Editor's comment

Patricia Smith Churchland is Chair of the Philosophy Dept. at UCSD. Her work relating recent brain studies with philosophy is extensive.

Reading on: Neuroscience and free will

Churchland, Patricia Smith Brain-Wise: Studies in Neurophilosophy The MIT Press, Cambridge, MA 2002 [abridged – 2300 words]

Much of human social life depends on the expectation that agents have control over their actions and are responsible for their choices. In daily life it is commonly assumed that it is sensible to punish and reward behavior so long as the person was in control and chose knowingly and intentionally. Without the assumptions of agent control and responsibility, human social commerce is hardly conceivable. As member of a social species, we recognize cooperation, loyalty, honesty, and helping as prominent features of the social environment. We react with hostility when group members disappoint certain socially significant expectations. Inflicting disutilities (e.g., shunning, pinching) on the socially erring and rewarding civic virtue help restore the standards.

In other social species too, social unreliability, such as a failure to reciprocate grooming or food sharing, provoke a reaction likely to cost the erring agent. In social mammals, at least, mechanisms for keeping the social order seem to be part of what evolution has bequeathed to our brain circuitry. The stability of the social-expectation baseline is sufficiently important to survival that individuals are prepared to incur some cost in enforcing those expectations. Much of our behavior is guided by expectations of specific consequences of events, not only in the physical world, and but also in the social world.

If the reward and punishment system is to be effective in shaping social behavior, the actions for which the agent is rewarded or punished must be under the agent’s control. The important question, therefore, is this: What is it, for us or baboons or chimpanzees, to have control over our behavior? Are we ever really responsible for our choices and decisions? Will neuroscientific understanding of the neuronal mechanisms for decision making change how we think about these fundamental features of social commerce? These are the places where issues about free will bump up against practical reality and our developing understanding of what is fair, what is reasonable, and what is effective in maintaining civil society.

One tradition bases the conditions for free will and control on a contrast between being caused to do something and not being so caused. For example, if someone falls on me and I fall against you, then my hitting you was caused by the falling body; I did not choose to hit you. I am not responsible, therefore, for hitting you. Were you to punish me for hitting you, it would not help me avoid such events in the future. Examples emanating from this prototype have been extended to the broader idea that for any choice to be free, it must be absolutely uncaused. That is, it is suggested that I make a free choice when, without any prior cause and without any prior constraints, I make a decision that results in an action. This contracausal construal of free choice is known as libertarianism. Is it plausible? That is, are the paradigm cases of free choices actually uncaused choices?

As Hume demonstrated in l739, the answer is no. Hume argued that our free choices and decisions are in fact caused by other events in the mind: desires, beliefs, preferences, feelings, and so forth. Hume made the deeper and more penetrating observation that an agent’s choices are not considered freely made unless they are caused by his desires, intentions, and so forth. Randomness, pure chance, and utter unpredictability are not preconditions for attribution of responsible choice; Hume puts the matter with memorable compactness:
“Where [actions] proceed not from some cause in the characters and disposition of the person, who perforni’d them, they infix not themselves upon him, and can neither redound to his honor if good, nor infamy, if evil.”

Logic reveals, Hume argued, that responsible choice is actually inconsistent with libertarianism (uncaused choice). Someone may choose to climb onto his roof because he does not want the rain to come into his house, he wants to fix the loose shingles that allowed the rain in, and he believes that he needs to get up on the roof to do that. His desires, intentions, and beliefs are part of the causal antecedents resulting in his choice, even though he may not be introspectively aware of them as causes.
If, without any determining desires and beliefs, he simply went up onto the roof—for no reason, as it were—his sanity and hence his self-control would be seriously in doubt.

More generally, a choice undetermined by anything the agent believes, intends, or desires is the kind of thing we consider out of the agent’s control, and is not the sort of thing for which we hold someone responsible. Further-more, desires or beliefs that are uncaused (if that is physically possible), rather than caused by other stable features of the person’s character and tempera-ment, likewise fail to be conditions for responsible choice. If a desire suddenly and without antecedent connection to my other desires or my general character were to spring into my mind—say, the serious desire to become a seamstress— I would suspect that someone must be “messing with my mind.” The brain presumably has no mechanism for introspectively recognizing a desire to fix the roof as a cause, just as it has no way of detecting in introspection that growth hormone has been released or that blood pressure is at 110/85. A desire, nevertheless, is most certainly a cause.

Neither Hume’s argument that choices are internally caused nor his argument that libertarianism is absurd have ever been convincingly refuted. Notice, moreover, that his arguments hold regardless of whether the mind is a separate Cartesian substance or a pattern of activity of the physical brain. And they hold regardless of whether the etiologically [etiology – philosophical study of causation] relevant states are conscious or unconscious.

In fact, moreover, the brain does indeed appear to be a causal machine. So far, there is no evidence at all that some neuronal events happen without any cause. True enough, neuroscience is still in its early stages, and we cannot absolutely rule out the possibility that evidence will be forthcoming at some later stage. Given the data, however, the odds are against it. Importantly, even were uncaused neuronal events to be discovered, it is a further, substantial matter to show that precisely those events constitute choice. They might, for all we can know, have to do with features of growth-hormone release or variations in the sleep/wake cycle.

