The phrase free will is carrying several ideas at once. That makes it easy for two people to argue for an hour while defending different things.
Table of contents
On this page
Several Kinds of Freedom
Start with ordinary choice. I mean the familiar process of noticing options, weighing reasons, forming an intention, and acting. A choice can be thoughtful or impulsive, wise or dreadful. Calling something a choice says that the action came through the person’s decision-making capacities. It does not yet say whether those capacities were determined, indeterministic, or self-created.
Imagine I have promised to help a friend move on Saturday. My alarm rings. It is raining, I slept badly, and staying in bed is enormously attractive. I remember the promise, consider sending an apology, picture my friend carrying boxes alone, and get up.
That is a choice in the ordinary sense. The fact that my reasons, character, and tiredness all have causes does not turn the scene into a reflex.
Freedom from coercion adds something practical. Nobody threatened me, physically forced me, or bypassed my deliberation. Compare that with handing over my wallet to someone holding a knife. I may still select one option over another, but the threat radically narrows the acceptable options. Freedom from coercion concerns how external pressure bears on my decision. It does not answer whether the whole process has prior causes.
Then there is leeway, often expressed as the ability to have done otherwise. This is control through genuinely available alternatives. The phrase is slippery. On a strong reading, with the complete past and laws of nature held fixed, I could still have stayed in bed. On a weaker, dispositional reading, I could have stayed in bed if my judgement or intention had differed, and my ordinary capacities would have carried that intention into action.
Those are not the same claim. The philosophical map of free will distinguishes this freedom to do otherwise from another family of ideas about control.
That other family is sourcehood. An action has sourcehood when it issues from the agent in the right way, rather than from a seizure, a threat, hypnosis, or direct manipulation. Sourcehood need not mean that the agent is uncaused. Many compatibilists say my action genuinely comes from me when it comes through my values, reasons, and deliberation, even though I have a causal history.
Ultimate sourcehood asks for more. It asks whether I am the origin of the relevant kind of person, values, and decision-making machinery, not merely the place where they now operate. This extra word is where much of my doubt lives.
Three Serious Positions
Libertarianism, in this debate, has nothing to do with the political label. It combines two claims: free will is incompatible with determinism, and at least some people really have free will.
Libertarians disagree about what makes this possible. Some place indeterminacy within deliberation. Some argue that agents, rather than only events, can cause actions. Some emphasise rare self-forming choices that help build character. The developed incompatibilist theories are attempts to secure both openness and control.
It would be unfair to answer libertarianism with “that is just randomness”. Serious libertarians know that randomness alone does not provide authorship. Their challenge is to explain how an undetermined outcome is settled by the agent rather than merely happening to the agent. I am not convinced that this challenge has been met, but it is the challenge they are trying to meet.
Compatibilism is the claim that free will can exist even if determinism is true. It does not, by itself, say that determinism is true. Nor does it reduce every free act to doing whatever desire happens to be strongest.
Contemporary compatibilist accounts ask whether action comes through capacities such as understanding reasons, revising intentions, controlling behaviour, and identifying with one’s motives. Some defend a compatible kind of leeway. Others say moral responsibility needs guidance over the path actually taken, not access to another path.
Hard incompatibilism says that the control required for basic moral desert is unavailable whether the world is deterministic or indeterministic. Determination leaves the source outside our ultimate control. Mere indeterminacy introduces luck.
This is broader than hard determinism because it does not need a verdict on the physics. It is also narrower than the bleak claim that nobody chooses, reasons, changes, or matters. A moral responsibility sceptic can retain all of those things while rejecting a particular desert-based kind of responsibility.
That is closest to where I currently stand.
What Frankfurt Cases Change
The most famous pressure on leeway comes from Harry Frankfurt’s 1969 paper.
In the familiar structure of a Frankfurt case, Jones is deciding how to act. Black wants Jones to choose one particular option and has a way to force that choice if Jones begins to go the other way. Jones chooses Black’s preferred option for his own reasons, so Black never intervenes.
Jones appears unable to do otherwise. Yet the thing removing his alternative played no role in the choice he actually made. Adding an idle controller to the story does not seem to change the quality of Jones’s actual deliberation.
The intended target is the Principle of Alternative Possibilities, which says that a person is morally responsible only if they could have done otherwise. Frankfurt cases suggest that what explains an action in the actual sequence may matter more than an unused alternative.
They are not a settled result. Critics argue that Jones retains some small “flicker” of freedom before intervention, or that making Black’s prediction reliable quietly assumes the very determinism an incompatibilist rejects. The detailed debate over Frankfurt-style cases remains open.
