Anchoring: the first number wins

Lesson 2 · Thinking Fast and Slow × AI · about 10 minutes

Mission: stop AI outputs and my own estimates from being pulled toward a number I supplied.

The core idea

Anchoring is the pull an initial number or suggestion exerts on later estimates, even when that number is irrelevant. Kahneman's book covers the anchoring experiments in its Part II heuristics discussion (see Sources).

It works on AI in two directions. When your prompt contains a number, a draft, or a conclusion, the model's answer tends to drift toward it. And when the AI gives you an anchor first, your own judgment drifts toward it.

What the research shows

A 2025 benchmark tested anchoring across tiny, standard, advanced, and reasoning models, including GPT-4o, DeepSeek-R1, and Qwen3. Anchoring appeared in all of them, and the mitigations it tested reduced the effect only modestly, by up to about 10 percent. Reasoning was the exception and helped somewhat (Huang et al., arXiv 2505.15392). A separate mechanistic study on smaller open models found anchors reweight the output distribution, and its effects varied across prompt designs (Valencia-Clavijo, arXiv 2511.05766, a preprint).

Practice

Choose the bias the scenario shows.

1. You ask the AI, "Is the average price above $400,000?" Then you ask, "Estimate the average price." Its estimate lands near $400,000. You have seen this pattern before.
Correct. The first number pulled the estimate toward it.
Not quite. The earlier number set the starting point for the later estimate. That is anchoring.
2. You paste in your own $80,000 budget and ask, "Is this reasonable?" The AI's revised range sits close to $80,000.
Correct. Your draft number is an anchor the model will drift toward.
Not quite. Asking the model to check your number invites it to adjust toward that number.
3. You want an unbiased estimate from a colleague and from the AI. Which order reduces anchoring?
Correct. Forming your own estimate first keeps the AI's number from setting your starting point.
Not quite. Forming your estimate before seeing anyone else's keeps your starting point independent.

Takeaway

Anchors in your prompt steer the answer, and anchors in the answer steer you. Form your own rough estimate before you ask, and ask for a range rather than a single number.

Try it this week: ask the AI for an estimate with no number in the prompt, then compare it to one you gave it. Note the gap.

Sources

Questions? Stuck on any idea? Ask your teacher in this conversation. Previous: Lesson 0001. Next: Lesson 0003, availability.
← Previous Next →