See also "How People Die By Suicide," my review of Thomas Joiner's book, Why People Die By Suicide, challenging Joiner's refusal to consider an adaptive model for suicide and attempted suicide.
Suicide, like filicide, seems upon first consideration to be a ludicrous act, viewed from the perspective of evolutionary biology. However, the act of filicide (the killing of one's child or children) is clearly adaptive in many cases. Not only that, but it can be shown through statistical evidence that actors seem to differentiate between adaptive and non-adaptive filicides when they "decide" to commit filicide (as well as other homicides). What about suicide?
We can define an act to be adaptive when the act increases the genetic fitness of the actor - that is, when the act's benefits - in terms of survival, procreation, or nepotistic distribution of resources to one's genetic relatives - exceed the act's costs, in the same terms. When considering adaptiveness of acts or traits, it is important to consider not only the particular acting organism and its existing offspring, but all its genetic relatives, including possible future offspring and non-offspring relatives. (A judgment about whether an act is adaptive or not implies no moral judgment. An adaptive act may be praiseworthy, horrible, or neither.)
Filicide is adaptive when the resources to be expended raising a particular child would be better spent on others, in terms of benefit to genetic relatives, present or future. If a child is likely to die or otherwise fail to reproduce, or if raising the child will adversely affect the parent's ability to have future children that may be a better genetic "bet," then filicide - while certainly a gruesome act - must be said to be adaptive. Perhaps even more obviously, filicide of a partner's offspring that is not one's own genetic offspring is almost always adaptive.
Are people more likely to kill their children under circumstances where the act is adaptive? A major body of work in evolutionary psychology suggests that this is so. For instance, stepparents are much more likely to kill their stepchildren than birth parents are to kill their genetic children. People are much more likely to kill babies than older children, and younger mothers are more likely to kill their babies than older mothers. It seems that not only is filicide sometimes adaptive, but that humans possess mechanisms to prevent themselves from committing filicide when it is not adaptive, and to allow themselves to commit filicide when it is adaptive.
Does the same hold true for suicide? There are two related questions: first, is suicide ever adaptive? Second, if suicide is sometimes adaptive, do humans appear to possess mechanisms to limit suicide to cases where it is adaptive?
As to the first question, it can be clearly demonstrated, at least in the abstract, that suicide is sometimes adaptive. The easy case is one in which a person sacrifices her life so that genetic relatives may live. Such cases must be rare, and are so different from the usual connotations of "suicide" as to barely be considered suicide at all. The more common case where suicide is adaptive is this: one's total expected future contribution to one's genetic fitness is exceeded by one's total expected drain on the resources of one's genetic relatives.
To put this in more concrete terms, there are many cases - old age, crippling disability - where all measures of genetic fitness approach zero. Once one may no longer reproduce, and is no longer an effective nepotistic distributor of resources (including wisdom), one's expected contribution to one's own genetic fitness is likely to be nil. However, as long as one survives in this condition, he not only contributes nothing to his own genetic fitness, but also likely drains the resources of his genetic relatives. His continued survival is contrary to his genetic interests. Therefore, suicide, in this situation, must be said to be adaptive.
The second question is: given that suicide is sometimes adaptive, do human beings tend to commit suicide in circumstances where it is adaptive?
Unfortunately, I lack the quantitative tools and study design knowledge to answer this question. But I will point to two conflicting bodies of evidence on this question.
In support of the idea that humans tend to commit suicide when it is adaptive is data demonstrating that the suicide rate increases dramatically for elderly people. Using age as a rough estimate of expected genetic contribution, the suicide rate (inversely) tracks genetic fitness. Excluding children, the suicide rate is lowest for those with the highest expected evolutionary fitness - those ages 15-34. The suicide rate climbs from there. In 1950, before the age of nursing homes, the suicide rate for those ages 75-84 was more than double the average suicide rate for the country. In addition, suicides by the elderly are more likely to be planned, with less likelihood of warning prior to the act.
In addition to data on suicide among the elderly, suicide is particularly likely in elderly people with poor health, and particularly those with low vision - both conditions that, in the environment of evolutionary adaptedness, might be expected to indicate low expected contribution to the fitness of one's genetic relatives.
Contrary to this is the data on familicide. Familicide is the killing of one's spouse and children. It is almost the exclusive province of males. There are two key features of familicide that are relevant to the discussion of the adaptiveness of suicide: a familicide is much more likely to kill his biological children than a mere filicide, and a familicide is much more likely to commit suicide than a simple filicide or uxoricide (wife killer).
Familicide seems to be an extremely counter-adaptive act, especially since the killer's biological children are commonly involved, compared to simple filicides, which are more likely to involve stepchildren. The fact that suicide is much more likely in a familicide, as opposed to a simple uxoricide or filicide, suggests, by association, that suicide is frequently a counter-adaptive act.
Note: the question of whether suicide is adaptive is an entirely different question from whether it is rational, in the lifetime-utility-maximizing sense.