I am the principal investigator on a child-nutrition trial in twenty-four rural villages. Twelve receive a daily meal and nutritional support immediately; the other twelve continue without the program for two years while we measure differences in growth, school attendance, and survival. Our charity has enough funding to serve all twenty-four. We are withholding the intervention from half because feeding everyone would answer fewer questions.
Two children in the control group have died since enrollment. This is distressing, but it does not by itself establish that the study should stop. Mortality is one of the outcomes we are measuring. An experiment cannot be considered ethically acceptable only while the comparison group avoids the consequences that make the intervention worth studying. If the absence of food had no serious effects, our donors would have less reason to fund the program in the first place.
I believe the children will thank us for the knowledge. Some will be unable to do so personally, but the ethical case has never depended on every participant receiving an individual benefit. Future programs may reach hundreds of thousands of children. A small improvement in the allocation of those resources could outweigh the entire burden borne by our present sample. The parents who attend our meetings naturally focus on the children they brought with them. Someone has to represent the much larger number who are not in the room.
One father asked whether his daughter would have survived if she had been assigned to the meal program. I explained that randomization permits us to estimate average effects across groups; it does not reveal both potential outcomes for a particular child. He said that we had food and knew she was hungry. These facts are not in dispute. What I resisted was the attempt to move directly from them to a causal claim about an individual death, bypassing the uncertainty that justified the research.
The field team has struggled with this distinction. Several staff members knew the children before joining the project and continue to encounter their families outside working hours. One coordinator began purchasing food for a control household herself. She said that she could not keep weighing the children and then leave without doing anything. I reminded her that the weighing was something. Without accurate measurement, their suffering would remain largely invisible to the institutions with the resources to address it at scale.
She replied that making suffering visible was different from making it stop. That is correct. They are different functions, and our project has deliberately separated them. In the treatment villages, we are implementing an intervention. In the control villages, we are establishing the evidence that will allow future decision-makers to justify implementing it elsewhere. It is understandable that staff prefer the first role. The second requires a degree of discipline that is less immediately rewarded by the people with whom one works.
Several parents have asked to withdraw. They may do so, but withdrawal cannot make them eligible for the treatment program; otherwise the allocation would cease to be random as soon as participants understood its consequences. They say this means they are free to leave the research while remaining in the circumstances the research is observing. I have acknowledged that distinction in the revised information sheet. Clear communication is especially important when people are disappointed by what participation can offer them.
We have also considered whether to stop early if the treatment group appears to be doing much better. The difficulty is that early termination reduces what we can learn about longer-term outcomes. It may be tempting to feed the remaining children as soon as a favorable pattern emerges, but then we lose the opportunity to observe which effects persist, which compound, and which become irreversible. Those questions matter enormously to future allocation decisions. An irreversible harm can be particularly informative precisely because it cannot be reconstructed later under more convenient conditions.
My daughter has asked why I do not simply send the food. She is eleven and has inherited a strong impulse to help whoever is immediately in front of her. I regard this as a good starting point for moral development. Eventually, however, she will need to understand that compassion must survive multiplication: if feeding one visible child today prevents us from acquiring knowledge that could benefit many more, the apparent kindness may be difficult to defend. I have tried explaining this after dinner, when she is more patient.
The report will include a dedication to the children who died during the study. Their families have asked for language acknowledging that the allocation may have contributed to the deaths. Our advisers recommend caution, since individual causal attribution remains uncertain. In the accompanying funding proposal, we estimate that the knowledge could save many thousands of lives. These are different kinds of claim, and I have asked the communications team to preserve the distinction while making the scale of our contribution as clear as possible.
You can send the food today. There are children who will still be alive tomorrow if you do. I don't know how else to say this.
You describe adequate funding, a known nutritional benefit, and a control group without food. What uncertainty justified assigning children to that group? Precision about effect size is not by itself a reason to withhold an effective intervention.
The relevant uncertainty concerns the size and persistence of the effect at this scale. Donors make allocation decisions between programs, and a plausible benefit is not the same as a reliable estimate.
The families understood that participation was a route to meals. At the meeting they asked when the second group would begin receiving them. I did not hear anyone explain the mortality comparison in the terms used here.
You require individual causal identification before acknowledging harm and use a modeled population effect to claim credit for future lives saved. Why do the two directions have different evidentiary thresholds?
The dedication and the funding model serve different purposes and face different review standards. I have asked our advisers whether the wording can make that clearer without implying findings the study was not designed to establish.
If your estimate requires conditions created by withholding food that you have the money to provide, how representative is that of the program you are asking donors to fund? The study design may be manufacturing the contrast.