that this could be unfairly changing the definition of volition.
Next I have the question of whether an action (choice) that starts as action 1 and finishes as action 2 (or vice versa) should be regarded (i)as entirely as action 2, or be regarded (ii) as action 1 followed by action 2.
If (i) then all seems well in the good ship Compatibilism. Condition set B preceeded Action 2.
If (ii) then condition set A preceeded events, yet action 1 failed to occur, to me it would seem that both determinism and compatibilism fail and we are left Libertarians.
"What’s the outcome when Chooser—as seems entirely possible—wants to
pick contrary to prediction?"
I think either determinism fails, or yes, external events outside the chooser's control meant the choice was illusory (the best I can imagine is something like the chooser saying "BOTH - no, opaque" due perhaps to a time lag between conscious and subconscious thought, or if actions 1 and 2 were represented by buttons, an earthquake causing the chooser to push the undesired button).
If the idea of the subconscious is outdated I apologise.
"I am a little concerned that this could be unfairly changing the definition of volition."
DeleteYes, that's the problem with the account you offer. It's an account of a type of autonomy but not of volition, which is prior intent. "Subconscious" is usually rendered "unconscious" these days, but the concept is alive and well. But the problem of volition is about a state of consciousness.
"Next I have the question of whether an action (choice) that starts as action 1 and finishes as action 2 (or vice versa) should be regarded (i)as entirely as action 2, or be regarded (ii) as action 1 followed by action 2."
That's positing that the intention changes from 1 to 2 to produce 2. But there's no (psychological) reason for the intention to change. In essence, "psychology" predicts one intention, but physics predicts the other, and physics trumps.
"If (ii) then condition set A preceeded events, yet action 1 failed to occur, to me it would seem that both determinism and compatibilism fail and we are left Libertarians."
That's part of what's wrong with set A and set B as the relevant precursors. They're too encompassing to serve as stand-ins for conscious intent.
"Yes, that's the problem with the account you offer. It's an account of a type of autonomy but not of volition, which is prior intent."
ReplyDeleteSo I guess by this definition Libet has shown us volition is not actually prior to action.
"In essence, "psychology" predicts one intention, but physics predicts the other, and physics trumps."
I've been reading this over for a few days, and I'm not sure where any conflict between physics and psychology would arise in a hard determinists view of the universe. The way I understand it the physical laws that govern ion movement across cell membranes, the resulting potential differences, the behaviour of neurotransmitters at synapses etc,govern everything that makes up a "psyche". I understand why you are saying the physics "trumps" , but I fail to understand why a psychological prediction would be different.
"That's positing that the intention changes from 1 to 2 to produce 2"
Well, for example, say the choices were indicated by buttons, the individual could childishly intend to press button 2, but pass their fingers over button 1 beforehand in an attempt to trick, or test, the perfect predictor. Thus the intention, from the beginning, would be to press button 2, but the action would look like 1 changing to 2.
"That's part of what's wrong with set A and set B as the relevant precursors. They're too encompassing to serve as stand-ins for conscious intent."
If relevant precursors can't be modelled in this way, doesn't that suggest that determinism fails in describing conscious intent? Or do you mean the actual epistemological limits now?
"I've been reading this over for a few days, and I'm not sure where any conflict between physics and psychology would arise in a hard determinists view of the universe. The way I understand it the physical laws that govern ion movement across cell membranes, the resulting potential differences, the behaviour of neurotransmitters at synapses etc,govern everything that makes up a "psyche". I understand why you are saying the physics "trumps" , but I fail to understand why a psychological prediction would be different."
ReplyDeleteI appreciate your taking the time.
I'm not making a psychical exception from physical law. Note that Predictor isn't capable of predicting everything in the universe. It can't even predict what's happening in Chooser's brain. All it can predict is Chooser's overt choice. What would most likely be the case if Predictor could perfectly predict Chooser's overt choices? There are two relevant possibilities: 1) Predictor might get the overt behavior right and Chooser might manifest the proper intention, or 2) Predictor might get the overt behavior right and Chooser might manifest an intention to perform the other action.
