The Skeptics Society & Skeptic magazine

Testing Determinism

In a recent article in Skeptic, (“Misunderstanding Free Will (Which We Don’t Have).” 26.4 (2021): 54-56), David Reeve and Dennis Middlebrooks exchanged opposing views on the issue of free will. Reeve believes free will does not exist, and Middlebrooks believes it does. I favor Reeve’s position.

There has been quite a bit of confusion in the discussion of free will over the centuries, and this is due primarily to semantics, specifically the lack of clear definitions. Hoping to somewhat remedy this problem, I offer the following:

Determinism is the view that all our thoughts, decisions, choices, and behaviors are determined by prior factors and that given the recurrence of the same prior factors we would behave the same way; we could not behave otherwise. Prior factors in this context include things such as genetics, environment, prior experiences, childhood rearing, and their interaction.

Free will, sometimes called libertarian free will, is the view that all our thoughts, decisions, choices, and behaviors are caused by the self, which may be influenced by, but not determined by, prior factors, and that given the recurrence of the same prior factors we might not behave the same; we could behave otherwise.

Legal free will is the view that in certain circumstances persons make decisions that are not affected by immaturity, neuronal or mental defect, or coercion. So, an adult person might say “I signed this contract by my own free will.” Legal free will is compatible with determinism since immaturity, defect, and coercion are just particular kinds of prior factors which may be relevant in some situations but not in others. In the absence of these factors adults may make decisions considered to be legally binding, e.g., contracts.

Compatibilism is the view that some version of free will is compatible with determinism. Libertarian free will is not compatible with determinism, but legal free will is. The validity of compatibilism depends wholly on the definition of the particular version of free will on offer.

In the 2021 book Just Deserts, Daniel Dennett has described a version of free will “worth having,” which I will call morally responsible free will and which he claims is compatible with determinism. I agree. Dennett thinks that usually when we behave immorally or harm others, then we should be punished even if our behavior was determined. Other authors, e.g., Michael Shermer in his 2020 book Giving the Devil His Due, and Dan Barker in his 2018 book Free Will Explained, write of “degrees of freedom” in decision making, but as far as I can tell this idea just refers to the degree of complexity in our decision making. Some of our decisions are more complex than others, and decision making by humans is more complex than that of other animals.

In discussing free will, we may simply consider “will” to be equivalent to choosing. I think it is likely that all decision making can be reduced to choosing between two options, X or Y, and that when there appear to be more options, we only compare two at a time. Taken literally, “free” means unconstrained, untethered, unaffected, or independent. I doubt that anyone believes that the typical choosing between options is free in any absolute sense. Everyone knows on some level that their choices are at least influenced by factors out of their control, and so these choices cannot be free in an absolute way. Libertarian free will is simply not a legitimate view for most philosophers, given how much science has shown us about the various determinants of or influencers on our decisions.

The idea of determinism in human decision making, which I favor, is rather simple—human choices are just another part of the cause-effect world in which we live, and they can be fruitfully investigated through science. Libertarian free will, however, is a different matter. It may seem like we have libertarian free will. For example, we might think “I’m going to go to the store” and then we find ourselves in the store. We think that the thought caused the behavior. But this is probably due to a thinking error known as “post hoc, ergo propter hoc.” We are prone to think that if event Y follows event X, then X must be the cause of Y. Yet, there may be some other factor Z which caused both X and Y, and this is what is probably going on in our common experience. A brain event Z is probably causing both X (the intention to go to the store) and Y (actually going to the store). And thus, we experience what some call the illusion of free will.

I think one common mistake is to confound determinism with predictability when they really aren’t the same thing. Even if the choices of a subject are determined, we may never be able to perfectly predict those choices. We might lack knowledge of context, the inputs into the decision, the algorithm of the choosing process, and/or the past behavior of the subject. In other words, we might not know the prior conditions or the relevant laws of nature.

How could we go about making predictions of choices by a person? What information might we use to make accurate predictions? I see three possibilities:

  1. We could monitor states of the brain and determine which states are highly correlated with options subsequently selected by a subject. This approach has already been used with some success when brains were electronically monitored and choice between two button presses was predicted.
  2. We could collect data about many prior conditions and determine which of these are highly correlated with options chosen by a subject.
  3. We could collect information about a person’s prior choices to help determine his future choices. This reminds me of my graduate school training in psychology in the 1970s when we were taught the maxim “The best predictor of future behavior is past behavior.” All of these approaches have potential and could possibly be combined.

