The discourse on free will and determinism is over 2,000 years old, so in some respects I feel a little late to the party. But this question has recently become relevant in my own academic interests, so I've found myself exploring it more and more.

To the uninitiated, the debate goes like this: Determinism is the idea that every event has already been predetermined. Events are the result of what follows—cause and effect—and if you could process every factor in a chain of events you would not only be able to see how the result was a necessary end, but predict what events would follow. In terms of humans, determinism says that every decision you make is predetermined by external context, social and moral obligation, and by your own character. Your character, of course, is the result of a lifetime of experience shaping your brain into what it is now. So on and so forth...

Many of the early Behaviorist in psychology were Determinists, believing free will an illusion and that every behaviour was predetermined by a root cause. John B. Watson—considered the father of Behaviourism—famously said:

Give me a dozen healthy infants, well-formed, and my own specified world to bring them up in and I’ll guarantee to take any one at random and train him to become any type of specialist I might select—doctor, lawyer, artist, merchant-chief and, yes, even beggar-man and thief, regardless of his talents, penchants, tendencies, abilities, vocations, and race of his ancestors.

The opposing camp to determinism is Libertarian Free Will, which is the idea that you have full control and agency of your choices and your self. We have options that aren't set in stone. René Descartes defended this view, saying that even if the outside world to our selves is deterministic, we at least have free will of our own minds.

Nestled in between these two extremes is Combatibilism, sometimes called Soft Determinism, which tries to find a common ground where both determinism and free will can co-exist. David Hume took this approach, stating that we may have free will and agency of our actions but what we do is often constrained by our own character. The majority of people seem to sit in this uncomfortable middle ground, though there are staunch defenders of either extreme.

Dan Barker's Take On Things

Dan Barker sees this problem a bit differently. The issue, as he says, more likely comes down to language and how we are using these terms. In Free Will Explained, he tries to unravel the debate through numerous analogies and introduce his position as an Acompatibalist* By this term I think he means that the issue of free will vs determinism isn't an issue at all. It's comparing apples to oranges. These two concepts can co-exist because they don't actually oppose each other. The real debate is between determinism and indeterminism. Free will is something different entirely, and that different thing is an illusion.

*note on "Acompatibalist": As far as I know this isn't a word. The 'a' prefix can take many different meanings, including "toward", "not", "without", "in", "on". I don't know which of these he means, so if someone could fill me on a formal interpretation of this word that would be great.

Of course, Dan is not the first person to posit free will as an illusion. As I mentioned, the Behaviourists who dominated psychology in the early 20th century to the 1950's also viewed free will as an illusion. What Dan does, however, is present a coherent view of free will as a product of the judgement on behaviour. Free will is a social construct; It's not an intrinsic property, but rather an evaluation we judge on actions and behaviours. We perceive others of free will as we judge their behaviour, and we perceive ourselves of free will when we judge our own. Free will is a frame of reference.

The crux of his argument is what he calls harmonic free will. Free will may or may not exist at an individual level. You are the result of your past, predisposed to make the choices you make. But once you change perspective to consider behaviours between individuals in a social construct, free will arises. Within subjects, we may be determinstic (he calls this perspective melodic free will). But between subjects we begin to judge and evaluate each other's behaviours—even our own. This judgement, he argues, is where free will resides.

To be honest, I am not totally sold on his ideas and analogies. While use of analogies can be helpful to get across certain nuances in concepts, Dan's book relies heavily on this device throughout. The danger with analogies is that they can be misinterpreted without the reader realizing they have misinterpreted, and I certainly fell victim to this trap a few times. But his main point is this: Free will is an illusion that arises through behavioural judgement in a social context, everything social is an illusion anyways, and these are useful illusions to maintain our social existence.

