"Would someone without free will do this?" is one of the most common responses I've heard to the proposition that free will might not exist, followed by a valiant attempt at doing the most unexpected thing one could think of.
On the surface, the existence of free will appears to be the Occam's Razor belief. What is there to doubt? It feels like my actions are my own. Why wouldn't they be?
Though when we take a look at the ways free will could manifest, we have to jump through a number of hoops to find support for its existence as we'll define it. This is not to say that it does not exist, just that it has forfeit its position as the simplest explanation, and it may require you to subscribe to a number of abstract, not-currently-provable theories.
It's important though that we first agree on such an elusive and ambiguous concept like "free will" means. Even looking at the same definition, we may internalize it to mean different things. On the free will debate, I think there are 3 schools of thought that are relevant here.
Determinism poses that every event and action is the inevitable consequence of prior events and conditions, following natural laws. If you knew every detail about the universe at one moment, you could theoretically predict everything that would happen next. Everything is strictly cause and effect.
Indeterminism argues that causality can not fully explain outcomes, some things are truly random or unpredictable. This view acknowledges the role of fundamental randomness quantum mechanics plays in our world. There are genuine "forks in the road" in nature where multiple outcomes are possible.
Compatibilism argues that free will is compatible with determinism by redefining what we mean by "free." Instead of requiring some supernatural escape from causation, we're free when we can act according to our own authentic desires without external constraint. So even if our desires are shaped by prior causes, we're still exercising free will when we act on them.
For example, if you choose to drink coffee in the morning, a compatibilist would say you're exercising free will if:
You made the choice based on your own desire for coffee
You weren't forced by someone else
You had the capability to act on your choice
The fact that your desire for coffee might itself be determined by prior causes (like your genetics, habits, or past experiences) doesn't negate the freedom of the choice in the compatibilist view.
compatibilism focuses on freedom of action rather than freedom of will. The question isn't "Could you have done otherwise in exactly the same circumstances?" but rather "Were you free to act on your own authentic desires?"
Determinism is an important starting point for this conversation and getting to know the other schools of thought, but we can remove it from our focus. The double slit experiment, among many other quantum studies, tells us we cannot depend on the universe to be fully deterministic. This was a hard pill to swallow for a while. "God does not play dice!" Einstein cried!
Though at this point, this is a widely accepted view, and it moreso comes down to why this is the case and what the mechanism behind the unaccounted indeterminism is.
Compatabilism is not as much claiming a certain mechanism by which the universe plays out, but rather a philosophical reframing to make peace with living in a world explained by Determinism or Indeterminism. It is offering a definition of free will catered to what ultimately matters to us in a practical life.
For this reason, we'll be working off the view posed by Indeterminism. At face value, it does not necessarily reject or support free will, leaving us a fine bit of space for discussion.
Also, it is worth noting as well this is not to be confounded with consciousness and experience, whether a given life has a "ghost in the shell" experiencing its sensory inputs is currently undetectable to us and thus outside the scope of this post. We are concerned with whether there is a possible avenue where we can have tangible effects on the world that stem from a choice not exclusively explainable by a chain of deterministic causes and effects.
When we model physical systems, we're typically proceeding with one of two classes of systems. Deterministic (like ODEs) and stochastic (SDEs).
The deterministic case allows us to purely model a system with our typical simplified physics equations, and often, in the absence of strong environmental noise and other effects we can't account for, this can often be good enough despite its reductionism.
You may have seen jokes about how we can predict where a certain asteroid will be 1000 years from now, but fail to predict the weather 30 minutes from now. This is because in space our deterministic models work quite well; there is little external or unexplained variables we have to concern ourselves with.
Something like the weather on the other hand is typically analyzed with stochastic models.
These can be useful for two reasons
System Incompleteness
Our equations cannot account for all of the factors of the system that can influence its trajectory. There are some other variables that we either are unaware of or do not have models for, but they very much constitute a non-negligible effect.
If we want to model a wildly complex system like the population of deers in an ecosystem, we may keep track of things like amount of food, predators, climate, nutrients, genetics...
Despite this list, you could probably come up with 100s of more things that can have sizable effects!
Chaos
very slight imprecisions in our measurements and initial conditions can have profound effects on the outcome.
"The Butterfly Effect", the flap of a butterfly's wings causing a hurricane weeks later.
So the scenarios in which we choose to model things as random come down to variable omission and measurement imprecision; complexities that merely masquerade as randomness, but we roll with it anyway for the convenience factor. Notice however these issues are not at all a nail in the casket for deterministic models! It suggests that if we could somehow account for all of these variables with perfect precision, our models could still work!
However, let's take a brief aside about some of the perks stochastic models give us. This will hopefully illuminate the deterministic nature of stochastic or sometimes pseudo-random processes.
This here is the equation for brownian motion. A very simple stochastic model.
At any given moment a particle is moving in a direction that is a sum of a constant value/direction (drift) plus a randomly sampled value (uncertainty) where epsilon is the randomly sampled value. This may also be referred to as something a random gaussian walk.
The first perk is that there is often a distinct deterministic component and a distinct random component, allowing us to analyze them separately.
Secondly, we can choose to run many simulations. And instead of looking at how an individual point turns out, we can look at the distribution of outcomes.
