Definition
Altruism is behavior that raises another entity's chances of survival or reproductive success at a measurable cost to the actor. In its strict biological sense — the one Dawkins and Hamilton work with — both cost and benefit are tallied in reproductive output, not in motive or feeling. In its everyday human sense the cost is paid in effort, resources, time, or risk. Either way, the actor gives something up so that another gains.
The reason altruism gets so much theoretical attention is that it looks, at first glance, like something natural selection should erase. If selfish individuals out-reproduce generous ones in every generation, helpfulness should bleed out of the gene pool. That it persists — in suicide-stinging bees, alarm-calling birds, mothers nursing through famine, strangers donating kidneys — is the puzzle every topic below tries to answer.
Why it matters
How it works
The level-of-selection question
The first move in understanding altruism is asking what selection acts on. Pre-1970s biology often spoke loosely of adaptations "for the good of the species," with individual sacrifice imagined as something groups evolved for their own preservation. Dawkins, building on Williams and Hamilton, dismantles this. Imagine a population entirely composed of restrained altruists. A single mutant who breeds without restraint leaves more descendants; those descendants inherit the selfish trait; within a few generations the cooperative population is invaded and overrun. Group-level adaptations are continually undermined from below — they cannot resist the arithmetic of within-group competition.
Move the unit of selection down to the gene and the puzzle inverts. The body is just one vehicle a replicator has built; copies of the same gene sit in close relatives. A gene that builds a body willing to sacrifice for the bodies of its kin is, behaviorally, helping its own copies prosper. What looks like bodily altruism is gene-level self-interest one register deeper.
Hamilton's rule — the arithmetic of kin altruism
W. D. Hamilton turned this insight into a single inequality: a gene for altruism spreads when rB > C. Here r is the coefficient of relatedness — the probability that an allele in the actor is identical by descent with one at the same locus in the recipient. B is the reproductive benefit to the recipient; C is the reproductive cost to the actor. For diploid sexuals, r is ½ for parent-child and full siblings, ¼ for half-siblings and nephews, ⅛ for first cousins. A parent should be willing to pay up to half its own reproductive output for a child; a cousin earns only an eighth of that willingness.
Hamilton's rule does not claim animals literally compute r, B, and C before acting. It claims that the genes that built bodies which behaved as if the inequality held are the genes that spread. The proximate behavioral mechanism is usually a crude rule of thumb — "feed the chicks in my nest," "help the female who smells like me" — that approximated Hamilton's rule reliably enough in the ancestral environment. Haldane's famous quip, that he would not die for his brother but would for two brothers or eight cousins, is a teaching device, not advice: it is the arithmetic of inclusive fitness made vivid.
Inclusive fitness, kin recognition, and Hymenoptera
Classical Darwinian fitness counts only an individual's own offspring. Inclusive fitness counts an individual's direct reproductive output plus the indirect contribution to relatives' reproductive output, each weighted by r. Once fitness is measured inclusively, "selfish gene" and "altruism toward kin" stop being opposed predictions and become two sides of the same accounting. Kin altruism then requires a recognition mechanism: spatial cues (the chicks in my nest are probably mine), olfactory matching (this lamb smells like my own), phenotype matching, and the rare green-beard effect where a gene recognizes copies of itself in others. Selection tunes recognition to be good enough — better than random, calibrated to ancestral conditions — not perfect.
The cleanest demonstration of Hamilton's logic is the eusocial Hymenoptera. Under haplodiploidy, sisters share r ≈ ¾ through their common father, more than the r = ½ they would share with their own daughters. A worker bee maximizes her gene's spread by helping her queen mother produce more sisters rather than by reproducing herself. The entire architecture of ant, bee, and wasp colonies — non-reproductive castes devoted to the queen's brood — falls out of the inequality with no hand-waving.