Though all events in the brain may be caused, this does not imply that actions are predictable. Causality and unpredictability are entirely compatible. Causation concerns conditions that bring about an event, whereas predictability concerns what we know about such conditions. When an event occurs in a complex system, we may know that the event is causally governed, even though on any given occasion we may not know exactly what conditions actually obtain, and hence are unable to predict precisely the nature of the event. Nevertheless, despite our inability to make precise predictions, we can often make useful general predictions. Thus I might be able to predict that a dollar bill dropped from the top of the Eiffel Tower will fall to the ground in less than two minutes, but I will be unable to predict exactly the fluttering pattern and its precise downward trajectory. Those subtle changes in movement will depend on moment-by-moment changes in air currents, and these changes will occur much faster than I can take relevant measurements and do the relevant computations, even if I were lucky enough to have very powerful computational equipment. Every movement of the dollar bill is, nonetheless, caused.

Similarly, brain events relevant to decisions and choices are probably all caused events, but this does not imply that I can predict with any great precision what you will say if I ask you for directions from UCSD to the Salk Institute. I can predict roughly what you will say, however, if I know that you are familiar with the area, that you are alert, paying attention, are hot easily disoriented, and that you tend to be forthcoming when asked for directions. I can also predict with considerable confidence that given the opporiunty, a human will go to sleep at night for at least a few hours, that he will want to eat and drink at some time during a 24-hour period, that he will not want to sit for very long naked on an iceberg, and so on. I can predict that a neonate [a baby from birth to four weeks] will suckle, a puppy will chew shoes, and that most undergraduates will name carrots as the first vegetable that comes into their minds. But these are rough and general, not precise, predictions.

The brain is a dynamical system of enormous complexity. The human brain is calculated to have about 1012 neurons and about 1015 synapses. The time scale for neuronal events is in the millisecond range. If we assume that synaptic events and neuronal events are the only causally relevant events, then to a first approximation, this means that the human brain has about 1015 parameters that can vary over roughly 1 - 100 milliseconds. (This is a conservative estimate, since there are intraneuronal events, such as gene expression, that are also
relevant.) These figures mean that it is not physically possible to take all the relevant measurements and perform all the relevant computations to grind out a precise prediction in real time. So predicting on a neuron-by-neuron or synapse-by-synapse basis is even less realistic than predicting the precise path and flutter of the dropped dollar bill. The logical point, therefore, is this: causality does not entail predictability, and unpredictability does not entail noncausality. Put another way, causality and unpredictability are entirely consistent.

As we reflect on what would have to be true for us to have free choice, we tend to be impressed by the fact that absolutely precise prediction of an agent’s behavior is really impossible given the relevant variables and time scales. We nurture the hunch that if you cannot predict whether I will choose a green salad or a beet salad, or whether I will choose to say “Hi” or “Good morning,” then my choices are really uncaused and therein lies my freedom to choose. The hunch may be the more compelling if it gets support from this tacit assumption: since “uncaused” implies “unpredictable” “unpredictable” implies “uncaused.” As I have shown, however, this is quite mistaken. The implication goes only one way. “Unpredictable” does not imply “uncaused.” Once the logic of the relation between causality and predictability are clarified, no logical rationale remains for deriving expectations of noncausality from facts of unpredictability.

Nonetheless, the idea that randomness in the physical world is somehow the key to what makes free choice free remains appealing to those inclined to believe that free choice must be uncaused choice. With the advent of quantum mechanics and the respectability of the idea of quantum indeterminacy, the suggestion that somehow or other quantum level indeterminacy is the basis for a “solution” to the problem of free will remains attractive to some libertarians. Stripped to essentials, the hypothesis claims that although an agent may have the relevant desires, beliefs, etc., he can still make a choice that is truly independent of all antecedent causal conditions. On this view, the agent, not the agent’s brain or his desires or his emotions, freely chooses between cappuccino and latte, for example. It is at the moment of deciding that the indeterminacy or the noncausality or the break in the causal nexus — whatever one wants to call it — occurs. The subsequent choice is therefore absolutely free.
This is meant to be an empirical hypothesis, and as such, it needs to confront neurobiologically informed questions. For example, what exactly, in neural terms, is the agent who chooses? How does the idea of an agent who chooses fit with what we understand about self and self-representational capacities in the brain? Under exactly what conditions do the supposed noncaused events occur?

Does noncausal choice exist only when I am dithering or agonizing between two equally good — or perhaps equally bad — alternatives? What about when, in conversation, I use the word “firm” rather than the word “stubborn”? Does it exist with respect to the generation of desires? Why not? There are also questions from quantum physics, such as these: What is the mechanism of amplification of the nondeterministic events? Were quantum effects of the envisioned kind to exist, how could they fail to be swamped by thermal indeterminacy?
These are just the first snowballs in an avalanche of empirically informed questions. Part of their effect is to expose the flagrantly ad hoc character of the hypothesis. That is, it is based more on a desire to prop up a wobbling ideology than on factual matters.

Provisionally, therefore, let us adopt the competing hypothesis, namely that Hume is essentially right and all choices and all behavior are caused, in one way or another. The absolutely crucial point, however, is that not all kinds of causes are consistent with free choice; not all kinds of causes are equal before the tribunal of responsibility. Some causes excuse us from culpability; others make us culpable because they are part of the story of voluntary action. The important question is what are the relevant differences among causes of behavior such that some kinds play a role in free choice and others play a role in forced choice. That is, are there systematic brain-based differences between voluntary and involuntary actions that will support the notion of agent responsibility? This is the crucial question, because we do hold people responsible for what we take to be their actions. When those actions are intentionally harmful to others, punishment, varying from social disapproval to execution, may be visited upon the agent. When, if ever, is it fair to hold an agent responsible? When, if ever, is punishment justified?