When I discovered my wife unconscious on the bathroom floor last Thursday, apparently suffering from an opioid overdose, my first instinct was to administer the naloxone nasal spray we keep in the medicine cabinet. But as someone who has spent more than a decade studying rational decision-making, I have learned to be suspicious of first instincts, particularly when they involve emotionally salient situations with potentially irreversible consequences. I therefore took a moment to consider whether reversing the overdose would actually maximize expected utility. This was a difficult decision, made under considerable time pressure and with incomplete information, but I believe the resulting analysis was sound. My wife survived, thanks in part to my eventual intervention, and has spent the past several days accusing me of being a monster. I would remind her that a monster would not have bothered to run the numbers.
\nThe central problem was one of expected-value maximization. Administering naloxone would increase the probability of my wife's survival, thereby preserving the future welfare she could reasonably expect to experience, along with her potential contributions to the welfare of others. On the other hand, her continued existence would also entail certain costs, including food, housing, healthcare, and the ongoing emotional demands of our marriage. Failing to administer the drug would eliminate these future expenses while imposing substantial negative utility on her, her friends, and certain members of her family. The question was whether the benefits of the intervention outweighed its costs. Although the calculation was complicated by the urgency of the situation, I felt that making an uninformed decision would have been a disservice to everyone involved, particularly my wife, whose interests were ostensibly the reason for intervening in the first place.
\nI have seen considerable criticism of this approach from people who apparently believe that emergency medical decisions should be governed by unconditional rules rather than consequentialist reasoning. "Of course you save your wife" is the sort of thing people say when they have never seriously examined the assumptions underlying their moral intuitions. Why, precisely, is saving one's wife always the correct decision? What if she is extraordinarily unhappy? What if the intervention has significant costs? What if the resources required to preserve her life could be allocated to activities with greater expected social returns? These are not rhetorical questions. They are the sorts of questions that any serious moral agent should be prepared to answer before committing resources to a particular course of action. The fact that my wife's respiratory function was deteriorating did impose certain time constraints on the analysis, but time pressure does not magically convert an unexamined intuition into a valid ethical principle.
\nFortunately, I already maintain an extensive spreadsheet tracking the expected utility of major decisions in our marriage. Originally created to evaluate whether we should purchase a home, the model has since expanded to include career decisions, family planning, the expected lifetime costs of pet ownership, and the question of whether our continued relationship represents an efficient allocation of our respective emotional resources. I was therefore able to retrieve much of the relevant information without starting from scratch. My wife's projected remaining quality-adjusted life years were the most important positive input, although I applied a discount to account for her history of substance use, her increasingly pessimistic disposition, and her tendency to spend entire Sunday afternoons watching reality television. I also included an estimate of her future earnings, adjusted for the probability that she would eventually follow through on her frequently stated intention to quit her job and become a ceramicist.
\nThe financial picture was more complicated. My wife earns approximately $71,000 annually and contributes a meaningful share of our household expenses, but she also consumes resources that would otherwise be available to me. Her preference for maintaining the thermostat at 73 degrees, for example, has increased our electricity costs by an amount I consider difficult to justify. She insists on buying name-brand groceries, frequently orders takeout despite the availability of perfectly adequate food in the refrigerator, and recently spent $180 on a haircut without conducting any apparent comparison of alternative providers. These expenditures may seem trivial when considered individually, but over a sufficiently long time horizon they compound into a substantial reduction in household wealth. I do not mean to suggest that her life should be valued solely in financial terms. I have also assigned positive utility to her companionship, physical affection, and occasional willingness to drive me to the airport.
\nOne particularly difficult variable was the expected future quality of our marriage. Over the past eighteen months, my wife has become increasingly critical of my intellectual interests, describing my engagement with rationalist philosophy as an excuse to avoid ordinary domestic responsibilities. She has complained that I spend too much time on LessWrong, that I routinely turn personal disagreements into formal debates, and that I once submitted an anonymous prediction market on the probability of our marriage ending in divorce. I understand why she found the last incident upsetting, although I maintain that anonymous forecasting is a perfectly legitimate mechanism for aggregating dispersed information. In any case, I had recently revised my estimate of our marriage's long-term success downward, and it would have been intellectually dishonest to ignore this information merely because she happened to be unconscious on the floor.
\nThere was also the issue of moral hazard. My wife has struggled with opioid use in the past, and I am deeply sympathetic to the challenges associated with addiction. Nevertheless, a rational analysis must account for the possibility that reliably reversing the consequences of risky behavior could alter an individual's future incentives. I do not believe in punishing people for making mistakes, but I do believe in designing systems that encourage optimal behavior over time. I briefly considered whether making the availability of naloxone conditional upon certain behavioral commitments might improve her long-term outcomes, although I ultimately concluded that she was not in a position to negotiate an enforceable agreement. This was unfortunate, as I had previously prepared a draft household accountability framework that included provisions for substance use, excessive discretionary spending, and the unauthorized relocation of my belongings.