I don’t ride motorcycles. I don’t even like them. However, my son-in-law, whom I shall call “KP,” is a motorcycle enthusiast. For the sake of argument, let’s also suppose that he supports libertarian free will over determinism. Let’s try to imagine an experiment with results that would lead him to change his position. I asked KP to imagine this scenario:

You won a lottery. The prize is a really nice motorcycle, valued at $30,000. You can select it yourself—the only catch is that you cannot test ride it. However, if you don’t like it within the first 30 days, you can exchange it (but only once). You will express your preferences for the prize motorcycle through a process of comparing motorcycles, two at a time, for 200 pairs. For each pair, you will be shown a photo and a description that specifies information on six objective factors, and you will express your preference for one of the two in each pair within 30 seconds. So, in your mind, what would be the six most important factors relevant to your choice of the prize motorcycle?

After about 30 minutes of explaining and clarifying with me, KP specified these six factors:

  1. Engine Configuration with nine different categories (Parallel Twin, Parallel Triple, Parallel 4, Parallel 6, V-Twin, V-4, Boxer Twin, Boxer 4, and Boxer 6).
  2. Engine Power in units of horsepower.
  3. Torque in units of foot pounds.
  4. Riding Style with five categories (Cruiser, Adventure, Standard, Sport, and Touring).
  5. Luggage Capacity in units of liters.
  6. Range at Full Capacity in units of miles.

In our hypothetical research, the experimenter would assemble a sample of 150 different motorcycles with pictures and descriptions of each on the six factors. In a series of 200 trials KP would choose which of two motorcycles he would prefer to receive as a prize, each trial lasting only 30 seconds. For each trial, the motorcycles for comparison in pairs would be chosen at random from the sample of 150, with replacement after each trial.

The experimenter would utilize a special computer enabling machine learning through Artificial Intelligence (AI). For the “learning phase” the computer would be given the input of pictures and descriptions of the two compared motorcycles, the same as presented to KP, and also the output of KP’s choice for the first 180 trials. For the “prediction phase,” consisting of the final 20 trials, the computer would still be given the input of pictures and descriptions of compared cycles as given to KP, but the computer would predict what KP’s choices would be! An impressive rate of correct predictions, pointing in favor of determinism, might be 90 percent plus. Before revealing the predictions to KP, he would be asked and would presumably agree that his will was free in making all his choices. Of course, the greater the number of learning trials for the AI, the better it would become at prediction. The computer learns to simulate the decision making process of the subject and thereby make accurate predictions of choices. KP would be likely to give up his belief in free will if the prediction accuracy rate were quite high.

According to my model of determinism, KP has some Neural Choice Mechanism (NCM) in his brain that captures relevant information from all his prior conditions and which leads to values associated with the different categories or states on the six different factors pertinent to each motorcycle. Obviously, all this information will have come from reading about motorcycles, talking with others about them, watching them, tinkering with them, and especially riding them for thousands of miles over many years. The information constitutes KP’s competence and value basis, which enables him to make his 200 binary choices in the hypothetical experiment. By some implicit algorithm in KP’s NCM, the values on the different factors are combined to get an overall value for each motorcycle, and then KP naturally selects the motorcycle in the pair with the highest overall value. What the AI does is simulate his brain, allowing future choices (the last 20) to be predicted from past choices (the first 180). In this experiment the researcher would be using past behavior to predict future behavior (approach #3).

This type of experiment could be repeated with different subjects and different kinds of choices between two candidate items, such as jobs, employees, relationship partners, cars, houses, colleges and universities, clothing, and perhaps even worldviews, etc. High rates of accurate prediction would not only support determinism but might persuade many advocates of libertarian free will to change their minds.

We still have much to learn about human decision making, but we now have methods and tools to help us get answers, moving the controversy a little more from philosophy to science. END

About the Author

Gary J. Whittenberger Ph.D. is a freelance writer and retired psychologist, now living in North Hollywood, California. He was formerly a leader in many freethought groups in Tallahassee, Florida. He received his doctoral degree from Florida State University after which he worked for 23 years as a psychologist in prisons. He has written many published articles on science, philosophy, psychology, and religion. He is the author of two books: God Wants You to be an Atheist: The Startling Conclusion from a Rational Analysis, and God and Natural Disasters: A Debate Between an Atheist and a Christian.

This article was published on January 18, 2023.

Skeptic Magazine App on iPhone


Whether at home or on the go, the SKEPTIC App is the easiest way to read your favorite articles. Within the app, users can purchase the current issue and back issues. Download the app today and get a 30-day free trial subscription.

Download the Skeptic Magazine App for iOS, available on the App Store
Download the Skeptic Magazine App for Android, available on Google Play
Download the Skeptic Magazine App for iOS, available on the App Store
Download the Skeptic Magazine App for Android, available on Google Play
SKEPTIC • 3938 State St., Suite 101, Santa Barbara, CA, 93105-3114 • 1-805-576-9396 • Copyright © 1992–2023. All rights reserved • Privacy Policy