Cognitive Biases

Confirmation bias is the most deeply rooted of human cognitive defects: we seek out, prefer, and remember information that supports what we already believe, while overlooking or discounting the evidence that contradicts it. This isn't laziness — even when you're actively researching, the bias quietly warps how you collect and interpret information. The neural mechanism rides the dopamine reward system: confirming information triggers a small pleasure spike; disconfirming information triggers an amygdala threat response. Confirmation bias is, at the physiological level, "reward the bias, punish the truth."

The non-trivial insight: the most insidious mode of confirmation bias isn't rejecting opposing evidence — it's distorting the search strategy itself. You don't intentionally ignore contrary information; your search terms, choice of sources, and attention allocation are already biased toward confirmation before you start reading. In the filter-bubble era, recommendation algorithms are a force-multiplier for this. Worse, confirmation bias is self-sealing — the people most under its influence are the ones least likely to believe they have it.

Practice: ① For each important judgment, set up a "red team" — assign a person (or AI) whose sole job is to surface contrary evidence; ② When searching, deliberately use opposite keywords (bullish on AI? search "AI bubble" / "AI overvalued"); ③ Build a steel-manning habit — before critiquing the opposing view, restate it in its strongest possible form. Make sure you're attacking the strongest version of their argument, not a straw man.

Confirmation Bias is the tendency to search for, favor, and remember information that confirms our existing beliefs — while discounting contradictory evidence. It operates unconsciously even in active research, and is dramatically amplified by algorithmic recommendation systems. Countering it requires deliberately seeking out the strongest opposing arguments, not just acknowledging they exist.

The anchoring effect describes how people, when making a judgment, over-rely on the first piece of information they encounter — the "anchor" — as a reference point. Even when the anchor is random and irrelevant, it measurably shifts every subsequent estimate. Two psychological channels are at work: "anchor and adjust" — the brain starts from the anchor and makes adjustments that are almost always insufficient; and "selective accessibility" — the anchor activates memories and information consistent with itself, pulling later judgments in its direction.

The non-trivial insight: anchoring doesn't only operate on "uncertain estimates" — it also operates on "judgments you feel sure about." Even seasoned experts — real estate appraisers, judges, doctors — are demonstrably shifted by random anchors in controlled experiments. Which means "I'm experienced, so I'm not anchorable" is itself a dangerous flavor of overconfidence. In negotiation, whoever quotes first holds a structural advantage. In investing, a stock's all-time high becomes a hidden anchor, and "40% off the peak" is a zero-information sentence that nonetheless drives behavior.

Practice: ① Before exposure to any external anchor, form your own independent estimate range — "what do I think this is worth?" must be answered before you see the asking price; ② In negotiation, if you can't quote first, at least explicitly reject the first anchor's reference value before resetting the frame; ③ For investments, replace historical price with an intrinsic-value model (DCF, etc.) as your reference — price is somebody else's anchor; value is your own judgment.

The Anchoring Effect describes how the first piece of information we encounter disproportionately influences our subsequent judgments — even when that information is arbitrary or irrelevant. In negotiations, pricing, and investment decisions, whoever sets the anchor gains a structural advantage. The defense is to form an independent estimate before encountering any external anchor.

The core mechanism of survivorship bias is the asymmetric extinction of data: failures systematically disappear from the observable sample, leaving us to infer general laws from a heavily filtered subset. Evolution actually reinforced this — humans innately attend to "what's alive, present, and audible," because in the ancestral environment, the absent didn't matter for survival decisions. In a modern information environment, the same instinct makes us systematically overestimate the efficacy of "successful" strategies and underestimate the role of randomness.

The non-trivial insight: the real danger of survivorship bias isn't "you can't see the failures" — it's that it flips the direction of causality. You think A caused success, but in reality success filtered for samples that happened to have property A. Successful entrepreneurs are mostly "high risk tolerance," not because risk-taking caused success, but because the conservative founders already exited your sample. Which means: the "strategies" extracted from success cases are often by-products of the survival filter, not real causal drivers.

Practice: ① For any case study, force yourself to find a counterexample — someone who used the same strategy and failed. One case can't prove anything, but one counterexample can refute a lot; ② For any statistic, ask: "what was the inclusion criterion for this sample? who got excluded?"; ③ Build a "graveyard" habit — before investing, study the companies in the same sector that died; before founding, study the failed founders in the same model; before adopting a learning method, find the contexts where it didn't work.

Survivorship Bias occurs when we draw conclusions from visible survivors while ignoring the invisible failures. It makes successful strategies look more reliable than they are, because the failures using the same strategy have disappeared from our dataset. The corrective is to actively ask: "What would the full distribution of outcomes look like, including the cases I'm not seeing?"

The sunk cost fallacy is continuing a course of action you should abandon, simply because of past time, money, or effort already invested. Rational decisions should weigh only future costs and benefits; sunk costs are unrecoverable and should not influence forward-looking choices. The psychological roots aren't just "I can't bear to walk away" — it's a double hostage situation: loss aversion (quitting confirms the loss) plus self-consistency (quitting concedes the past decision was wrong, threatening self-image).

The non-trivial insight: the most dangerous form of sunk-cost reasoning isn't "I don't want to lose the money" — it's "I don't want to lose the narrative." After heavy investment in a project, you've sunk not just resources but a whole story about why this is worth doing. Quitting demolishes that narrative — and humans will keep losing money rather than admit the story was wrong. That's why organizational sunk-cost behavior is worse than individual: every reporting layer is protecting its own narrative.

Practice: ① For any "continue or stop" decision, run the reset test: "If I were starting fresh today with no commitment to this project, would I choose to start it?" If the answer is no, the only reason to continue is sunk cost; ② Set a stop-loss line up front — "if by date X we haven't hit metric Y, we terminate" — defined before emotion can hijack the call; ③ Don't decide while in a losing state — when emotion is highest, decision quality is lowest. Let it cool 24 hours and re-evaluate.

The Sunk Cost Fallacy is the irrational tendency to continue an endeavor because of past investments — time, money, or effort — rather than future prospects. Rational decision-making should be purely forward-looking: sunk costs are gone regardless of your next action. The key question is always "what are the future costs and benefits from this point forward?" — not "how much have I already spent?"