I don't know whether 1) or 2) would be the result that occurs, but my point is that it's possible (even likely) that the way the Predictor's prediction would be upheld is by Chooser's conduct being counterintentional. This result is incompatible with compatibilism, which defines voluntary conduct as the result of intention.
"If relevant precursors can't be modelled in this way, doesn't that suggest that determinism fails in describing conscious intent? Or do you mean the actual epistemological limits now?"
Yes, determinism fails in linking conscious intent with conduct. I think one way to put it is that hard determinism insists on determinism applying to both conduct and "experience," but denies that there's a deterministic relationship between "experience" and conduct, which is asserted by folk psychology and supposedly combined with determinism in compatibilism.
Joseph,
ReplyDeleteYou're right: the way I've described the problem in this discussion, there's no "psychological" reason to assume intention will deviate from overt action. I answered one of your questions incorrectly: Chooser *is* informed of Predictor's prediction, a point that I should have been made clearer in the text. Thank you for helping me to clarify this.
In the essay, I try to make two points: 1) The incoherence of Newcomb's paradox explains why there seem to be two perfectly good solutions (Not, as usually thought, one must somehow be right.); 2) Newcomb's paradox can also be used to demonstrate how compatibilism fails. In my eagerness to accomplish the two elegantly, I didn't take note of the fact that point 2 involves a change in the usual Newcomb paradox description: **Chooser is told the prediction before making it.**
The "psychological" reason that Chooser will intend to do differently than Predictor's announcement is that when Predictor predicts the opaque box and then announces the prediction Chooser has an incentive to intend to choose both boxes, contrary to prediction. Then psychology can't be squared with Predictor's correct prediction in those cases where Predictor predicts the opaque box.
"I've been reading this over for a few days, and I'm not sure where any conflict between physics and psychology would arise in a hard determinists view of the universe. The way I understand it the physical laws that govern ion movement across cell membranes, the resulting potential differences, the behaviour of neurotransmitters at synapses etc,govern everything that makes up a "psyche". I understand why you are saying the physics "trumps" , but I fail to understand why a psychological prediction would be different."
ReplyDeleteI'm not sure I answered this directly enough. The ultimate reason hard determinism can imply the divergence of mentalistic and physical predictions is that psychology isn't exactly true: it's only a useful approximation. The mentalistic concept of intention doesn't reduce to physics--not because there's something besides physics but because psychology is inexact. Taking psychology as precisely true leads to the reflexive paradoxes of determinism because receiving a prediction can alter the intentional description without altering the causal properties of the world relevant to the choice.
For me, the weakness in Newcombe's Paradox comes from the way it 'predetermines' that reconsideration, by smuggling in the possibility of a perfect predictor at the outset. There can be no such thing as a perfect predictor, unless all that is, is in conformity with deterministic rules (thus susceptible to faultless prediction). Ergo, the paradox cashes out in determinism. Rather begging the question, no?
ReplyDelete(ps. you're leaning awful hard on the Libet data, Stephen - are you *certain* it can bear the load?)
This comment has been removed by the author.
ReplyDeleteThis doesn't refute incompatibilist FW.
ReplyDeleteCompatibilists and Incompatibilists define FW differently.Compatibilists define it as the ability to do what you want, or to act without duress.That is not affected by prediction.
Incompatibilists define FW in ways amounting to radical unpredictability. An Incompatibilist who agreed to the premises of the Newcombe problem would face a paradox.But no Incompatibilist voluntarist should agree to the premise of a perfect predictor in the first place, so nothing has been established.
If compatibilists and incompatibilists define free will differently, there would be no basis for debate between them about its existence. Perhaps this is your claim.
DeleteHere, I go after compatibilists. (I don't see voluntarists as worth the game.) My contention is that the prediction is what, in fact, the actor would do; and the paradox demonstrates circumstances in which this is not what the actor wants.
If you are not even bothering to deal with voluntariasts, you do not have a decisive arguement against free will.
Delete