Randomness and Free Will

But there is one quote in particular that stood out to me throughout Dan's book, and it is something I wanted to dive deeper into:

Free will gains nothing from randomness

This claim stood out to me as it hits closer to home with my own research pursuits. Here is a little background on myself; I am by no means a philosopher. My main practice is in neuroscience—specifically, computational neuroscience. I study how the nervous system perceives and processes information and work on models to explain this. More specifically, I'm exploring this within your sense of touch. A typical experiment might find me presenting some physical stimulus to you, either passively on your arm, or "actively" allowing you to pick something up yourself.

Anyways, I think there are two statements Dan is making with his quote:

Whether or not randomness exists doesn't matter because Free Will is something different.

As I said before, Dan is an acompatibalist*. He sees free will as a completely separate argument from determinism. Instead, randomness would play a role in the debate for indeterminism. There, randomness exists, then not everything in reality is deterministic as things could have happened "another way".

I see his point, but I still want to maintain that the presence of randomness has some relevance for free will. I don't deny his view of free will—of course it's an illusion. Any feeling or perception relating to free will is internally generated, just as everything else in your life. Some things we perceive have tangibility. Some things don't. But one part of what everybody will be referring to with "free will" is indeterminism, the ability for things to have happened differently. Indeterminism may not be necessary to justify free will, but it certainly helps validate it.

The findings of randomness (or more precisely, probabilistic stochastic processes) in quantum mechanics is not relevant to Free Will in our macro-scaled world.

I think I see where Dan is coming from by this statement. An important intuition that has come out of quantum mechanics in Physics is that reality at the subatomic level seems to operate probabilistically. Because of this, quantum mechanics is often used as a way to justify free will. Since our reality fundamentally allows for probabilistic randomness, that might allow for randomness—and thus free will—to exist at the macro scale. This, as Dan calls it, is the question of micro-determinism.

Dan makes an important distinction from micro-determinism and what he calls macro-determinism. This is the level of complex structures and organisms. The level of animals and brains. Through this illustration he makes the comparison between modern Quantum physics and classical Newtonian physics. At the subatomic level, interactions are probabilistic. But when you are dealing with macro-sized objects, like the motion of a car or the flow of a river, quantum mechanics becomes ill-suited to explain all of the properties. That randomness washes out, and we're left with predictable events. Randomness may exist in the quantum realm, but in our macro realm it is no longer relevant.

Trying to provide evidence for free will through the subatomic properties that make us up is a fools errand. You have to deal with the properties on the macro scale. Thus, free will gains nothing from randomness.

Image from the Human Connectome Project Gallery

But there are actually other ways that randomness can come into our biological systems which make it relevant again for free will at the individual level (melodic free will, as Dan put it). A lingering question in Cognitive Neuroscience is what decision strategy our nervous system employs when trying to decode what it perceived. In other words, how does your brain deal with uncertain information? The most intuitive strategy is a deterministic one: given a few possible perceptions, the brain may always choose the most likely hypothesis. This is certainly the prevailing strategy in any scenario; if there is truly a 70% chance of A and a 30% chance of B, you should always choose A and you will be right 70% of the time! But the other possible strategy is called Probability Matching (or sometimes Bayesian Sampling), which states that the observer in our example would respond A only 70% of the time. Most of the time they would choose the most probable option, but sometimes they would choose the alternative.

This theory has been difficult to prove definitely; it's likely our brains are capable of both strategies and switch between them under different contexts. But we have not been able to decisively shelf the idea either.

A Visual Example

For an illustration, let's consider the processing underlying vision. As it turns out, seeing is not as simple as recording an image on a sensor. Your visual system is a complex process of deconstruction, processing, and reconstruction starting at the photoreceptors in your eye and ending in the back of your brain—your visual cortex. By then, the information is generally thought to be intact in the mapping as how the photons fell on your retina. Researchers at UC Berkeley made a lot of headlines in 2011 when they were able to reconstruct images from brain activity in the visual cortex matched to sample video clips (Nishimoto et al., 2011).

