The adaptationist view holds that the mind is neither a blank slate nor a general-purpose CPU, but a bundle of special-purpose modules — recognizing faces, detecting cheaters, choosing mates, avoiding predators — each handled by its own "psychological circuit." Every circuit is an "adaptation" that selection shaped to solve a recurrent ancestral problem.
Non-trivial: (1) the functional question comes before the mechanism question. To understand a trait, first ask "what problem was it selected to solve?" and reverse-engineer the mechanism. (2) Two layers of "why" — proximate and ultimate: the proximate cause of hunger is low blood sugar; the ultimate cause is that not eating reduces reproductive chances. Both are correct, but they answer different questions; conflating them is a common error. (3) Special-purpose, not general-purpose, is the very source of systematic bias: a circuit that was efficient ancestrally misfires in the modern world. (4) A crucial self-correction: not every trait is an adaptation. Some are byproducts (the nose evolved to breathe, but happens to hold up glasses); some are random drift. Inventing an "evolutionary story" (a just-so story) for everything is this lens's worst abuse. The healthy use is to generate falsifiable functional hypotheses, not to rationalize after the fact.
The Wason selection task: the abstract logic version ("if a card has a vowel on one side, it has an even number on the other — which to flip?") yields very low accuracy. But its logically identical social-contract version ("you must be 18 to drink — whom do you check?") sends accuracy soaring. Same logic, dressed as "detect the rule-breaker," and the dedicated cheater-detection module switches on. Strong evidence that the brain isn't running general logic, but a suite of domain-specific programs.
(1) This is isomorphic to reading undocumented legacy code — you can't read the implementation directly, so you infer "what requirement did this originally satisfy?" Apply the same move to the brain: for a seemingly irrational emotion, first ask "what did it solve ancestrally?" (2) Contrast with AI: an LLM approximates "general computation," while the human brain is "modular assembly" — exactly why humans have systematic biases and strong general models don't, and why the real value of human-AI collaboration is complementarity, not making humans imitate general computation. (3) Parenting: a child fearing the dark or large unfamiliar animals is far more "natural" than fearing cars or power sockets — the former were real ancestral threats. Understanding this stops you from "reasoning" with a fear circuit that doesn't operate at the level of reasoning.
Adaptations were tailored to the ancestral environment. The problem: cultural and technological evolution far outpaces genetic evolution. So circuits that were once highly adaptive now systematically misfire in the modern world. It's not that "people are broken" — it's that people didn't change, the environment did. The bug isn't in us; it's in the speed gap.
Non-trivial: (1) the supernormal stimulus is the key mechanism — an exaggerated version of a real ancestral cue that hijacks reward circuits. Junk food (industrially concentrated sugar/fat once scarce), social-media likes (reputation, once earned in person, now a digital counter), short-form video (an endless drip of novelty) all pull on us harder than the real things they imitate. (2) Mismatch is specific, not blanket: it's not "everything modern is bad," but particular circuits failing under particular modern stimuli. (3) The cure is to design your environment, not grind on willpower: since the problem is the stimulus, the most effective intervention is changing what you're exposed to, not repeatedly fighting a circuit engineered to hijack you.
A strong craving for sugar and fat was a life-saving advantage in food-scarce ancestral environments — every extra stored calorie meant a better shot at survival. The same taste preference, transplanted into a world where sugar and fat are everywhere, becomes a driver of obesity and metabolic disease. The circuit didn't change one bit; only its environment flipped from "scarcity" to "surplus."
(1) To you this just is distribution shift in machine learning: a model excels on its training distribution (ancestral environment) but collapses when deployed on a new one (modern environment) — the human brain is a model trained on the old distribution, forced to run on the new. (2) It meshes directly with dopamine reward-prediction error (see Day 35): supernormal stimuli keep manufacturing "better-than-expected reward," recalibrating the reward circuit to a baseline reality can never meet. (3) Most practical for parenting: a child's "addiction" to short video/games is a supernormal stimulus hijacking circuits built for real social interaction and exploration. The fix isn't a daily tug-of-war with willpower but designing the environment — make it the non-default, add friction, and crowd it out with real high-quality stimuli (exercise, face-to-face play). Changing the environment is far cheaper than changing the child.
