Ulysses Pact

A simple explanation

Ulysses wanted to hear the sirens. He also knew that, hearing them, he would steer the ship into the rocks. So he asked his crew to tie him to the mast and to refuse — under any later order from him — to untie him. He heard the song. The ship sailed past.

The move is not willpower. The move is removing the lever from the hand that will, in the hot moment, reach for it.

This is a Ulysses pact: a pre-commitment that the future self cannot undo. The cold-state self binds the hot-state self by making the feared action structurally unreachable, not merely discouraged. The pact survives the moment of wanting because there is nothing for the wanting to act on.

An everyday example

You know that, after the third drink, you will text the ex. You have known this for two years. You have promised yourself, in the cold state, that you will not. The promise has lost.

Today, before going out, you hand your phone to a friend and ask them not to give it back until tomorrow morning, no matter what you say. You go out. You have the three drinks. The wanting arrives on schedule. There is no phone. The wanting passes.

The friend with the phone is the mast. The version of you that would have texted at midnight does not have the lever.

What is a Ulysses pact?

It is the structural-impossibility class of pre-commitment. The cold-state self anticipates that the hot-state self will act against the cold-state self's values, and removes the means rather than relying on the will. The pact's defining property is that the bound self cannot, by itself, undo the binding.

This is Jon Elster's distinction, made precise in Ulysses and the Sirens and refined in Ulysses Unbound. Elster separates pre-commitment in general from the strong form — the form that holds even against the considered later wish of the binding party. The strong form is the Ulysses pact.

How is a Ulysses pact different from a regular pre-commitment device?

A pre-commitment device, broadly, raises the cost of the undesired action. A Ulysses pact removes the action from the option set entirely, for a defined period.

The difference is that a revocable device works only as long as the cold-state valuation is louder than the hot-state pull. When the pull crosses a threshold, the device gets revoked. The promise to oneself, the sticky-note on the fridge, the app that can be uninstalled — these are pre-commitments, but they are not pacts. The Ulysses pact is the form that holds even when the bound self genuinely, in the moment, wants the binding gone.

The two are not opposed. Most lives need both. The point is calibration: knowing which class of situation requires which strength of binding.

The behavioral loop

A short loop with a load-bearing pre-step:

1. Cold-state anticipation — the cold self sees, with clear eyes, that the hot self will predictably act against shared values. The clarity is what makes the pact possible.
2. Pact construction — the cold self designs a structural intervention: a third party who will not yield, a credential thrown away, a substance removed, a contract signed. The construction is the work.
3. Entry into the hot state — the trigger arrives. The wanting begins.
4. Encounter with the wall — the hot self reaches for the lever and finds it absent. The wanting does not disappear; the action is unreachable.
5. Passage — the hot state subsides on its own clock. The action did not happen, not because the will held but because the structure did.
6. Harvest — hours or days later, the deposit lands as something the cold-self knew it could not get any other way: I did not do the thing I would have done.

The loop's signature is that the meaning is downstream of the structure, not upstream of the will.

Emotional drivers

Three feelings cluster around the move.

First: relief at the cold-state construction. The pact is built when the system is honest with itself about its own failure modes, and that honesty itself releases a kind of pressure.

Second: resentment in the hot moment. The bound self can read the cold-state self as a tyrant. This is expected. The pact does not require the hot self to consent to it; that is the whole point.

Third: harvest afterward. A specific quiet form of self-trust that did not have to be manufactured in the hot moment. The cold self protected the future self. The future self knows.

What your nervous system does

The hot and cold states are not metaphors. They are different nervous-system configurations. In the cold state — well-fed, well-slept, distant from the trigger — the prefrontal valuations are loud and the limbic pull is quiet. In the hot state — under stress, cue-exposed, sometimes drug- or hormone-modulated — the limbic system runs the room and the prefrontal voice that built the pact can become inaudible.

Willpower is the attempt to override the limbic system from the prefrontal, in real time, in the hot state. It works sometimes. It fails predictably along certain dimensions: high cue salience, low glucose, sleep debt, social disinhibition, the particular substances and contexts the person has not yet learned around.

The Ulysses pact moves the locus of the work. The decision is made when the prefrontal is loud and the limbic is quiet; the binding then holds physically, requiring no further prefrontal effort once the hot state arrives. The pact is, structurally, an extension of the cold-state self into a region of nervous-system configuration that the cold-state self otherwise cannot reach.

