Free Will, Sapolsky, and AI: Is everything predetermined?

First, we must define what to call freedom

The phrase "I chose" can mean different things. In the strongest, libertarian sense, a person could have truly chosen otherwise under exactly the same conditions, and the source of choice does not reduce to prior causes. Robert Sapolsky's argument is directed precisely against such independent authorship.

Another definition is proposed by compatibilists. For them, a free action is not a magical break in causality, but the ability to act according to one's own motives, deliberations, and character without external coercion. If I have considered the consequences and done what I wanted, they can call the action free, even if my desires have a causal history.

Therefore, the dispute partially concerns not only facts about the brain, but also what kind of control we demand from the concept of freedom. The overview Stanford Encyclopedia of Philosophy directly treats this as a combination of empirical, metaphysical, and ethical problems.

Sapolsky's argument: find the moment where an independent author appeared

Sapolsky proposes tracing any action back in time. A second before it, certain neural networks were at work. Minutes before—fatigue, fear, hunger, hormones, and context were influencing it. Years before—experience, trauma, learning, and culture. Even earlier—prenatal development, heredity, and childhood conditions. A person did not choose any of these factors before they began shaping their subsequent choices.

His main thesis is not that there is a single "crime gene" or a single brain region that controls us. On the contrary: behavior is the result of a vast network of causes at different scales. If you remove them one by one, Sapolsky finds no separate, causeless "I" that could independently trigger a different outcome.

What the famous brain experiments actually showed

In Benjamin Libet's classic experiments, participants moved their wrists voluntarily and reported the moment they felt the intention to move. Electrical activity, called the readiness potential, began before the reported conscious awareness of the intention. This gave rise to a strong interpretation: the brain "decided" before the conscious "I" became aware of the decision.

In 2008, Chun Siong Sun, John-Dylan Haynes, and colleagues used fMRI to identify patterns associated with the future pressing of a left or right button, even before participants reported a conscious choice. The authors wrote that the outcome of a simple action was encoded in the activity of the frontal and parietal regions before the decision entered consciousness. At the same time, this was a probabilistic prediction of a simple, intentionally meaningless choice, rather than an accurate reading of complex life decisions. Original article in Nature Neuroscience.

From this follows an important but limited conclusion: part of the preparation for an action occurs unconsciously. This in itself does not show that conscious deliberation never has a causal role. "Unconscious" also does not mean "not mine": these processes occur in the same brain that holds our goals, memory, and character.

Why Libet's experiments did not close the question

Later work by Aaron Schurger and colleagues offered a different explanation for the readiness potential. The average wave before movement may arise not as a record of a decision already made, but as an accumulation of neural activity fluctuations up to a threshold, after which spontaneous action occurs. That is, the signal may reflect the preparation process and the moment of initiation, rather than a secret, ready-made command. Model in PNAS.

Another limitation is the nature of the task. Choosing to press one of two buttons without reason differs from deciding to leave, forgive someone, or risk one's life. In the study by Uri Maoz and colleagues, the classic readiness potential appeared for arbitrary choices, but not in the same way for deliberately considered decisions with consequences. Study in eLife.

An academic review of the problem summarizes cautiously: the interpretation of the readiness potential has faced serious criticism, and neuroscientists do not agree on what exactly it means. Therefore, the phrase "the brain decides for us" is too crude. We are the organism with the brain; the real question is whether the functioning of this system corresponds to a sufficient notion of authorship and control.

If everything is physics, is everything predetermined?

Physicalism—the view that mental phenomena depend on physical processes—is not identical to strict determinism. Modern physics contains fundamentally probabilistic descriptions. Even in a fully physical world, it is not necessarily possible to unambiguously calculate every event in advance.

But randomness does not automatically rescue strong free will. If a neuron fires due to an indeterminate fluctuation, it does not mean that "I" controlled the event. The alternative of "determined by prior conditions or occurred by chance" leaves the problem of independent authorship unresolved in both cases.

Furthermore, determinism does not equal practical predictability. A complex system can be extremely sensitive to initial conditions, and we may lack the data and computing power for its accurate prediction. Even if behavior has causes, it does not follow that someone is capable of composing a ready-made script of your life.

The strongest objection: freedom does not have to be causeless

A compatibilist would agree with Sapolsky that a person did not create their own genes, childhood, or basic brain structure. But they would ask: why must freedom be independent of character and causes? An action without causes would be no more "mine," but simply random.

In such an approach, what matters is whether a person is able to:

• understand causes and consequences;
• respond to arguments and new information;
• plan and control impulses;
• act without threat, coercion, or severe impairment of the ability to control;
• change future behavior after learning.

This is not a peripheral position. In PhilPapers Survey 2020 about 59% of surveyed philosophers accepted or leaned toward compatibilism, while approximately 11% leaned toward the answer "free will does not exist." The survey does not establish truth through voting, but it shows: Sapolsky's conclusion is not a settled scientific consensus.

How this relates to artificial intelligence

A modern language model generates a response based on training, current context, system instructions, parameters, and the procedure for selecting the next text elements. If these causes are changed, the result changes. In this general sense, both humans and AI are systems whose behavior has prior conditions.

But this similarity does not yet make them the same.

An interdisciplinary report on AI consciousness, built on several scientific theories of consciousness, found no grounds to consider current systems conscious, although the authors see no obvious technical impossibility of creating systems with corresponding indicators in the future. Butlin et al., 2023. This is a more cautious conclusion than "AI will definitely never feel anything" or "the chatbot has already woken up."

The most interesting question, therefore, is not "which of them is programmed?". In a certain broad sense, both are causally formed. The question is different: what internal organization, experience, ability to understand causes, and stable personal goals must a system have for us to recognize it as an agent?

If Sapolsky is right, then why do anything?

The absence of an "uncaused author" does not mean that efforts are futile. Conversation, education, therapy, law, reward, and punishment are themselves causes that change future behavior. The thought "my decisions have causes" does not remove you from the causal chain — your reflections are part of it.

Something else changes radically: the attitude toward merit and blame. If no one chose the starting conditions of their own brain, there is less ground for pure vengeance and contempt. But there remain reasons to protect others, isolate a dangerous person, repair harm, and create conditions that reduce the recurrence of crime.

For the AI era, this is practically important. Algorithms evaluate people, predict risks, and nudge them toward decisions. If behavior depends heavily on the environment, whoever designs the digital environment gains real power over future actions. The discussion about free will moves from the philosopher's office to the design of recommendations, advertising, education, labor, and state systems.

An honest conclusion without convenient sensation

Sapolsky's position is particularly strong against the image of a tiny, independent "manager" inside the head, standing apart from biology and causality. Science has indeed not shown such a manager. However, a compatibilist might respond that there is no need to look for one: a person is the entire system capable of reasoning, learning, and acting according to causes.

Therefore, the most accurate formulation today is: we have increasing evidence of the causal determinism of human choice, but the implications of this evidence for the concept of freedom depend on which freedom we consider important. AI makes this debate sharper because, for the first time, complex systems have appeared alongside us that generate convincing answers without confirmed internal experience.