Phil: Why the long face?
Sophie: My thought experiment failed.
The conviction that we have free will runs deep in many people, in part because they perceive themselves to have free will. For most, the conviction is grounded in experience, not rational or philosophical argument. Whenever a person can imagine having acted differently than they actually did act, a bit of experience is added to reinforce the conviction that free will is possible. An instance in which the person experiences constraint in their choice to act cannot subtract from the essential reality of free will. The conviction of having free will only says that we have it under the right conditions. It is easily granted that often these conditions are not met.
This arrangement, the idea that we can sometimes decide freely, and that experience can only support but not question the existence of free will, makes the idea virtually unassailable at the level of data and evidence. When there can only be verification but no falsification, what do you expect? This arrangement, however, does not amount to a second-level proof of free will. It makes no sense to say ‘If free will exists, then it can only be reinforced by experience, but not refuted.’ A statement of this structure could be made for any kind of metaphysical claim, but that does not make them true. A religious believer might accept the idea that ‘If god exists, then I will experience him through his works, and no event can cast doubt on his existence. Now I am in fact experiencing his works, therefore he exists.’ This is a reverse inference. Some reverse inferences work probabilistically (Krueger, 2017), but this one is particularly bad because it ensures that it cannot be questioned.
Libertarians, and other philosophers who argue in favor of free will, make two mistakes. One mistake is that they even bother trying to make a rational argument for free will. They don’t have to convince those who already believe it based on their experience. If anything, a difficult, quasi-logical argument might only confuse these believers. The other mistake is that – and this is the reason why I refer to these arguments as “quasi-logical” – much like induction from experience fails to prove free will, so does any attempt at logic. In a recent essay, I described how philosopher Pink (2004), after beautifully reviewing the history of the free will debate, surrenders to a set of irrational claims in order to defend the libertarian view, that is, the view that when we choose X without evident constraint, we are free because we could have chosen Y. That we can imagine to choose Y proves nothing; it only begs the question.
Pink knows this. So he has to tell us more. One of his offerings is to attack the view that there are two known ways to model events: causal determinism and chance. The project of science is to understand and model events and their regularities. This means that through observation and experimentation, causal determinism is pushed as far as it can go. The remainder of the variance in the observations is attributed to chance. This project is a subtraction game. Chance = Total Variation – Causally Determined Variation. This view implies that variation in human (your) behavior can be modeled this way as well. Psychology’s mission is to explain what it can in terms of causally comprehensible laws and attribute what is left to chance, at least for the time being. No one needs to worry that the limit will be reached such that all behavior will have become predicted and explained. It’s just too complicated, and perhaps there is a percent variation that is due to chance in the truest sense, that is, irreducibly uncertain.
The cause-plus-chance model does not allow for a third possibility. This is where libertarians say ‘But there is: free will. Human Behavior = Causes + Free Will + Chance.’ Pink’s defense of this equation is that scientists have not proven that causation and chance is all there is. Therefore, Pink concludes, a third force can be added. This argument is a variation of the rhetorical claim that anything in my imagination may be real, or deserves to be real, or is real, unless you prove to me that it cannot be so. This line of argument leaps over two barriers that have been erected in the past, and for good reason. One barrier is parsimony: a two-force model is better than a three-force model, all else being equal. The second barrier (which is a corollary of the first) is that if you’d like to add a force, you need to show that it is necessary to do so. It is insufficient (and unfair) to pass the burden of proof onto the opposition. If you want to have free will, prove its existence positively! A lesser standard is to show that the standard model of cause-plus-chance is insufficient. If insufficiency can be shown, then free will may be introduced as a candidate to fill the gap, until a time comes when we have positive evidence. The presumed dark energy in the universe is an example of how this works. No one has any evidence for dark energy (it is dark after all), but if it existed, we could understand why the universe is expanding at an accelerated speed. Here, the standard gravitational model is clearly insufficient.
In the case of the universe, there is an undisputed fact, the accelerated speed of expansion, that must be explained, and therefore we resort to postulating a yet undemonstrated force to keep the mathematical model going. Where is the analog in the realm of human experience? There is no human behavior that, on the face of it, is inconsistent with standard cause-plus-chance models. Therefore, no additional force needs to be postulated. What we do have is the widespread belief in free will. To think, however, that we need to postulate free will to explain the belief in free will returns us to the logically most barren form of reverse inference. Indeed, the belief in free will is open to explanation within the purview of the standard cause-plus-chance model.
To make all this a bit more imaginable, put yourself mentally into a T-maze. You walk down the hall and you know that you must make a left or a right turn at the T-section and reach one end of the perpendicular hall. You have a free choice. No one is forcing or incentivizing you. It is up to you. In that sense you are free. But this is not the libertarian understanding of free will. This understanding says that you are not making a right turn because you have a pre-existing tendency to make a right turn or because you roll a psychological or a physical die. At the intersection, you are literally free. You can go either way. There are no pre-existing psychological tendencies or leanings; all those would be causes and causes negate libertarian freedom.
Now we see that you took a right turn. With this singular event, there is little hope to discriminate between causation and chance (let alone free will). Perhaps a concurrent brain scan reveals that seconds before turning right, your brain already preparing for the move. If so, this would be a point for causation. Now suppose, you did the maze 100 times. We’d have two pieces of information: the proportion of right turns and the autocorrelations. Autocorrelations, if they are not zero, support causation because they mean that a given turn can be predicted from past turns. The size of these correlations speaks to the relative weight of causation and chance. Free will would also demand zero correlations. It is easier to separate free will from causation than it is to separate it from chance. The proportion of right turns speaks to causation inasmuch as it is close to 0 or 1. But even a proportion of 50% of right turns is consistent with causation if there is a high autocorrelation. Perhaps you are alternating the direction of the turn. If you do, how can you show that this is freely willed? You made up your mind, you say, to alternate, but could as easily turned right every time. In short, whatever choices you make in the fictional maze, the data have no bearing on free or unfree will. They only speak to the battle between necessity and chance. The ‘confirmation’ of freedom lies only within the subjective experience, which again is no proof.
Krueger, J. I. (2017). Reverse inference. In S. O. Lilienfeld & I. D. Waldman (Eds.), Psychological science under scrutiny: Recent challenges and proposed solutions (pp. 110-124). New York, NY: Wiley.
Pink, T. (2004). Free will: A very brief introduction. Oxford, UK: Oxford University Press.
 Is there a replication crisis in thought experiments?