Does Free Will Still Exist If an AI Can Predict Your Next Decision?

A few months ago I read about a study in which a machine learning model predicted, with measurable accuracy, which political candidate a person would vote for — based not on their stated views but on their social media behaviour, purchasing patterns, and location data. The prediction was made before the person had consciously decided.

I sat with that for a while.

Not because it surprised me. The research had been pointing in this direction for years. Cambridge Analytica made this practically famous in 2016. What surprised me was my own reaction — a specific, unnameable discomfort that was not quite anger and not quite fear. Something closer to the feeling you get when a magician reveals how a trick works. The wonder does not disappear, but it changes.

The question underneath the discomfort: if my choices can be predicted before I make them, am I actually making them?

What Does Determinism Actually Say?

The debate between free will and determinism is one of the oldest in philosophy and one of the most resistant to resolution.

The determinist position is this: every event in the universe, including every thought and decision a human brain produces, is the result of prior causes. The physical state of your brain at any given moment is the result of everything that came before it — your genes, your experiences, your environment, the chemistry of your last meal. Given complete information about all prior causes, your next thought could theoretically be predicted.

The compatibilist response — developed by philosophers including David Hume and, more recently, Daniel Dennett — is that this does not actually threaten meaningful free will. What we mean by free will is not freedom from causation but freedom from coercion. You are free when your actions flow from your own desires, values, and reasoning, even if those desires were themselves caused by prior events.

This debate has been mostly theoretical. AI makes it practical.

Does AI Prediction Actually Prove Determinism?

Here is what I think is actually happening, and why the determinist conclusion does not follow as cleanly as it first appears.

When an AI predicts your behaviour, it is not reading your internal state. It is finding patterns in your observable behaviour that correlate with outcomes. The prediction does not tell you what is happening inside your mind. It tells you that your observable behaviour is patterned — which you already knew, because you are a consistent person with preferences and habits.

The prediction is probabilistic, not certain. It is right more often than chance, which is interesting, but it is not a direct window into the mechanism of your decision-making. There is a meaningful difference between “this person usually votes this way given these conditions” and “this person’s brain state at t=0 will produce vote X at t=1 regardless of any other input.”

Neuroscientist Robert Sapolsky, in his 2023 book Determined, argues that free will is genuinely illusory — that the biological machinery producing decisions is not under voluntary control in any meaningful sense. This is the most honest version of the hard determinist position I have encountered. I do not fully accept it, but I take it seriously.

The Part That Troubles Me More

I want to separate two questions, because they are often conflated.

Question one: Do humans have libertarian free will — the ability to have made genuinely different choices given identical prior conditions?

This is probably no. Physics has not found a mechanism for it. Quantum indeterminacy introduces randomness but not agency. The brain does not appear to have a component that is exempt from physical causation.

Question two: Does anything important change when AI systems can predict and therefore influence your behaviour?

This is yes, and this is the one that actually matters.

The problem is not that AI can predict your decisions. The problem is that prediction enables manipulation at scale. Recommendation algorithms do not just observe your behaviour — they shape it, by controlling what you see, when you see it, and what emotional state you are in when you decide. The system is not a passive predictor. It is an active participant in the causal chain that produces your choices.

This is the meaningful threat to agency, and it is not a philosophical abstraction. It is a design decision being made by engineers and product managers right now.

What I Think

I think the free will debate, in its classical form, is less important than it first appears. Whether or not your choices are ultimately caused by prior physical events does not change the phenomenology of choosing. You still deliberate. The deliberation still changes outcomes. The outcomes still reflect your values, even if those values were themselves produced by prior causes.

But I think the AI prediction problem surfaces something real and urgent that the classical debate does not: the difference between your choices being the causal result of your history and your choices being the engineered result of a system designed to produce specific outcomes in you for someone else’s benefit.

The first is fate, in a sense. The second is manipulation. These are not the same thing, even if they look similar from the outside.

What I want — what I think any honest account of human freedom requires — is to be the author of my own causes. Not free from causation, but free from having my causation directed by systems I did not choose and cannot see.

That freedom is worth protecting. Not because it is metaphysically robust, but because its absence makes us into instruments rather than people.