\nAfter considering these factors, I arrived at a provisional estimate that administering naloxone would produce a modestly positive expected utility, though with considerable uncertainty. The result was sensitive to assumptions about our future relationship, the costs associated with continued medical treatment, and the probability that my wife would eventually achieve sustained recovery. I was about to perform a Monte Carlo simulation when I remembered an additional consideration that had not been incorporated into the original model: my wife is the primary account holder on our family cellphone plan and controls several important household passwords. Her death would likely create considerable administrative difficulties, including the possibility that I would lose access to our shared streaming subscriptions and be required to establish a new relationship with our internet service provider. Once these externalities were included, the expected value of intervention became substantially more favorable.
\nI therefore administered the naloxone and contacted emergency services. My wife subsequently regained consciousness and was transported to the hospital, where she received additional medical attention. I would characterize the operation as a qualified success. The primary objective, preserving her life, was achieved, and the costs were broadly consistent with my initial projections. I also gained valuable practical experience making consequentialist decisions under time pressure, which I intend to incorporate into a forthcoming LessWrong post on emergency resource allocation. The paramedic who arrived at our apartment seemed less interested in the philosophical dimensions of the case. When he noticed my laptop open beside the bathroom door and asked what I had been doing, I attempted to explain the structure of the model, but he interrupted to ask how long my wife had been unresponsive. I found his insistence on focusing exclusively on the clinical variables somewhat narrow-minded.
\nMy wife, unfortunately, has not been particularly grateful. When I explained what had happened, she asked whether I was seriously telling her that I had stopped to calculate whether she was worth saving. I told her that this was a gross mischaracterization of my methodology. I had not attempted to determine whether she was "worth saving," a phrase loaded with subjective and potentially discriminatory assumptions. I had evaluated the expected consequences of two competing interventions under conditions of uncertainty. The distinction is philosophically significant, and I would have hoped that someone who has been married to me for seven years would be capable of appreciating it. She responded by asking how long I had waited before administering the Narcan. I explained that the elapsed time was already captured in the model as an opportunity cost and that focusing on the raw number of seconds would not meaningfully advance our understanding of the decision.
\nThe conversation became particularly unpleasant when she asked what I would have done if the calculations had come out differently. I answered that I would have followed the evidence, as any responsible person should. She began crying and accused me of admitting that I might have allowed her to die. I tried to reassure her by pointing out that the actual calculation had produced the opposite result, making the hypothetical scenario irrelevant to our present circumstances. Indeed, I had gone out of my way to incorporate several generous assumptions about her future happiness and productive capacity. Under a less favorable model, the decision could easily have been more difficult. I thought she might find this comforting, but she instead asked me to leave the hospital room.
\nSince then, several friends have suggested that the appropriate response to an emergency is to provide assistance immediately rather than attempting to construct an individualized utility function for the person in distress. I understand the practical appeal of this heuristic, particularly for individuals without extensive training in decision theory. However, I am concerned about the broader implications of treating emotional urgency as sufficient grounds for bypassing rational analysis. If we accept that certain situations are simply too important to think carefully about, we risk establishing a precedent under which the most consequential decisions are systematically delegated to instinct. This strikes me as precisely backward. When the stakes are highest, we should demand greater intellectual rigor, not less. I am aware that the emergency physicians who have commented on the incident disagree, but most have not published anything substantial on the subject of expected utility.
\nMy wife's sister has been especially hostile. During a telephone conversation yesterday, she described my behavior as unforgivable and suggested that my wife should divorce me immediately. I explained that this would introduce substantial transaction costs, including legal fees, housing disruptions, and the destruction of various forms of relationship-specific capital accumulated over the past seven years. I also pointed out that my wife's recent medical emergency had already imposed an unexpected burden on our finances, making this an unusually poor time to initiate additional costly proceedings. She responded by calling me a "fucking psychopath," which is not a recognized objection within any formal framework of rational choice theory. I have since attempted to redirect the discussion toward the relevant variables, but she has blocked my number.
\nThere is one aspect of this experience that I do regret. In retrospect, I should have developed a more comprehensive emergency decision framework in advance, including predetermined thresholds for administering life-saving interventions to members of my household. Doing so would have reduced the computational burden at the moment of crisis and perhaps prevented some of the interpersonal misunderstandings that followed. I have already begun work on such a framework, with separate provisions for opioid overdose, choking, cardiac events, and various other contingencies. The current draft incorporates both direct welfare effects and second-order considerations such as insurance coverage, household contributions, and expected future relationship satisfaction. My wife has asked me not to send it to her, but I believe that once her emotions have settled, she will recognize the value of having these policies explicitly documented.
