My eleven-year-old daughter has cancer, and while I love her very much, I am struggling to justify the expense of her chemotherapy when the same money could be used to improve the welfare of tens of millions of farmed shrimp. This is an uncomfortable conclusion, and I want to emphasize that I did not arrive at it lightly. I spent the better part of a weekend reviewing the relevant literature, constructing a cost-effectiveness model, and conducting sensitivity analyses across a range of assumptions about crustacean sentience. The results were remarkably consistent. Even under relatively conservative estimates, the expected welfare gains from redirecting our daughter's treatment fund to shrimp welfare interventions substantially exceed those from prolonging the life of a single human child. My wife has accused me of abandoning our daughter. I would argue that she is allowing a deeply ingrained speciesist bias, compounded by an unusually strong genetic attachment to one particular individual, to override her commitment to doing the most good.

The organization I have identified as the optimal recipient is Shrimp Welfare Project, which works with aquaculture producers to reduce suffering among farmed shrimp, including by introducing electrical stunning equipment to render the animals unconscious before slaughter. Its public impact estimates suggest that its programs improve the welfare of approximately 1,400 shrimp per donated dollar per year. At that rate, the $23,400 our family would need to contribute toward the proposed course of treatment could instead help approximately 32.8 million shrimp. My daughter, by comparison, is one person. I am aware that this framing may strike some readers as insensitive, but I would encourage them to consider whether their discomfort is based on a defensible moral principle or merely the evolutionary accident that human beings find children with faces more emotionally compelling than small aquatic invertebrates.

I first raised this issue during a meeting with my daughter's oncologist, who had been explaining the proposed treatment regimen and its likely side effects. When she finished, I asked whether the hospital had performed a comparative cost-effectiveness analysis incorporating potential benefits to nonhuman animals. She appeared confused, so I explained that modern moral philosophy requires us to take seriously the interests of all sentient beings, regardless of their taxonomic classification. She said that her job was to treat children with cancer. I acknowledged that this was a perfectly understandable institutional objective, but pointed out that optimizing for the survival of pediatric oncology patients was not necessarily equivalent to optimizing for total welfare across all morally relevant beings. She asked whether I had any questions about the chemotherapy. I said that I had several, beginning with whether she considered shrimp capable of experiencing suffering.

My wife later described this exchange as humiliating, which I find particularly disappointing given her previous enthusiasm for effective altruism. When we first met, she was impressed that I donated ten percent of my income to highly effective charities and had written several essays about the moral importance of impartial beneficence. She even attended an EA Global conference with me, although she spent much of the weekend complaining that nobody seemed interested in talking about anything other than AI. It is therefore difficult to understand why she has abandoned these principles now that they have become relevant to a concrete decision. She keeps insisting that our daughter is different because she is our daughter, as though genetic proximity were a morally relevant variable. I have asked her repeatedly to specify the coefficient by which she believes biological relatedness should multiply an individual's welfare weight. She has not provided a number, preferring instead to accuse me of being a sociopath.

I am not suggesting that my daughter has no moral value. Quite the opposite. Her welfare is explicitly included in my model, and I have assigned it considerable weight. She is a bright, affectionate child with a promising future, and I would be delighted if she survived long enough to finish middle school, attend college, and perhaps eventually contribute to important work in AI alignment. But we cannot simply ignore the opportunity costs associated with bringing about these outcomes. Every additional dollar spent on her medical treatment is a dollar that cannot be spent preventing shrimp from experiencing avoidable suffering. Every hour of potentially healthy life we purchase for her must be evaluated against the aggregate welfare of the vast number of animals whose interests we could have advanced instead. There is no principled reason to exempt my daughter from this calculation simply because I taught her how to ride a bicycle and have several thousand photographs of her on my phone.

One objection I frequently encounter concerns the uncertainty surrounding shrimp sentience. Critics point out that we do not know precisely what shrimp experience, how intense their suffering might be, or how to compare it with human suffering. These are legitimate epistemic challenges, but they do not justify assigning shrimp welfare a value of zero. The appropriate response to moral uncertainty is to model it explicitly. My spreadsheet therefore includes parameters for the probability of shrimp sentience, the relative intensity of shrimp suffering, and the number of shrimp affected by a given intervention. I have also included a parameter for the probability that my daughter survives treatment, although my wife has asked me to stop referring to this as the "human intervention success rate" in front of her. Naturally, the model is sensitive to the assumptions one makes, but I have found that even very small welfare weights can produce enormous aggregate values when multiplied by sufficiently large populations. This is a mathematical fact that no amount of crying in the hospital parking lot can invalidate.

My daughter herself has been surprisingly resistant to the argument. Last Tuesday, while I was reviewing the latest version of the spreadsheet at the kitchen table, she asked whether I was really considering refusing to pay for her treatment. I explained that I was not refusing anything; I was attempting to identify the allocation of our family's limited resources that would maximize expected welfare. She asked why I cared more about shrimp than about her. This was an extraordinarily loaded question, and one that betrayed a fundamental misunderstanding of impartial consequentialism. I told her that I cared about her exactly as much as I should, and that the problem was that there were simply too many shrimp for her interests to outweigh theirs. She began to cry, at which point my wife entered the room and demanded that I put away my laptop. I would have preferred to finish explaining the distinction between total and average utilitarianism, but the conversation had clearly become too emotionally charged to proceed productively.