Dr. Jack Gallant discussing his fMRI image recreation work

But before it gets to that point, the central visual pathway is a long process breaking down orientations, colours, patterns, and combining these between the visual fields of both your eyes. And the journey doesn't stop here; that information must be further processed to actually make sense of what you are seeing. What is it? Where is it in space? What parts of the scene belong to "it" and what is separate? Any visual illusion will immediately inform you that what you see is not necessarily reality. There are finely tuned processes underlying every part of your perception.

Being "finely tuned" does not mean being perfect. The nervous system is a surprisingly noisy place. Neurons don't operate in a binary active-inactive state. They are actually always active, sometimes firing randomly. What changes when a signal is added is that this firing rate often increases, coordinating with other neurons in bundles and projections, effectively becoming a coherent signal.

This summary is a gross simplification undercutting the beautiful complexities of the nervous systems and the dynamics that occur, but what I want to get across is that it is not "on or off". Even your photoreceptors don't behave in the way you might think. You may have thought that a cone or rod on your retina only send a signal once it receives a photon. Actually, the opposite is true. The default state of a photoreceptor is to release neurotransmitters at its maximum rate, and as it receives a photon that firing is inhibited.

The point I'm trying to make is that your nervous system is noisy. Your sensory measurement of the world is noisy, processing that information is noisy, and putting it back together in your consciousness (whatever that is) is noisy. And with noise, comes randomness. What this all boils down to is that it is theoretically possible to give the brain the same input and get different results. You may call our societal and moral rules, our character, and our internal processing biases deterministic as they restrict the possible options we can make. But if that hypothesis is true—that the brain can be given the same input and get different results—then we are at least capable of randomness at the macro level. We are capable of perceiving and deciding things differently.

Proving Illusions With Illusions

Now with all of this in mind, here is one of my favourite visual illusions—The Neckercube.

When you see this cube, which way do you see it facing? Is it popping out of the screen, or inset going into it? I find this illusion immensely fascinating because, for most people, it's both. The common experience is to jump back and forth between the possible perceptions. This is what is known as a bistable percept. Both possible ways to perceive the cube are just as likely, and so your brain randomly jumps between them. This simple illusion provides evidence—and hope—that we may sometimes employ this fabled Bayesian Sampling.

Unfortunately, this is by no means a "smoking gun" of evidence. Trying to test this idea with the illusion empirically quickly falls into an inescapable trap. In order to definitely say that we are not deterministic, we would need to be able to test someone at the same point in time multiple times and see if their responses can change. Each test would need to be absolutely independent; that means the observer needs to be in exactly the same state, with exactly the same experience, every measurement. This isn’t possible because time constantly moves forward. We can’t go back in time to redo a test, so each test we make has to occur later in time.

And here is our fatal flaw: these measurements are not independent samples. Each subsequent perception of the Neckercube is contaminated by each previous one. Our observer is no longer in exactly the same state. So you could argue that your current perception of the Neckercube is the deterministic result of your life long experience up to this point in time, including your time looking at and processing the Neckercube. We cant test between multiple observers at the same point in time because of course, each person has a different brain with different life experiences up to this point and deterministically could come to a different observation. This really is not a trivial experiment to design.

Truthfully, this debate in decision strategies has been going on for a while. Researchers Michael Kubovy and Amos Tversky made an excellent observation on this matter back in 1971 when trying to test between a deterministic and probabilistic decision model in a simple binary detection task. They were comparing against research by Wayne Lee (1964) who supposedly found evidence for what he called the micromatching hypothesis (essentially, probability matching):

Lee's [subjects] learned the distributions in the course of a relatively short experiment (400 or 500 trials). Under these circumstances, the location of the cutoff points probably shifted during the study. It is not surprising, therefore, that Lee's [subjects] produced data that are closer to the probabilistic model than the presented data.

The implication is this: a deterministic decision strategy with a shifting decision bound looks like probability matching. More simply, If there is variation in the information your brain is using to make its decision, its decisions won't always be accurate. And that behaviour can look an awful lot like probability matching. This variation could be due to internal noise, but it could also be due to differences in stimulus presentation, or in other external factors in the environment or the rest of your body constantly changing, or even just due to the changes in your brain after the previous experience. This is especially true in an experiment where the participant is still learning the categories they are deciding on.

