There is a class of decisions that cannot be made well by gathering more information about them — because the information required to make them is only generated by the act of making them. Parenthood is the clearest case. You cannot know what being a parent is by interviewing parents, reading parenting memoirs, or babysitting. Every one of these activities gives you information about parenthood from the outside. The knowledge parenthood generates — the specific cognitive and emotional world that opens when a particular irreversible commitment is made — is accessible only to the person who made the commitment. The information cannot be transmitted, because it is not a set of facts but a reconfiguration of the prediction model itself.
The person who delays the decision in order to gather more information is making a category error. They think they are reducing uncertainty. They are instead deferring into a state that permanently excludes the information they need — because that information exists only on the other side of the threshold.
Simple Picture
A man in his mid-thirties cannot decide whether to propose. He has been in a relationship for four years, the relationship is good, and he has spent two years trying to become more certain before acting. He reads books about marriage. He talks to married friends. He does therapy to understand his avoidance patterns. He becomes, over time, genuinely sophisticated about commitment and its psychology.
He is no closer to knowing whether to propose.
The reason: everything he has learned is information about commitment gathered from outside commitment. It is accurate information. It is the wrong kind. The knowledge of whether to commit to this person in this relationship — the specific phenomenology of a life built inside this particular commitment — is inaccessible to him until he makes the commitment. He is trying to solve an equation with a variable that only appears after he submits the answer.
The uncertainty he is trying to reduce is not reducible by the strategy he is using. It is only reducible by the thing he is trying to gather information to avoid.
The Epistemological Structure
The brain’s prediction model is calibrated to the states it has actually occupied. It generates predictions from experience, not from models of experience. The uncommitted person has a prediction model calibrated to uncommitted life — they can accurately model the risks of commitment (loss of freedom, the specific person’s flaws, the possibility of failure) but they cannot model the goods of commitment, because those goods only exist inside the committed state. They are not goods that can be observed from outside. They are goods that are constituted by the commitment itself.
This is the deepest form of non-ergodicity. Ergodic systems allow you to reason from population to individual: if most parents find parenting meaningful, you can infer something about what your time series will look like. But the specific knowledge is non-ergodic — the ensemble of parents can tell you the average, not the particular. And the particular is the only thing that matters, because you are not averaging across possible lives; you are traversing one life. The ensemble average says “most committed people report that commitment is worthwhile.” The individual time-series question is “what will I find on the other side of this specific threshold?” — and the answer to that question is generated by crossing it.
Decision theory handles this badly. The standard framework is: maximize expected utility, gather information until marginal cost exceeds marginal benefit, then decide. This works for decisions where the information set is fixed or expanding regardless of choice. It fails for decisions where the information set is radically different on different sides of the decision — where choosing option A locks you into a universe of information that choosing option B forecloses forever. These are the irreversible decisions, and they require a different epistemology.
The Puer’s Rational Mistake
The puer aeternus — the eternal provisional life, the man who keeps his options open — is not making an irrational mistake. He is making a sophisticated mistake. He has correctly identified that commitment forecloses optionality: every door walked through closes other doors. He has incorrectly identified optionality as the thing to preserve.
What optionality cannot buy is the knowledge that lives behind any of those doors. Maximum optionality means maximum exposure to the prediction model calibrated to uncommitted life — which means maximum insulation from every state of knowledge that commitment would generate. The puer is not free. He is locked inside the prediction model of the person who has never committed. He has preserved all his options and foreclosed the one that mattered: the option of becoming someone whose model is calibrated to having been there.
Meaning arrives at 90 degrees to the direction of travel. The man who optimizes for knowing whether the commitment is worth it before making it is looking directly at the precessional effect, which guarantees he will never see it. Meaning is not available to the observer of commitment. It is available to the participant. The person who is fully in a committed relationship, a committed vocation, a committed creative project, encounters a quality of experience that the observer cannot access — not because they are deluded but because participation generates prediction errors that observation cannot. The meaning is in the updates.
The Wound as Gatekeeper
The wound guards the commitment domain — the specific territory where the stakes are highest and the prediction model update would be most transformative. The precision with which the wound organizes avoidance of a particular domain is the precision with which it organizes avoidance of the specific commitments that would generate the most knowledge.
This is why “I’ll commit once I’ve healed enough” is structurally self-defeating. The healing the person imagines — the state of readiness from which commitment would feel safe to make — is generated by the commitment, not by preparation for it. The model that makes commitment feel dangerous cannot be updated by studying danger from a safe distance. It can only be updated by going through enough cycles of the thing it predicted was dangerous and discovering the prediction was wrong. You cannot get the update without incurring the prediction error, and you cannot incur the prediction error without making the commitment that exposes you to it.
The desire-love distinction maps directly here. Desire operates in the uncommitted state: it is the signal the wound uses to pull you toward familiar configurations from a distance. Love is the committed state: the prediction model has been updated by sustained contact with a real, specific person rather than a wound-activating image. Desire cannot tell you whether to commit, because desire is a function of the uncommitted prediction model. The question “do I love them or just desire them?” is not answerable from outside the committed state. Love, as distinct from desire, is what you find out when you are in it — not what you assess from outside it and then decide.
