Free Will and Determinism

I recently read the introduction to History of Civilization in England by Henry Thomas Buckle (1821-1861). In this long essay, he outlined his dissatisfaction of history study being a collection of facts and called for its transformation into a proper science, which generates knowledge in the form of causal connection. He made many insightful remarks and one that got me thinking is about free will vs. determinism. He Obviously took the determinist position when he claimed that if he has enough information about the circumstance which his well acquainted friend is in, he can predict with certainty what his friend will do next. He keenly pointed out a problem of free will. Before we can have a meaningful discussion, I want to make explicit the essential conception of a free will. We think of a free will as an active agent capable of dictating one's thoughts and actions independent of one's experience or the given circumstances. Buckle's insight is that in order to freely will, we need to have another authoritative agent independent of deterministic laws to inject such decision into our consciousness given that the physical universe evolves according to deterministic laws. But how does that agent operate? Given that it lives in a world also evolves according to some deterministic laws, since it has to be free as well, it would require yet another free active agent. (This kind of reasoning is not unlike Descartes's notion of mind-body duality, which is wittily summarized as "ghost in the machine" by Gilbert Ryle.) Thus we enter an infinite regression that unsuccessfully evades the contradiction between determinism and free will.

This is a tough question because one is reluctant to give in entirely to either side. Some people suggested that quantum physics might hold the answer due to the truly random events allowed in quantum physics. I initially felt unsatisfied by such proposal because we want to think of will actively, instead of  randomly decides on things. But lately I came to realize that in observation of a single act, the two are indistinguishable! My worry was due to my subjective experience of activeness. Let's put aside the subjective content and imagine an experiment in which Buckle watches his well acquainted friend for a while and then predicts what his friend will do next. If free will exists, Buckle's friend will be able to decide on things independent of his experience and as a consequence, Buckle will not be able to predict with certainty even with all the information. In observation, that is precisely what you expect from a random event whose outcome is independent of the past. Although our subjective experience is too intricate to give us direct clues, the unexpectedness of a certain decision might have germinated from a seed of randomness allowed by quantum physics.

This resolution proposes that we give up the deterministic view, which arguably is suggested and supported by classical physics, to allow for some non-deterministic events, which is supported by quantum physics, our current best knowledge. This answer might still upset people believing in free will, because randomness seems to dismiss one's active control over his or her life. But the pursuit for the possibility of control only pulls us back to the infinite regression. While the moral implication in absence of an active agent is interesting to consider, I think I will spend more time on pondering about how quantum mechanics allows for randomness when all its laws, like those in classical physics, are deterministic (in probability). In quantum mechanics, randomness creeps in at the act of measurement whose outcomes' probabilities are predicted by input and laws, but not the outcome itself. I will subject further discussion on this topic to a future article. Let's revisit our thought experiment and introduce a modification: instead of predicting the exact action of his friend, Buckle will predict the probability of each possible action and we shall repeat the experiments many times (maybe with identical clones of his friend). If free will can actively control one's decision, Buckle's friend can decide how frequently he wishes to do a certain thing and thus he can manipulate the frequencies of each action rendering Buckle's prediction of probability incorrect. This is why the condition of "a single act" is needed in my statement of indistinguishability. However, this experiment is hard, if not impossible, to carry out because after each decision, the state of Buckle's friend changes and cloning the quantum state of Buckle's friend is impossible: we need to find some decision whose probability does not change when repeated. Hence a definite answer to whether free will exists, and if so whether it exists in the weak, random form or the strong, ghost form, is not known yet.


  1. I think this is a well written, thought out post. A few things -

    1) This is a trivial matter, but I think that bringing quantum mechanics into your argument is unnecessary. The true heart of your argument regarding QM can more easily described simply with statistics - a state may be described by a series of parameters while estimated reactions to stimuli may simply be described by a singular probability as a function of these parameters (p-value). Unless you claim that the probability of a mechanical or chemical reaction in the brain (or whatever your source of free will) is dominated by quantum uncertainty, it seems safer to move away from analogies to quantum mechanics.

    2) The problem with your distinction of "a single act" is that you are dealing with issues of time scale. If we define in binary the success of "a single act," we must first define "a single act," which it seems you recognize as being extremely difficult to define. You may say someone did X, or someone didn't do X. However, we may also state that someone hasn't done X YET (which is the standard go-to of failed prophets throughout history). So in this case, your argument fails as we haven't defined an exact time scale by which "a single act" may or may not happen.
    More fundamentally, a decision to do X and it's opposite, a decision to NOT do X is ill defined. In human terms, it is generally not true that deciding action Y is a decision not to do X. Simply put, there exist no such things as "single acts" in any meaningful, measurable, experimental way, at least if the experiment involves simply observation and prediction of reaction to non-controlled stimuli.

