Experiment 14 · Famous experiments
The spinning wheel that moved people's guesses
A number from a rigged wheel, which everyone knew was random, pulled people's estimates toward it.
- Amos Tversky and Daniel Kahneman
- Hebrew University of Jerusalem, Israel
- 1974
- 3 min read
On this page
Before you readWhat would you do?
You want an AI tool's help to set a price for a used motorbike you are selling. Which prompt is least likely to be pulled by an anchor?
Pick an answer now. The file tells you whether you were right.
01The file in one minute
1A random number
A wheel set to stop at 10 or 65 shifted people's estimates of African countries in the UN from 25% to 45%.
2Experts too
Estate agents' valuations followed the asking price, and judges' sentences followed demands made by dice.
3Why it works
People start from the number in front of them and adjust too little.
4What it means for AI
A number in your prompt can anchor the answer. Ask without one, then compare.
02Timeline
September 1974
Tversky and Kahneman describe the wheel-of-fortune anchor in the journal Science.
1987
Northcraft and Neale show listing prices sway estate agents' house valuations.
2006
Englich, Mussweiler and Strack find dice-rolled demands affect judges' sentences.
03The experiment
In their 1974 paper on judgment under uncertainty, the psychologists Amos Tversky and Daniel Kahneman described a simple demonstration. A wheel of fortune, secretly set to stop at either 10 or 65, was spun in front of each participant.
People were then asked whether the percentage of African countries in the United Nations was higher or lower than the number on the wheel, and then to give their own estimate. Those who saw 10 gave a median estimate of 25%. Those who saw 65 gave a median of 45%. The number was plainly random, yet it moved the answers.
Later studies found the same pull in expert settings. In 1987, Gregory Northcraft and Margaret Neale showed that the asking price on a house changed estate agents' valuations, though most agents said the price had not influenced them. In 2006, Birte Englich, Thomas Mussweiler and Fritz Strack found that experienced German judges gave longer sentences after seeing a higher sentencing demand, even when that demand came from a roll of dice.
04What it showed
When people have to estimate something uncertain, they often start from whatever number is in front of them and adjust, and they usually stop adjusting too early. The first number frames the search for an answer. Telling people about the bias helps only a little.
Prompts carry anchors too. Ask a chatbot whether a price of 500 million dong is fair, and its answer may orbit that figure. Ask an AI to estimate how long a project will take after you mention three weeks, and three weeks may come back. The number you write shapes the number you get.
The idea behind this caseLesson: The four-part prompt →
05What it means when you use AI
- Make your own estimate before you look at anyone else's number, including an asking price or an AI's guess.
- When you want an estimate from a chatbot, ask without giving a number first, then compare with yours.
- In negotiations, expect the first offer to act as an anchor, and prepare your own range before talks begin.
- Ask for a range with reasons, such as a low, likely and high case, instead of a single number.
07 — Sources
- Judgment under uncertainty: Heuristics and biases · Science · 1974-09-27
- Playing dice with criminal sentences: The influence of irrelevant anchors on experts' judicial decision making · Personality and Social Psychology Bulletin · 2006