The DojoWell interpretation

The Ulysses pact is the Meaning System's structural-impossibility intervention. Other Systems have analogues — the Threat System uses literal locks and distances, the Reward System uses removed cues, the Belonging System uses public commitments — but the meaning form of the pact is what makes it a pact and not merely a device.

What the Meaning System is protecting, in a Ulysses pact, is the part of the self that would lose coherence if the hot-state action ran. The cold self knows: if I do this, the residue will be larger than any deposit, the path I am on will fork, my self-image will require a story to absorb it, and the absorbing of that story will cost me something I do not want to spend. The pact is built to spare the self that cost.

The substitute — the loop that wears the garb of the pact and is not — is the revocable pre-commitment in an extreme context. The promise to oneself when alone in a hotel room with a bottle. The app one installs to block the site and which one can uninstall in thirty seconds. The financial constraint that is one swipe away from being lifted. These are real interventions in some contexts. In contexts where the limbic pull is high and the prefrontal voice will go quiet, they are the substitute. They share the outer shape of the pact — a constraint, a friction — and they fail in the precise mode the pact was designed to refuse: they bend at the moment of wanting.

Read on the equation: a Ulysses pact pays its effort in the cold state, accumulates near-zero residue (the felt friction reads as freedom on reflection), and lets the deposit harvest as self-trust later. Density: high. The substitute pays effort during the hot state, accumulates residue (the broken promise to the self, the small story about why the device had to come off this time), and delivers a deposit that is near-zero or negative. Density: low.

The framework's reading is not that all desires require pacts. Most do not. The Ulysses pact is a precision instrument for the narrow class of situations where two things are simultaneously true: the cold-state valuation is clear, and the hot-state pull will predictably exceed the will's capacity to hold against it. Outside that class, the pact is over-engineered. Inside it, every other intervention is the substitute.

This is also where the framework breaks with the simpler reading that "constraint is bad and freedom is good." The pact reveals that constraint, chosen by the cold self and binding the hot self, is the form a certain kind of freedom has to take. Freedom of the cold self to remain itself across the hot state is the freedom the pact protects. Elster's phrasing is precise: the pact is constraint as freedom, because the alternative is not a freer self but a self whose values are routinely overridden by a state of itself it has not learned to trust.

What are real examples of a Ulysses pact?

A short, mixed list — domains the framework treats as canonically Ulysses-shaped:

• Substance use in recovery. Throwing out everything in the house at the start of a sober period. Giving the spouse the key to the alcohol cabinet. The constraint must survive the 11pm version of the self that would otherwise drive to a store.
• Spending. Disabling credit cards and leaving them at home before a vacation, where the cold self knows the hot self will overspend in ways the cold self will regret.
• Information diet. Giving the phone to a partner overnight when the cold self knows the hot self will scroll, or installing a content blocker that requires a friend's password to undo for a defined period.
• Advance medical directives. A formal Ulysses pact in the legal sense — instructions that bind the future self against later wishes, used in psychiatric advance directives and end-of-life planning. The framework treats these as the strong-form pact made institutional.
• Decision-making. Public commitments where the cost of reversal is reputational and external — useful in domains where the hot-state pull is to abandon a long arc for an immediate distraction.

Across the list, the test is the same: the cold-state valuation is clear, and the hot-state version of the self will, without structural binding, predictably act otherwise.

Practical steps

1. Identify the narrow class. Most desires do not require a pact. The class that does is the one where you have already lost the same fight more than twice in the same shape. Three failures is enough data.
2. Build the pact in the coldest state available. Not after the trigger, not the morning of, not while already negotiating. The cold-state quality of the decision is what gives the pact its later force.
3. Make the undoing structurally impossible, not merely costly. A device that can be undone with effort will be undone. A pact that requires another person or an institutional process to undo will, in most cases, hold.
4. Use the smallest pact that closes the loop. A pact larger than the loop produces residue without proportionate deposit. A pact smaller than the loop is the substitute.
5. Accept the resentment in the hot moment without negotiating with it. The bound self is supposed to dislike the binding. Disliking it is not evidence that the binding was wrong.

Reflection questions

• What is one specific situation where you have lost the same fight, in the same shape, more than twice? What would a structural-impossibility intervention look like there?
• Where in your life are you using a revocable pre-commitment device in a context the cold self knows requires a pact? What is the cost of the substitution?
• Is there a domain where the cold self does not yet trust the hot self enough to build a pact — where the work first is to admit the failure mode honestly?
• What is one place where a constraint you chose is now the shape of a freedom you would not have without it?

Frequently Asked Questions