Anchoring bias occurs when an initial piece of information — the 'anchor' — disproportionately influences subsequent judgments and decisions, even when the anchor is arbitrary or irrelevant. In Tversky and Kahneman's classic experiments, spinning a random number wheel before asking participants to estimate the percentage of African countries in the UN significantly shifted their answers toward the wheel's number. The bias persists because people adjust insufficiently from the anchor rather than reasoning from scratch. Anchoring affects negotiations, pricing, sentencing decisions, and any context where a starting value is presented before a judgment is made.
Run simple anchoring experiments with friends: present different starting numbers before asking estimation questions and observe how answers cluster around the anchors. Then examine real-world applications in retail pricing ('was $100, now $60'), salary negotiations, and legal sentencing to see how anchors shape outcomes.
From your study of cognitive biases, you already know that human reasoning regularly departs from the norms of formal logic and probability. Anchoring bias is one of the most thoroughly documented and practically consequential of these departures. It describes the tendency for an initial piece of information — even one that is explicitly random or irrelevant — to pull subsequent numerical estimates toward it. The mechanism is not gullibility; it appears to operate at a level below conscious deliberation.
The classic demonstration is Tversky and Kahneman's spinning wheel experiment: participants watched a wheel land on a number (rigged to land on 10 or 65), then estimated the percentage of African nations in the UN. The median estimate for the group that saw 65 was 45%; for the group that saw 10, it was 25%. The wheel was visibly random. The number was obviously irrelevant. Yet it shifted estimates by twenty percentage points. The reason appears to be that people begin from the anchor and adjust insufficiently — they move in the right direction but stop too early, leaving the anchor's influence embedded in the final answer.
What makes anchoring particularly powerful is its resistance to awareness. Unlike some biases that fade when people are warned, anchoring largely persists even when participants are told "there will be an anchor in this study and it will bias you — try to correct for it." This is unlike the correction most people imagine: you cannot simply subtract the anchor's influence because you don't know how large it was. You can adjust *some*, but you cannot introspect your way to the unanchored answer. This connects to your broader understanding of cognitive biases as features of system-one processing that aren't fully accessible to deliberate reflection.
The practical implications reach across high-stakes domains. In negotiations, the party who names a number first sets an anchor that constrains the bargaining range — knowing this, skilled negotiators deliberately anchor high before compromising. In retail pricing, "was $199, now $79" anchors on the original price to make the discount seem larger. In legal sentencing, studies show that even prosecutors' randomly assigned sentencing demands shift judges' actual sentences — a disturbing finding given that judicial decisions are supposed to be based on facts and law. In any context where an initial value is presented before a judgment, the cognitive ground is already tilted.
The corrective implications are modest but real. Because adjustment from an anchor is the mechanism, strategies that force reasoning from multiple reference points — generating arguments for why the true value might be much higher, much lower, and then synthesizing — produce less anchored estimates than simply "trying to be objective." Similarly, generating your own estimate before seeing any anchor inoculates you better than being told about anchoring after the fact. Knowing about anchoring doesn't make you immune, but it does let you design decision environments that reduce its distorting influence.
Topics in reflective domains aren't scored by quiz answers. Read, reflect, and mark when you've thought it through.