My takeaway is modest. Leeway and sourcehood can pull apart. Making an act unavoidable is not necessarily the same as producing it. But even if Frankfurt defeats the claim that alternatives are always necessary, it does not show that Jones created the motives from which he acted. It shifts the argument towards sourcehood. It does not settle how deep that source must go.
The Best Case for Compatibilism
The strongest compatibilist reply is that ultimate self-creation was never a sensible condition of freedom.
A cause can be mine while also having causes. When I get out of bed to help my friend, the promise matters because I can understand it as a reason. My concern for my friend matters because it can guide my behaviour. If the situation changed, perhaps because I was ill or my friend no longer needed help, my deliberation could produce a different action.
This kind of reasons-responsiveness is not magical, but it is substantial. Influential accounts ask whether the mechanism producing an action can recognise and react to a suitable range of reasons, and whether the agent owns that mechanism. The reasons-responsive approach distinguishes deliberate action from hypnosis, irresistible compulsion, and a movement caused by someone pushing my arm.
Those distinctions track real capacities. We teach a person who can understand reasons. We negotiate with someone who can revise a plan. We excuse a seizure and challenge a calculated lie. We ask more of a rested adult than of a terrified child. None of this becomes pointless if the capacities have causes.
Compatibilists can also say that moral responsibility belongs to human relationships, not to some view from outside the universe. Promises, apologies, trust, resentment, and forgiveness depend on the quality of a person’s will towards others. Requiring people to have authored themselves from nothing may discard the only freedom social creatures could need.
That is a powerful view. It preserves agency without hiding an uncaused soul inside the explanation. It also avoids making responsibility depend on a random break in causation. If free will means mature, uncoerced, reasons-responsive self-government, I have little interest in denying that we sometimes have it.
My reservation is about what follows from it.
The Source Behind the Source
Galen Strawson’s Basic Argument presses exactly here.
When I act for a reason, what I do depends on how I am mentally. It depends on my values, priorities, habits of attention, and capacity for self-control. If I am to be ultimately responsible for the action, Strawson argues, I must be ultimately responsible for those relevant features of myself.
Perhaps I deliberately change them. I practise patience, question a prejudice, or build a habit of keeping promises. That self-shaping is real. But the decision to reshape myself is made by an already existing me, using values and capacities I did not first choose. To be ultimately responsible for those starting principles, I would need to have chosen them using still earlier principles. The explanation retreats again.
This is not an argument that people cannot change themselves. We plainly do. It is an argument that self-change never becomes self-creation. At every stage, the person doing the shaping already has some way of judging what is worth shaping and some capacity to act on that judgement.
A compatibilist can reject the step from responsibility for an action to responsibility for the relevant features of oneself. Perhaps it is enough that the action expresses my values and responds to reasons now. I did it knowingly, without coercion, through capacities that are genuinely mine. Demanding that I also made myself may simply ask for the wrong thing.
That identifies the real disagreement. For everyday answerability, the compatibilist standard may be enough. For ultimate desert, I do not think the extra demand is arbitrary. If I deserve suffering or praise simply because of the kind of will I exercised, it matters that I never chose the foundations of that will.
The Basic Argument may not force someone with a less ambitious concept of responsibility to change vocabulary. It does explain why that vocabulary does not give me the deeper desert I am asking about.
Keeping Agency, Losing the Cosmic Ledger
By basic desert, I mean deserving blame or praise just because one knowingly performed the act, independently of whether the response protects anyone, repairs harm, teaches, deters, or improves a relationship. That is the sense isolated in the literature on moral responsibility scepticism. It is this sense, especially when used to justify suffering, that I doubt.
I still want the language of agency. An agent can form intentions, act for reasons, learn, plan, and affect what happens.
I also want the language of autonomy, understood as self-government that comes in degrees. Coercion, manipulation, addiction, fear, exhaustion, knowledge, and social power can all affect how autonomous a choice is. There is no need for autonomy to mean independence from every cause. Philosophical accounts of personal autonomy largely try to distinguish influences that participate in self-government from those that undermine it.
These concepts do practical work. Saying “I chose to help” identifies where deliberation operated and where future reasons can have an effect. It does not say I invented my temperament, concern for promises, ability to imagine another person’s burden, or available options.
Accountability survives too. We can name a wrong, ask for reasons, require repair, set boundaries, withdraw trust, and prevent further harm. We can praise generosity because gratitude is fitting, recognition matters, and examples shape people. What I lose is the claim that anyone earns pain for its own sake, or that fortunate character makes its owner ultimately superior to someone shaped less fortunately.
So I can accept much of what compatibilists call freedom. I just do not see the bridge from useful self-government to ultimate moral desert. A reasons-responsive person is a remarkably capable source of action. They are still a source with sources.
The next question is what those sources are. Part 3, Every Choice Has a History, will follow a decision backwards through the person, the situation, and the world that made both possible.