A simple illustration of Kubovy and Tversky's (1971) observation using Signal Detection Theory. Imagine that you have two hypothetical groups of people, group A (blue) and group B (red). These distributions represent measured heights of the people in each group. Since the groups are set by the experimenters, they have a fixed average height and variance that shouldn't change (let's say the average height of group A is 5'6", and group B is 5'10"). The experiment is simple, you are given a height and need to categorize it, But as somebody trying to learn these two categories, you don't have access to those fixed population numbers. Instead, you are constantly trying to determine these underlying distributions every trial of the experiment, which is why they are shifting above. The yellow vertical bar represents the decision boundary for that trial; any time you get a height to the right of the bar, a deterministic strategy says you should always choose group B (red). Otherwise, choose A (blue). But since these distributions are shifting as you are learning, that boundary changes. The resulting behaviour, Kubovy and Tversky argue, can produce data closer to the probabilistic model.

But Bayesian Sampling is still being explored. There is a growing body of research slowly establishing that aspect of perception can be modeled and viewed as probabilistic inferences (Moreno-Bote, Knill & Pouget, 2011). Bayesian Inference—a probabilistic framework for comparing hypotheses given prior expectations and incoming data—is finding more and more success in explaining human behaviour and perceptual phenomenon (for cool examples, see Alais & Burr, 2004; Knill & Pouget, 2004; Körding et al., 2007). So I wouldn't rule out randomness at the macro-scale just yet.

Of course, maybe randomness still doesn't play a role in free will. Being capable of perceiving things differently is one thing; we don't necessarily choose how we perceive things, and we can only act on our end-result inferences. Are we actually just deterministic beings in an indeterministic world? (comically, this is the opposite of Descartes' argument)

I think at the very least, providing evidence of randomness in individual behaviour gets us pretty far in this 2,000 year old debate. There is still much of the brain we do not understand. Maybe there is still some network or process yet to find capable of directing neural activity or exerting control—will—over perception. Something to explain our source of agency.

But I don't know, this is just a random blog on the internet.

References

  • Alais, D., & Burr, D. (2004). The Ventriloquist Effect Results from Near-Optimal Bimodal Integration. Current Biology, 14(3), 257–262. https://doi.org/10.1016/j.cub.2004.01.029
  • Barker, D. (2018). Free Will Explained. Toronto, ON: Sterling Publishing.
  • Knill, D. C., & Pouget, A. (2004). The Bayesian brain: the role of uncertainty in neural coding and computation. Trends in Neurosciences, 27(12), 712–719. https://doi.org/10.1016/j.tins.2004.10.007
  • Körding, K. P., Beierholm, U., Ma, W. J., Quartz, S., Tenenbaum, J. B., & Shams, L. (2007). Causal Inference in Multisensory Perception. PLoS ONE, 2(9), e943. https://doi.org/10.1371/journal.pone.0000943  
  • Kubovy, M., Rapoport, A., & Tversky, A. (1971). Deterministic vs Probabilistic strategies in detection. Attention, Perception & Psychophysics, 9(5), 427–429.  
  • Lee, W., & Janke, M. (1964). Categorizing externally distributed stimulus samples for three continua. Journal of Experimental Psychology, 68(4), 376–382. https://doi.org/10.1037/h0042770
  • Moreno-Bote, R., Knill, D. C., & Pouget, A. (2011). Bayesian sampling in visual perception. Proceedings of the National Academy of Sciences, 108(30), 12491–12496. https://doi.org/10.1073/pnas.1101430108
  • Nishimoto, S., Vu, A. T., Naselaris, T., Benjamini, Y., Yu, B., & Gallant, J. L. (2011). Reconstructing visual experiences from brain activity evoked by natural movies. Current Biology, 21(19), 1641-1646.