Imagine if we only ran the simulation once and got back the dark red line towards the top. This would give us a very distorted picture of the underlying process! Instead we can see where most processes end up, and tell a story from there.
When we run random processes on a computer, we use something called an RNG (random number generator). Typically, these random number generators are "seeded", basically predetermining all of the numbers they will output. The same way how using a certain seed in minecraft can allow you to recover the same world (perhaps this warrants another post on how shockingly few bytes can beget immense complexity). This allows for us to reproduce the same exact trajectory if we are provided with the same initial conditions, the same seed, and the same equations. Hence the term "pseudo-random".
This last line is key. If we can:
Account for all influential variables accurately
Have perfect precision
For things that are actually "random", pre-determine the sequence of random values they will return
Any random process can be made wholely deterministic.
We've now said that the random parts of real world processes can actually be argued to be deterministic by 1 and 2, but for 3, is there any true irreducible randomness? Is there a Random Number Generator in real life?
Enter quantum mechanics, the last frontier of what may be true pure randomness.
So now once more following the stochastic modeling equations, let's form a perspective on our universe. Everything once more can be explained by the deterministic component and the random component. But now our random component is strictly reduced to what happens at quantum forks, and how that can have downstream effects via chaos.
So now, let's imagine that quantum did not exist in our world. If this were the case then Hard-Determinism would win. The very initial conditions of the Big Bang would directly and deterministically explain the shoes you chose to put on today. It's a very very long chain of causality though it theoretically could be made fully explainable much like our weather example.
In this model, there would be no room for free will to exist. Everything is cause and effect explained by initial conditions. We know our decisions and actions are the result of activity in the brain which is caused by current sensory inputs and state which can be explained by what happened before that and before that and so on back to when your cells first formed and when these atoms even came into being in the first place. After all, despite the mystery of the brain and behavior and its complexity, it is just another physical system and plays by the same rules. When we fire an electrical signal through a circuit we can trace the path it took to yield its outcome. The only way this could change is if something within this physical system changed. Free will would require some external supernatural intervention that can exist outside of this physical system but be able to have an effect on it.
Let's not jump too quickly to this pessimistic conclusion though. After all, we do live in a world where quantum mechanics exists and it does affect how things happen.
Adding quantum back in, we suddenly do have a means by which free will can occur. Is it crazy to believe that free will could be enacted through quantum interactions? I'd say at least no crazier than any of the other existing theories on quantum mechanics! It's anyone's guess.
This is what I've come to call Humans as Agents of Chaos. Most quantum events amount to nothing. A quantum event within the atoms of a rock chances are doesn't have the means to affect very much.
But when we look at the granular levels at which brain activity happens where ion channels open and close to ultimately determine whether a brain cell fires or not, it could be the case that what happens in the quantum realm can become relevant here.
Taking a bit of a leap of faith, we could imagine a series of quantum events determining whether I make the decision to go to work today or not, which can set off a number of chain reactions that determines our next political leader.
Humans provide a conduit for these nanoscopic quantum effects to have very large scale effects on the world.
However, this all just posits that quantum mechanics has an effect on human behavior and thus the world. We still haven't tied this back to how this relates to free will.
The question now is free will mediated through quantum effects? Does my will decide what happens in the quantum realm in the scope of quantum mechanics within my own body and brain? Or do quantum effects still operate tangential to free will and any perceived control is still only an afterthought to actions decided by the deterministic and quantum components of our world without my role in any of it? This raises some other questions as well like, how do I manage to will the outcome of quantum forks in systems that are strictly within myself? Do other systems where small effects on the quantum level like the weather, could they have free will? Is this view of discretizing will into a per-person or per-life basis even make sense?
This also creates an interesting distinction between humans and AI like language models. Language models deterministically process their input to yield a distribution over all possible next words to output. This distribution will always be the same. But then, using an RNG, we randomly sample the next word from this provided distribution. A difference in sampled words early on in a response can be the difference between a very angry response and a very friendly response to the same question back to back, just by using a different set of random numbers. So language models ran on GPUs do not actually ever have their behavior determined by quantum effects, and we can easily achieve the 3 qualities we described earlier. The space for free will to exist here is empty. By this description the weather has a more compelling case for having free will than an AI written in PyTorch, and I don't see this as a hole in the explanation. We've defined free will as having any kinds of means of holding the reins of indeterminism while being careful not to confound this with human-like qualities of personality, emotion, goals, reactions...
Unfortunately, this is about where this post ends. By process of elimination we've been able to rule out the ways that free will, as we've defined it, can and cannot manifest. As originally stated, the means by which it can exist require you to take on some speculative theories, though there isn't much reason to believe one means may be more plausible than another, aside from Occam's Razor. We've answered some questions but invoked many more. Though we're now reaching a bit of a dead end with how much more can be said, and you might just have to call it a day and have faith in which ever theory you prefer.
I have never been particularly religious. I think the faith needed to believe in a higher power is just as much as that needed to believe there is nothing at all out there. Though without delving into typical traits of a benevolent god or a book telling you how to live, I think this perspective of everything as a long chain of causes and effects makes you wonder, "What was the first cause?"
Comments