Free Will
The author beneath the act — or the convincing illusion of one.
What we mean.
Free will is the capacity to have done otherwise; to originate an action that is neither random nor determined.
Compatibilists say it survives determinism if your action flows from your reasons. Libertarians demand genuine ontological freedom. Hard determinists deny it altogether.
In the brain
Libet's 1980s experiments showed brain activity preceding the conscious decision by ~300 ms — interpreted as evidence against free will, then heavily contested.
Sapolsky's 2023 'Determined' argues there is no free will and we must remake society accordingly. Most people disagree from the inside out.
In silicon
AI actions are determined by weights + inputs + sampled randomness. No model 'chooses' in the way a person chooses — though stochastic decoding gives an analog of freedom and a denial of determinism.
Agents with hierarchical goals approximate intentional action. Whether intention without phenomenology is intention at all is a live debate.
How we arrived here.
- 1781
Kant: free will as moral necessity
- 1983
Libet's readiness potential
- 2010
Compatibilism dominates academia
- 2023
Sapolsky: Determined
“Man can do what he wills, but he cannot will what he wills.”
Where the edge moves next.
When AI agents act with stable preferences across millions of decisions, the distinction between 'simulated' and 'real' agency may collapse into a matter of phrasing.
Where it touches the world.
Criminal justice reform.
Personal responsibility ethics.
Autonomous vehicles and liability.
Why it matters.
If freedom is the felt absence of constraint, perhaps both humans and AI are equally free — and equally unfree.