I should also address the accusation that I am being unusually strict about cost-effectiveness when it comes to my own family's expenses. My wife has pointed to my recent purchase of a $2,700 gaming computer, a $900 ticket to a rationalist retreat, and several hundred dollars' worth of nootropic supplements as evidence that my commitment to impartiality is selective. This criticism misunderstands the concept of instrumental value. My computer allows me to run local language models and perform the calculations necessary for high-impact philosophical work. The retreat provides valuable opportunities to network with people who may one day influence the development of transformative AI. The supplements support the cognitive functioning upon which much of my future contribution to humanity depends. My daughter's chemotherapy, by contrast, is principally intended to benefit my daughter. While I certainly appreciate the emotional significance of that goal, it is important to distinguish between investments in people with unusually high expected impact and expenditures motivated primarily by sentimental attachment.

Indeed, the most troubling aspect of this controversy has been my family's insistence that I privilege their interests over those of strangers. My mother called yesterday to tell me that she would personally contribute toward the treatment if I was unwilling to do so. I thanked her for her generosity but explained that this would simply transfer the moral error from one household to another. I suggested that she consider donating the same amount to Shrimp Welfare Project instead, and offered to help her set up a recurring contribution. She told me to go to hell, which I found needlessly hostile. My father has since begun organizing a fundraiser among our relatives, apparently under the impression that raising enough money to cover treatment would resolve the underlying ethical problem. It would certainly resolve the immediate financial problem, but it would also demonstrate the extent to which our society systematically overallocates resources to photogenic beneficiaries with compelling personal narratives.

The fundraiser is particularly frustrating because it has received an extraordinary amount of attention despite providing almost no information about comparative cost-effectiveness. My sister posted a photograph of my daughter in a hospital bed alongside a request for donations, and within two days the campaign had raised more than $18,000. I subsequently posted a detailed explanation of the opportunity costs associated with this expenditure, including a chart showing the approximate number of shrimp whose welfare could be improved at various funding levels. My sister deleted the comment and blocked me. The following morning, I discovered that the fundraiser's description had been amended to specify that all proceeds would go directly toward my daughter's medical expenses. This is a classic example of donor preference rigidity, in which charitable contributions are artificially restricted to a predetermined beneficiary even when substantially more cost-effective opportunities exist elsewhere.

Several people have suggested that I am losing sight of the purpose of effective altruism, which is supposedly to make the world a better place rather than to provide an intellectual justification for neglecting one's family. I find this distinction incoherent. If the goal is to maximize welfare, then we must be prepared to accept conclusions that conflict with our immediate emotional intuitions. Otherwise, we are simply practicing ordinary charitable giving with a more impressive vocabulary. I have spent years criticizing people who donate to local animal shelters rather than interventions that help vastly larger numbers of animals. It would be hypocritical of me to suddenly abandon that position merely because the emotionally salient beneficiary happens to sleep in the bedroom across the hall. Intellectual consistency demands sacrifice, and I am prepared to make an enormous sacrifice on my daughter's behalf.

That said, I am willing to acknowledge certain limitations of my analysis. The cost-effectiveness estimates for shrimp welfare depend on uncertain assumptions, and the marginal impact of a large donation may differ from the average impact reported by an organization. There are also reasonable disagreements about how to aggregate welfare across species and whether human lives possess certain values that are not adequately represented by suffering-adjusted time. These issues deserve serious consideration, and I would welcome a rigorous technical critique from anyone capable of engaging with the model on its own terms. Unfortunately, the criticisms I have received so far have consisted almost entirely of personal attacks, anecdotes about my daughter's kindness, and emotionally manipulative references to the possibility that she might die. None of these constitutes a sensitivity analysis.

My wife has now informed me that she intends to proceed with treatment regardless of my objections and has asked me to stay with my brother for the time being. I respect her autonomy, although I believe she is making a serious moral mistake, and I am disappointed that she has chosen to end our discussion before we could reach a consensus. I have offered to meet with her, my daughter's oncologist, and an independent animal-welfare researcher to review the relevant evidence in a neutral setting. So far, nobody has accepted the invitation. In the meantime, I am continuing to refine the model and have submitted a preliminary version to several members of the effective altruism community. One respondent suggested that I was confusing a speculative aggregate welfare estimate with an actual moral obligation to withhold treatment from my child. I have asked him to elaborate, but he has not replied.

I remain hopeful that my daughter will eventually understand the reasoning behind my position. I have spent eleven years teaching her the importance of kindness, compassion, and caring for those less fortunate than herself. We have discussed animal suffering at length, and she knows that moral concern should not be limited to those who look like us. She used to get upset when she saw insects struggling in the swimming pool, and I was always proud of her for insisting that we rescue them. It is therefore painful to see her reject the same principles now that the beneficiaries are less familiar and the costs more personal. I love my daughter more than anything in the world, but there comes a point when love must give way to reason.

I only wish she cared about the shrimp as much as I do.