Anchoring Effect — Judgment Bias Simulation

What is this?

The Anchoring Effect is one of psychology's most reliable findings: when people are shown an arbitrary number before making an estimate, their final answer is disproportionately pulled toward that number — even if they know it's random.

In 1974, Tversky and Kahneman showed that people asked "Is the percentage of African countries in the UN higher or lower than 65?" gave answers averaging 45%. Those anchored to 10 guessed 25%. Same question. Wildly different answers. This simulation runs that experiment on 200 virtual agents with varying levels of domain expertise, so you can see exactly who gets fooled and by how much.

How to read the charts

Histogram (top): The distribution of all 200 agents' estimates. The gray "ghost" shows where they were before the anchor. The shift tells you the anchor's pull.
Red line = Anchor. The random number the wheel produced. Notice how the histogram bulges toward it.
Blue line = Mean estimate. The crowd average. Compare it to the green dashed "Actual" line (₹50,000) to see the adjustment gap.
Scatter plot (bottom): Prior knowledge (X) vs estimate error (Y). Dots near the zero line are well-calibrated. Dots far above/below (especially low-knowledge ones, left side) are being dominated by the anchor.
Try: Drag the knowledge slider all the way left. Spin. Watch the histogram collapse toward the anchor. Then drag right and spin again — experts resist much more.

Anchor —

Mean Estimate —

Actual Value ₹50,000

Adjustment Gap —

% Anchored —

Distribution of Agent Estimates — 200 Agents (gray = pre-anchor baseline)

Prior Knowledge vs. Estimate Error — Who Gets Fooled Most? (green = accurate · red = badly anchored)

The Adjustment Gap: Psychologists find that people anchor on the given number and then "adjust" toward what seems plausible — but always stop too soon. So a high anchor produces a too-high estimate and a low anchor produces a too-low one. The gap between the mean estimate and ₹50,000 (the actual value) is the adjustment gap: how far people failed to move away from an arbitrary starting point. Low-knowledge agents show the widest gap; high-knowledge agents partially resist, but almost no one is immune.