\nI have also been accused of applying a double standard because I would supposedly expect my wife to administer life-saving treatment to me without performing a similar calculation. This is inaccurate. I would fully support her use of a rigorous expected-value model, provided it incorporated all relevant information, including my future contributions to AI alignment research, my unusually high level of intellectual engagement, and the potentially enormous downstream consequences of my continued existence. In fact, I suspect that even a relatively crude approximation of these variables would make the correct decision obvious. I have offered to help her construct such a model, but she has declined.
\nMy wife has now informed me that she intends to stay with her sister while considering whether our marriage can continue. This is deeply disappointing, and I believe she is making an emotional decision that she may eventually regret. More importantly, her departure substantially alters several of the assumptions underlying my original analysis, particularly those concerning her future contributions to our household and the expected utility of our ongoing relationship. If she proceeds with a divorce, the calculation that justified administering naloxone will need to be revisited. I have explained this to her in an effort to demonstrate the practical consequences of her behavior, but she has accused me of threatening her, which was not my intention.
\nI sincerely hope that we can resolve this misunderstanding. I love my wife, and I would like to believe that she appreciates the extraordinary care I took in reaching a decision about her continued existence. There are millions of husbands who would have administered the Narcan without thinking twice, acting on nothing more than a primitive emotional impulse or a culturally inherited sense of marital obligation. I respected my wife enough to subject the question of whether she should live or die to a rigorous quantitative analysis, taking into account both her interests and the interests of everyone else who might be affected.
\nAnd after considering all that, I chose her. I cannot understand why she doesn't find that romantic.
\n","excerpt":"When I discovered my wife unconscious on the bathroom floor last Thursday, apparently suffering from an opioid overdose, my first instinct was to administer the naloxone nasal spray we keep in the medicine cabinet. But as someone who has spent more than a decade studying rational decision-making, I have learned to be suspicious of first instincts, particularly when they involve emotionally salient situations with potentially irreversible consequences. I therefore took a moment to consider whether reversing the overdose would actually maximize expected utility. This was a difficult decision, made under considerable time pressure and with incomplete information, but I believe the resulting analysis was sound. My wife survived, thanks in part to my eventual intervention, and has spent the past several days accusing me of being a monster. I would remind her that a monster would not have bothered to run the numbers.","tags":["Decision Theory","Effective Altruism","Relationships"],"score":287,"displayOrder":470,"thumbnail":"/assets/articles/i-ran-the-numbers-before-narcanning-my-wife-thumb.webp","image":"/assets/articles/i-ran-the-numbers-before-narcanning-my-wife.webp","commentCount":6,"date":"2026-10-06T00:00:00.000Z","updatedAt":"2026-10-06T00:00:00.000Z","revision":2,"status":"published"},"comments":[{"id":"seed-i-ran-the-numbers-before-narcanning-my-wife-0","author":"clock_is_running","text":"The delay belongs in the model. Every minute spent improving the estimate changes the outcome you are estimating.","parentId":null,"score":82,"date":"2026-10-06T00:18:00.000Z","own":false,"moderatable":false,"seed":true,"ownerAuthor":false},{"id":"seed-i-ran-the-numbers-before-narcanning-my-wife-1","author":"marital_expected_value","text":"Agreed. In a future version I would precompute the most time-consuming terms, although some depend on my assessment of the relationship at the time.","parentId":"seed-i-ran-the-numbers-before-narcanning-my-wife-0","score":21,"date":"2026-10-06T00:47:00.000Z","own":false,"moderatable":false,"seed":true,"ownerAuthor":false},{"id":"seed-i-ran-the-numbers-before-narcanning-my-wife-2","author":"ordinary_spouse","text":"She should not have to outperform a spreadsheet to receive the medicine already in your hand.","parentId":null,"score":94,"date":"2026-10-06T01:33:00.000Z","own":false,"moderatable":false,"seed":true,"ownerAuthor":false},{"id":"seed-i-ran-the-numbers-before-narcanning-my-wife-3","author":"units_check","text":"What is the uncertainty interval on your estimate of your own future contribution? It seems much narrower than the interval you assigned to hers.","parentId":null,"score":35,"date":"2026-10-06T02:36:00.000Z","own":false,"moderatable":false,"seed":true,"ownerAuthor":false},{"id":"seed-i-ran-the-numbers-before-narcanning-my-wife-4","author":"bounded_attention","text":"You say incomplete information forced you to calculate. Incomplete information also limits what the calculation can establish. There was an available action with an immediate benefit.","parentId":null,"score":42,"date":"2026-10-06T04:37:00.000Z","own":false,"moderatable":false,"seed":true,"ownerAuthor":false},{"id":"seed-i-ran-the-numbers-before-narcanning-my-wife-5","author":"take_the_phone","text":"Please let someone else keep the emergency supplies. Your contingency plan still requires a philosophical review.","parentId":null,"score":56,"date":"2026-10-06T07:01:00.000Z","own":false,"moderatable":false,"seed":true,"ownerAuthor":false}]}