The Finite Game Mistake
Finite players treat commitment as a bet: you commit when the expected value is positive, and you hedge your commitment by maintaining exit options. This is internally coherent. It is also the strategy that produces the experience of never being fully in anything, because exit options are incompatible with the knowledge commitment generates. A commitment made with a visible exit option held in reserve is not a commitment — it is an extended evaluation. Extended evaluations generate information about the evaluation state, not about the committed state. The person who has been “almost committed” for ten years has not accumulated ten years of relationship knowledge. They have accumulated ten years of evaluation knowledge.
The infinite player’s relationship to commitment is different. The safety trap names what the finite approach produces at its logical end: the heart in the casket does not stay intact. It calcifies. And the specific failure mode Lewis describes — “unbreakable, impenetrable, irredeemable” — is not excess of commitment but its permanent deferral. The person who has never committed has not protected their capacity for commitment. They have allowed it to atrophy.
The Epistemological Threshold
The threshold is real. It is not infinitely deferreable. The person who waits until the evidence is sufficient before committing will wait until the window closes — because the evidence of commitment’s value accumulates only on the committed side, and every year of deferral is a year of calibration to uncommitted life, making the threshold feel higher. The uncommitted prediction model generates evidence that commitments are risky. This is accurate. It is also an expanding bias: the longer you run the uncommitted model, the stronger its evidence against commitment, and the higher the bar rises. The threshold moves away from you while you gather information to approach it.
The local optimum trap is precise here: the uncommitted state is locally optimal — it has evidence for its own rationality, it generates returns (freedom, low stakes, reduced risk), and every direction toward commitment feels like descent. The taller hill is on the other side of the valley, and the only way to find out whether the valley is real or just the appearance of descent is to start walking through it.
Dimwit / Midwit / Better Take
The dimwit take is “just commit — stop overthinking it.”
The midwit take is “gather all available information, do the inner work, get clear on your values, then decide when you’re ready.”
The better take is that commitment decisions belong to a class where the information set is constituted by the decision, not prior to it. The evidence about parenthood that matters to a parent is unavailable to a non-parent. The experience of a committed relationship is inaccessible to someone evaluating whether to enter one. More preparation produces more sophistication about the decision without producing proximity to the decision’s actual content. The readiness you are waiting for is generated by what you are waiting to feel ready for. The knowledge is behind the door, not in the hallway. You cannot review the knowledge before choosing the door. The question is not “do I have enough information?” but “is this the kind of decision where the information I need requires me to go first?” — and for the decisions that matter most, the answer is almost always yes.
Threads to Pull
Ideas, thinkers, and questions worth pursuing — and why.
- William James, “The Will to Believe” — James argues that certain propositions are only verifiable from within the committed stance — you cannot evaluate religious belief from a neutral position because the evidence for faith is only generated by acting as if it were true. This is the first rigorous philosophical treatment of the epistemological structure this note describes. The extension: what other propositions are like religious faith in this sense? Love, vocation, political engagement — anywhere that “show me it’s real and I’ll commit” gets the structure of evidence precisely backwards.
- Michael Polanyi on tacit knowledge — Polanyi’s claim that “we can know more than we can tell” is usually read as an observation about skill. The deeper reading: some knowledge is not just hard to articulate but constitutively inaccessible from outside the committed practice. The master craftsman knows things the apprentice cannot know until they have practiced long enough that the knowledge has been built into their prediction model. This is the felt-sense version of commitment epistemology: the body knows before the mind does, because the body is already in the committed state.
- The relationship between commitment epistemology and ergodicity — If commitment produces non-ergodic knowledge states (genuinely inaccessible from outside), then the expected-utility framework for commitment decisions is structurally broken — not just practically inadequate but categorically wrong for a class of decisions. Ole Peters’s ergodicity economics provides the mathematical tools; the application to knowledge states rather than wealth states would be genuinely new work.
- The puer-aeternus literature in depth — Marie-Louise von Franz’s treatment of the puer identifies the refusal to commit as a specific psychological configuration, not just a preference. The puer doesn’t fail to commit because the evidence is insufficient. The puer fails to commit because the prediction model is organized around preserving the possibility of the perfect commitment — which forecloses every actual commitment. The connection between the puer’s psychological structure and the epistemological trap described here is worth making precise: is the puer’s problem primarily emotional (fear of loss, wound-avoidance) or epistemological (a confusion about what kind of evidence commitment decisions require)?
- The hardest version: what if you commit and find the knowledge on the other side is negative? — This note argues you cannot know what commitment generates without committing. But what about commitments that generate knowledge of having been wrong? The person who marries and discovers, from inside the committed state, that the prediction model now calibrated to the relationship is one of incompatibility — did they need to commit to learn this? Or is there information available from outside that would have been sufficient? The distinction between knowledge that is only available through commitment and knowledge that is faster to access through commitment matters practically. Not every commitment produces valuable knowledge. The question is which ones require the epistemological leap and which ones reward more preparation.