Natural selection optimizes for "surviving"; sexual selection optimizes for "reproducing" — and the two often conflict. The peacock's tail hampers escape yet boosts mating success. What selection truly optimizes is never individual welfare, but gene propagation. Sexual selection runs along two paths: intrasexual competition (e.g., stags fighting with antlers) and intersexual choice (e.g., peahens picking by tail).
Non-trivial: (1) only a costly signal is an honest signal (the handicap principle): the peacock's tail is a credible "my genes are good" advertisement precisely because it is so expensive that only the truly fit can afford it. A fakeable signal is worthless; a hard-to-fake handicap carries information. (2) Fisherian runaway is a positive-feedback loop: preference and preferred trait reinforce and co-amplify each other, pushing the tail to absurd extremes — a rare case of "exponential runaway" in evolution. (3) Corollary: much human behavior that looks wasteful, exaggerated, or irrational reveals its function under the "signaling" lens — it broadcasts hard-to-fake quality to a specific audience.
The peacock's tail is the textbook case: it markedly lowers survival (easier to catch, resource-hungry) yet keeps being amplified by female preference. This shows sexual selection can push against natural selection — as long as the reproductive payoff outweighs the survival cost, the genes happily run the "loss-making" trade.
(1) Costly signaling is everywhere in modern careers: a technically "over-polished" open-source side project is essentially a peacock's tail — expensive, hard to fake, honestly broadcasting "I have the spare capacity to do this." Seeing this lets you consciously choose your signals: invest where they're hard to fake and where the target audience is actually looking, not in cheap self-promotion. (2) Conspicuous consumption and performative overwork follow the same logic — grasping their signaling function keeps you from being swept into a runaway loop. (3) This is also a textbook positive-feedback system (echoing feedback loops you know well): Fisherian runaway is structurally isomorphic to social-media "status races," which self-accelerate past rational bounds.
Why would organisms sacrifice for relatives? Hamilton's rule gives the stark answer: an altruistic gene spreads when rB > C — r is the coefficient of relatedness, B the reproductive benefit to the recipient, C the cost to the altruist. In other words, helping a relative propagates the very gene you also carry.
Non-trivial: (1) the unit of selection drops from "individual" to "gene" (the gene's-eye view). The organism is merely a "vehicle" genes temporarily build to copy themselves. This shift instantly makes otherwise inexplicable altruism self-consistent. (2) It explains both family bonds and their misfires: ancestrally, "the people around you were mostly kin," so the brain approximates relatedness via cues like familiarity, similarity, and co-residence — and in modern society those cues are easily fired by shared language, hometown, comrades-in-arms, a common flag, manufacturing fierce loyalty toward "fictive-kin" groups. Tribalism and nationalism are, at root, the kin circuit misfiring on non-relatives. (3) It demystifies selflessness without devaluing it: understanding the mechanism lets us consciously extend the circuit to those it wouldn't otherwise cover.
Worker bees don't reproduce, yet labor tirelessly to raise the queen's offspring — long a puzzle of altruism. Kin selection answers it: the haplodiploid genetics of bees make sisters related at r ≈ 0.75 — more effective at propagating their own genes than reproducing themselves would be. So "forgo reproduction, raise sisters" is actually the optimal solution at the gene level.
(1) The gene's-eye view is especially congenial to your distributed-systems background: treat a gene as a selfish "data replica," with organisms as nodes that copy and propagate it — strikingly isomorphic to the "data-centric, nodes-as-vehicles" view in distributed systems. (2) The most practical modern corollary is seeing through fictive-kin triggers: a company calling itself "one family," a brand addressing you as "fam," both invoke your kin-loyalty circuit — essentially borrowing an r value on non-relatives. Spotting it lets you tell which loyalties are worth giving from which are hijacked. (3) Parenting: a parent's near-unconditional, enormous investment in a child (r = 0.5) is kin selection's most intense expression — understanding this doesn't cheapen love; it reminds us that this circuit is so powerful it deserves to be consciously, and generously, extended beyond blood.