Over-relying on the first number or fact encountered, then adjusting insufficiently from that starting point.
Imagine you spin a wheel and it lands on 50, then someone asks you how many jelly beans are in a jar. Even though the wheel has nothing to do with jelly beans, your guess will magically drift toward 50. It's like your brain grabs onto the first number it sees and then uses it as a starting point, only scooting a little bit away — like a dog on a leash that can wander, but not very far from the post it's tied to.
Anchoring bias occurs when an individual's numerical estimate or qualitative judgment is disproportionately influenced by an initial reference point, even when that reference point is arbitrary or irrelevant to the question at hand. The effect is remarkably robust: it persists across domains ranging from trivia questions to multi-million dollar real estate transactions, and it affects both novices and trained experts. People generate their final estimate by starting from the anchor and adjusting — but the adjustment is almost always insufficient, leaving the final answer pulled toward the anchor. Critically, simply being aware of the bias or being motivated by financial incentives does not reliably eliminate it, making it one of the most difficult cognitive biases to counteract through willpower alone.
The same glitch looks different depending on the terrain. Finance, medicine, a relationship, a team — same mechanism, different costume.
Investors anchor on a stock's purchase price or its 52-week high, causing them to hold losing positions waiting for a return to the anchor price or to undervalue gains because the stock hasn't reached a prior peak. Initial analyst price targets similarly anchor subsequent market expectations, even when fundamentals shift significantly.
Physicians anchor on a patient's initial presenting diagnosis or referral impression, causing them to insufficiently adjust their diagnostic reasoning when new, contradictory evidence emerges. Initial lab values or vital signs can also anchor clinical expectations, leading to delayed recognition of deterioration or improvement.
Amos Tversky and Daniel Kahneman, 1974, in their landmark paper 'Judgment under Uncertainty: Heuristics and Biases' published in Science. The concept was further elaborated by Epley and Gilovich (2006) who distinguished between self-generated and externally provided anchors.
In ancestral environments, quick estimation under uncertainty was essential — judging the size of a herd, the distance to a predator, or the amount of food in a patch. Using a nearby reference point and adjusting from it was a fast, energy-efficient heuristic that produced 'good enough' answers most of the time. This shortcut conserved scarce cognitive resources while enabling rapid decisions in time-pressured survival situations.
Machine learning models can exhibit anchoring-like behavior when initial training data or hyperparameter values disproportionately influence the final model. In AI-assisted decision-making, when algorithms present a suggested value (e.g., a predicted price, risk score, or diagnosis), human operators anchor on that suggestion and fail to sufficiently override it even when they have access to contradicting domain knowledge. Recommendation systems anchor users on initially presented items, shaping downstream browsing and purchasing behavior.
This page taught you the name. The deck turns the name into reflex. 1,100+ swipeable scenarios, 1,100+ defenses, 650+ detection prompts — spaced-repetition Swipe Deck, unlimited Spot-the-Bias Quiz, Defense Playbook, Pre-Flight, My Blindspots, Cheat Sheets, Field Guide e-book. $39.53$59.
