People are reluctant to harm some people in order to help others, even
when the harm is less than the forgone help (the harm resulting from
not acting). The present studies use hypothetical scenarios to argue t
hat these judgments go against what the subjects themselves would take
to be the best overall outcome. When the outcomes in question are inc
ome gains and losses for two groups of farmers, subjects judge the har
m they would not impose through their action to be smaller than the ha
rm they would impose through inaction. Some subjects refuse to reduce
cure rates for one group of AIDS patients in order to increase cure ra
tes more for another group, even when group membership was unknowable
to anyone, so that, from each patient's point of view, the change woul
d increase the probability of cure. Likewise, they resisted a vaccine
that reduced overall mortality in one group but increased deaths from
side effects in another group, even when, again, group membership was
unknowable. Some people apply a do-no-harm principle to groups without
apparent understanding of how such a principle might be justified in
terms of its consequences. The capacity for such judgments makes them
vulnerable to learning principles that have no justification at all.