Behavioral vs psychological altruism
Throughout The Selfish Gene, "altruistic" and "selfish" are defined behaviorally, by their effect on survival probability, not by the actor's motives. This is a deliberately narrow definition, and it makes the same vocabulary describe a bee, a praying mantis, and a human. It also means many acts that feel altruistic can be behaviorally selfish at the level that matters for selection — and conversely, an act done for purely reputational reasons can still be behaviorally altruistic. The two senses of the word slide past each other constantly in popular discussion; the analytic discipline is to keep them separate.
Sapolsky picks up this same problem from the other side. He asks whether "pure altruism" — goodness detached from any expectation of reciprocity, acclaim, or self-esteem — even exists, noting that anonymous organ donors unnerve people precisely because their goodness seems disconnected and affectless. We trust warm-hearted goodness and fear cold-blooded badness, yet the biologies of strong love and strong hate, as Elie Wiesel observed about love and indifference, turn out to be strikingly similar.
Beyond kin — reciprocity, reputation, and culture
Kin selection covers a huge slice of biological altruism but cannot explain helping non-relatives, especially anonymous strangers in one-shot interactions. The second mechanism evolutionary biology offers is reciprocal altruism: when interactions repeat, and when actors can recognize and remember partners, a strategy of help-now-be-helped-later outcompetes pure defection. Indirect reciprocity extends this through reputation — helping the helpers, punishing the cheaters — which lets cooperation reach beyond the dyad to the whole observing community.
Human altruism toward anonymous strangers, however, exceeds what kin selection and reciprocity alone fully predict. Sapolsky's reading is that this excess is produced by a layered stack: evolved emotional machinery (empathy, guilt, gratitude), culturally transmitted norms of fairness, and the reputational incentives that institutions make legible. Dawkins, writing at the gene level, draws a sharper conclusion: if you want a society where people are generously cooperative, do not look to biology for help — teach altruism, because it does not come for free. Both authors agree that the human capacity for genuine other-regarding behavior is real; they differ on whether to call its installation "natural" or "cultural."
Context dependence — the same brain, two behaviors
Sapolsky's larger argument in Behave is that aggression and altruism are not opposites in the way folk psychology imagines. They are outputs of the same contextual brain, deciding which to deploy on the basis of cues — who counts as us, who counts as them, what the stakes are, what the audience expects, what hormones happened to peak that morning. Pulling a trigger and applying a bandage are different movements; bandaging an injured person and shooting a menacing alien can both be "doing right," and in a scanner the prefrontal circuitry that contemplates each looks remarkably similar. The motor act is the easy part. The meaning behind the muscles — and therefore the moral classification — is what shifts.
This is why the same person can heroically aid a stranger one week and ignore another the next. Altruism is not a fixed trait; it is a capacity, expressed or suppressed by circumstance. Any account that omits the context layer will mispredict.
Spotting "good of the species" claims
A useful field test, drawn from Dawkins's topic 1: whenever a popular-biology source claims an animal does something "for the good of the species" — population control, perpetuation of life, ecological balance — ask the mutant question. What would happen to an individual that defected from this group-serving behavior? If the defector would leave more descendants than the conformists, the trait cannot be a species-level adaptation. It must either be a gene-level adaptation (the appearance of altruism masking inclusive-fitness gains) or a misreading of what the animal is actually doing. The mutant test is the single most useful diagnostic for distinguishing real evolutionary explanations from pretty-sounding ones.
Converting any altruism claim into Hamiltonian terms
When you encounter biological altruism in the wild — alarm calls, food sharing, allogrooming, alloparenting — try to estimate r, B, and C. B and C are usually proxies (calories burned, time spent, survival probability foregone); r is the relatedness coefficient given the species' social structure. If rB > C makes the trait look advantageous, kin selection is your null hypothesis. If rB < C, you need another explanation: reciprocal altruism, manipulation, mistaken identity in the kin-recognition mechanism, or measurement error in B and C. Thinking Hamiltonianly is just this — a three-variable comparison that captures, in one inequality, when altruism toward kin is evolutionarily expected.