The Concept of Free Will

In an argument a few years ago I said something I had no intention of saying. Not a slip of the tongue - a whole sentence, aimed carefully, and chosen for damage. I heard it on the way out. By the time I knew it was coming it was already in the room.

I apologised, and the apology was sincere, and underneath it was a question I could not get rid of. Where exactly was I when that was decided?

The Experiment Everyone Half Remembers

You have probably heard some version of this. People were wired up, told to flick a finger whenever they felt like it, and asked to note the instant they became aware of deciding. Their brains showed a build-up of activity before that reported moment of awareness. Depending on who is telling it, the gap is half a second, or seven seconds, and the conclusion is always the same: the decision was made before you knew, so you are a passenger.

It is a genuinely important result and it has been oversold to an almost comic degree, so here is the less exciting version.

• The rising signal was measured by averaging many trials, which means it describes a tendency across attempts and not a decision on any given one. A strong competing account says it is not a decision at all, just ordinary background fluctuation drifting until it happens to cross a threshold - and what gets averaged together afterwards looks like a slow build-up that was never there in a single instance.
• The later brain-imaging studies that predicted a choice several seconds early did so at around sixty percent accuracy. That is above chance and well worth studying. It is also nowhere near the fortune-telling the headlines described.
• The reported moment of awareness comes from someone watching a fast clock and remembering where the hand was. Timing your own thoughts to a tenth of a second is exactly the kind of thing minds are bad at.

None of that rescues free will. It just means the case was never as closed as the summary suggests.

What Was Actually Measured

Here is the part that gets skipped every time, and it is the part that matters most.

The decision being studied is: flick this finger, at no particular moment, for no reason whatsoever. That is a task designed to have nothing behind it. Nobody has ever put a person in a scanner while they worked out whether to leave a relationship, take the job in another country, or tell the truth about something expensive.

Those decisions run on a completely different timescale. They involve going quiet for a week, talking to three people, changing your mind twice, sleeping badly, and noticing one morning that it has settled. Whatever is happening there, it is not a twitch with a readiness signal in front of it.

We measured an arbitrary finger movement and announced a result about the human will. The gap between those two things is the entire question, and it is the one part nobody tested.

The Question Has a Trap In It

Set the experiments aside, because there is a problem in the question that no amount of data will fix. Ask what a free decision would even look like if you got one.

If the choice comes out of your history, your values, your mood, what you ate, who raised you and what you have read - then it was caused, and the causes were in place before you noticed. That is the version people find threatening.

So try the alternative. A decision that does not come from any of that. Uncaused. Nothing about you determines it. Notice that this is not freedom, it is a dice roll, and a dice roll wearing your face is not more yours than a considered choice. It is less.

A decision that isn't caused by who you are isn't free. It is just noise with your name on it.

There is no third thing. We imagine one - a self that stands outside the chain of causes and reaches in - but that is the same character nobody can find when they go looking for who is behind the eyes. The free will being disproven by the experiments is a version that never made sense to begin with.

The Version Worth Keeping

Drop the question of whether the universe could have gone differently, which you will never settle and which changes nothing about your afternoon. Ask a smaller one instead: was that decision made by me, or was it made at me?

That distinction is not academic. It is the difference between signing a contract and being held down while someone moves your hand. Between choosing to stay in a job and not being able to leave one. Between taking the drink and being unable not to. Everybody already knows this difference, and every legal system on earth is built on it.

So here is a test that actually does something. Can reasons move you? If someone shows you a fact and your behaviour changes, then you are the kind of system that deliberation works on. That is what being a person who decides consists of. Not escaping causation - being the particular place where reasons get to be causes.

Which also explains why arguing with people is not pointless, and why it feels pointless with a certain kind of person. You are testing whether reasons reach them.

Where the Leverage Actually Is

If there is a practical lesson in the experiments, it is not that choice is fake. It is that the moment of choice is a bad place to spend your effort.

By the time the impulse arrives, most of the work is done. Willpower at that instant is a thin last line, and everybody who has ever said one more episode at one in the morning knows exactly how thin. The leverage is earlier, in the conditions that produce the impulse:

• Where the phone physically is when you sleep. Not a metaphor. A room away is a different decision from an arm's length away.
• What is in the house. Nobody has ever eaten the thing that was not there at midnight.
• What is set up in advance. Clothes out, bag packed, first sentence of tomorrow's work already written. You are handing the next version of you a decision that has already been half made.
• Who you spend your week with. Most of what you will want next month is currently being decided by whoever is around you now.
• Sleep, food, daylight. Almost every failure of self-control I have ever had was downstream of being tired, and dressing that up as a character flaw only made it harder to fix.

None of this requires a ghost in the machine. It only requires that earlier states of you shape later ones, which is precisely what the deterministic picture says. Determinism is not the enemy of self-improvement. It is the mechanism that makes it possible - if nothing you did today affected tomorrow, there would be no point doing anything.

Then Nothing Is Anyone's Fault?

This is where the idea usually gets abused, so it is worth being direct about it.

If nobody could have done otherwise in some cosmic sense, the appetite for punishment as pure payback does get harder to justify. Fine. But responsibility is not only about payback. It is also forward-looking: this person did the thing, this person can be reached by consequences and conversation, and that is exactly why we hold them to it. Holding someone responsible is one of the causes you apply.

There is also some evidence, argued over since it first appeared, that people who have just been persuaded they have no free will behave slightly worse in the next ten minutes. Take the size of that effect with caution. Take the direction seriously. A belief about agency is itself one of the inputs, which is a strange loop and a good reason not to adopt this idea as a posture.

Back to the Sentence

So: did I choose those words in the argument? In the mystical sense, probably not. Something assembled them out of a bad week, an old grievance and whatever I was like at twenty-three, and handed them to my mouth a fraction before it handed them to me.

The apology stands anyway, and not as a technicality. The process that produced the sentence is the same one that has to sit in the room afterwards, and it is a process that learns. I have not said that particular thing again. Something in there was reachable.

That is the whole of it, as far as I can tell. Not an author, free of causes, issuing commands from outside the world. Just a system complicated enough that reasons can get in and change what comes out next - which is smaller than what we wanted, and considerably more useful.