    3) Now for another thought experiment/clarification of your idea. Suppose we COULD clone Buckle's friend in a particular state (as Buckle's friend is too large to be defined by a quantum state, technically, he isn't bound by the no-cloning theorem). Now suppose we expose Buckle and all of his replica's to an external stimuli in a box with a binary response. Now suppose we collect sufficient statistics that the Law of Large Numbers applies and that the data converges on a single proportion mu.
    Now, contrast this case to the realistic case where we cannot clone Buckle's friend, but still subject him repeatedly to the same experiment, in the same box, with the same stimuli. Given the same number of experiments, would the data converge on the same proportion mu? I would say not, as history is likely a good explanatory parameter for any function predictive of reaction to stimulus.
    Why is this relevant? Because it seems that though we may detect noise in our data, we cannot define TRUE randomness in the latter system. True randomness can be defined as the variation in the first proposed experiment. However, lack of controls placed in the latter experiment causes doubts as to the source of the error and variability - is it a result of prior stimuli, or is it the result of random noise? If we cannot control for as many parameters as possible, we cannot estimate the uncertainty given those parameters, therefore we may not determine whether this error or "randomness" is systematic or natural. Therefore, there does not seem to be a way to be able to define the concept of randomness in terms of a singular action. Therefore, your point that in a singular action random action and will are indistinguishable is true, yet moot - we cannot define randomness outside of the context of repeated, controlled experimentation.

    1. 4) We know behavior modification works in all manner of animals, humans included. Therefore, reaction to a specific stimulus can be modified. Where then is free will? Is free will decreased? Is free will an entity than lies on some sort of continuum on which it can be moved/is there a degree of free will? It seems that if we admit that free will may be removed, can it be recovered? Where then does free will originate from?

      5) Is it free will when, as infants, we shit our pants, or is it free will when we are taught to hold in our shit until we may make it to an appropriate place to shit. It seems that the wondering whether we are deterministic beings, or beings endowed with free will is just scratching the surface of understanding the origins of our actions.

      Thanks for the insightful post, and keep up the good work!

  2. A very nice post. Just a first thought (I think I'll have more later):

    There's an interesting distinction I think should be drawn within your point about the "independence" of free action from circumstances.

    There's a way in which total independence from circumstances would really undermine our wanting to call someone free. For example, a crazy person may act (and think) completely independent of their circumstances, but we do not for that reason call them free. We say rather that they are a slave to their illness.

    On the other hand, we also want to say that a free agent *does* act in a way that is very much dependent on circumstances. When you weigh the advantages and disadvantages of a particular course of action, and then decide on the best one, we absolutely do not say "insofar as you deliberated, you were unfree." In fact, the opposite is true: a free person is someone who can deliberate on circumstances and freely choose the best course of action.

    I think this point might go some ways toward pushing back against your middle paragraph, where you say "If free will exists, Buckle's friend will be able to decide on things independent of his experience and as a consequence, Buckle will not be able to predict with certainty even with all the information." I do not think that freedom and unpredictability are identical.

    1. For an incompatibilist (determinism is not compatible with free will - seems to be Falcon's stance), free will requires unpredictability to some degree. If we have perfect predictability, then since the future can be known to us with complete accuracy, we will not be free to deviate from the predicted path (causal determinism). Of course, if we adopt a compatibilistic definition of free will, then we can be both free and completely predictable.

      Insanity opens up a whole new can of worms. Do we lose all freedom of will if we hallucinate once? How delusional do we have to be before we are not free? And then there are personality disorders. A person with severe depression cannot bring himself to get out of bed in the morning - is that lack of free will?

    2. Even if unpredictability is somehow essential to freedom (I remain agnostic), freedom could require it without being synonymous with it.

      There is also a question of where this unpredictability is supposed to go. On the one hand we might require every single free act to be somehow partly unpredictable. I think that would be a bad theory: it seems to me like that would mean virtually everyone is unfree in the majority of their actions. More reasonably, we might say that there has to be some larger unpredictability in the aggregate outcome of their actions.

      But as my previous comment suggested, I don't think either of these views (correct or incorrect as they may be in some negative way -- i.e., merely placing constraints on freedom) gets to the heart of the issue, because they ignore questions of both first-person experience and the meaning of reason. Talking primarily in terms of prediction presupposes the behaviorist viewpoint that it only matters what freedom looks like from the outside.

  3. The problem for free will is that human choices are either deterministic or they are random, neither prospect makes free will possible, and there is no third option.

    Here's the transcript of an episode of my show I did on this - http://causalconsciousness.com/Episode%20Transcripts/14.%20%20Why%20Both%20Causality%20and%20Randomness%20Make%20Free%20Will%20Impossible.htm