In short
The curse of knowledge is the difficulty that better-informed people have in imagining the perspective of people who know less, so that they assume their own knowledge is shared and overestimate how much others understand. It shows up in teachers who cannot see why a lesson is hard, in experts who write instructions nobody can follow, in sellers who assume buyers see what they see, and in anyone who has tapped a tune on a table and been amazed that the listener could not name it.
How the curse of knowledge works
Once you know something, the knowledge becomes part of how the world looks to you, and you cannot switch it off to see the world without it. A tapper hears the melody in her head while her fingers tap; the listener hears knocks. The tapper cannot stop hearing the melody, so the knocks seem obviously to carry it, and her estimate of what the listener will get starts from her own experience and adjusts too little. The curse is this failure to adjust: better-informed people do allow for what others do not know, but the correction falls short.
Camerer, Loewenstein and Weber gave the effect its name in 1989, in an economics paper that asked whether people who know something can reproduce the judgment of people who do not, when it pays them to. They could not, fully: the informed participants' predictions were pulled toward the figures they knew, even in trading markets with money at stake. More information, the authors concluded, is not always better for the person who has it.
The same mechanism runs through several better-known findings. Fischhoff's 1975 hindsight studies showed that people told the outcome of an event judged it to have been more predictable than people who were not told. Keysar and Henly showed in 2002 that speakers overestimate how often listeners catch their intended meaning, while overhearers told the intention did not. Birch and Bloom showed in 2007 that adults who knew where an object had been moved gave a higher probability that an uninformed person would look there than adults who did not know. In each case irrelevant knowledge leaked into the judgment.
The curse does not fade with expertise; it grows with it. The more you know about a subject, the larger the gap between your picture and a beginner's, and the less able you are to see what the beginner's lacks. That is why the people worst placed to write an introduction are often those who know the field best.
The original study
The term comes from Colin Camerer, George Loewenstein and Martin Weber, whose paper "The curse of knowledge in economic settings: An experimental analysis" appeared in the Journal of Political Economy in 1989. In the first stage, 51 Wharton students were given reports on eight real companies, with a decade of earnings figures, and asked to forecast each company's 1980 earnings, with a payment for accuracy. In the second stage, two months later, other participants were told the companies' actual 1980 earnings and asked to estimate what the first group had forecast. In individual judgments and then in experimental markets whose assets paid off according to the first group's forecasts, the informed participants' estimates leaned toward the earnings they knew. The markets cut the bias by roughly half but did not eliminate it.
The demonstration everyone remembers is older and simpler. In experiments for her 1990 Stanford dissertation, supervised by Lee Ross, Elizabeth Newton paired people as tappers and listeners. Tappers chose a well-known song from a list of 25 and tapped its rhythm on a table; listeners tried to name it. Before the listener answered, the tapper predicted the chance that the listener would get it right. Tappers predicted success about half the time. Listeners named about 2.5 percent of the tunes, 3 in 120. The tappers could hear the song as they tapped. The listeners heard knocks and had no way to supply the tune that made them meaningful.
Read the diagram and evidence as text
A tapper’s thought bubble contains a musical phrase. A listener has only a row of knocks, geometrically identical to the tapper’s knocks. Shared 0–100% rails show about 50% predicted tune identification versus about 2.5% observed identification. People, notation and tap marks are illustrative.
Will the listener name the song?
Share of listeners naming the tune, predicted against observed
Same rhythm tapped on a table; only the knowledge of the song differs
Tapper: knows the song and hears the melody while tapping: Predicted about 50% of listeners would name it
Listener: hears only the knocks: About 2.5% named it: 3 in 120
| Condition | Value and unit | Evidence quote |
|---|---|---|
| Tappers' predicted share of listeners naming the tune | 50 percent | people who tapped the rhythm of a well-known song on a table predicted that about 50 percent of listeners would name it |
| Listeners who named the tune | 2.5 percent | listeners named about 2.5 percent of the tunes, 3 in 120 |
| Listeners' correct answers | 3 tunes named | 3 in 120 |
| All tapping attempts | 120 tunes tapped | 3 in 120 |
Newton's dissertation was not published as a journal article; the 50 percent prediction and the 2.5 percent hit rate on this page come from Griffin and Ross's 1991 review, in which Ross, Newton's advisor, described the study.
Keysar and Henly's 2002 experiments brought the finding into ordinary speech. Speakers said a sentence with two possible meanings, such as a man chasing a woman on a bicycle, to a listener who then chose between pictures. Speakers believed they had been understood 72 percent of the time; listeners chose the intended picture 61 percent of the time, and when they had misunderstood, speakers thought they had understood in 46 percent of cases.
Everyday examples
The curse appears wherever someone who knows has to communicate with someone who does not. The rows below are illustrations, not findings from the studies cited on this page.
| Situation | What the bias does | What helps |
|---|---|---|
| An engineer writes the setup guide for a product | Steps obvious to the writer are left out, and the guide is unusable by the people it is for | Have someone who has never seen the product follow the guide while you watch in silence |
| A teacher explains a concept for the tenth year | The explanation is fluent and complete to the teacher and skips the step students are stuck on | Collect the wrong answers; they show where the explanation assumes what it should be teaching |
| A manager sends a two-line message about a decision | The context that makes the two lines sensible is in the manager's head and nowhere else | Write the message as if to someone who was not in the meeting, because the reader was not |
| A seller prices a house they have lived in for years | Every feature they know about seems visible to the buyer, who sees paint and square footage | Price from comparable sales, which are made by people who did not know the house |
| A specialist gives a diagnosis in technical terms | Terms that are everyday to the specialist are opaque to the patient, and the nodding is politeness | Ask the patient to say back what they will tell their family tonight |
How to counter it
You cannot unknow what you know, so the counters replace imagination with evidence of what the other person actually understood.
- Test with a real novice. The listener in the tapping study is the only judge of whether the tune came through. Watch someone use the instructions or solve the problem without your help.
- Ask for the playback. Having the other person say what they understood, in their own words, exposes the gap before it costs anything.
- Borrow from the market, not the bonus. Camerer and colleagues found that paying for accuracy did not reduce the curse; trading in a market, where less biased traders acted more often, cut it by about half. A reviewer paid to find what readers will miss is an untested parallel.
- Assume the adjustment is too small. When you have corrected your estimate of what others know, correct it again.
- Keep records of what you did not know. Notes from when you were learning, or a newcomer's questions from last month, are the closest thing to a beginner's mind.
- Give the key before the message. A listener who knows the song's name can hear it in the taps. Telling people the point first lends them the knowledge the curse assumes they have.
What it is not
The curse of knowledge is one of a family of effects in which a person's own mental state leaks into a judgment about someone else's. Its neighbors differ in what leaks.
- It is not hindsight bias, though Camerer and colleagues built on Fischhoff's hindsight work and the two share a mechanism. Hindsight is about judging the past once the outcome is known; the curse of knowledge is about judging what another person knows once you know more.
- It is not the illusion of transparency, the overestimate of how visible your own feelings and thoughts are to others. The illusion of transparency concerns your inner states; the curse of knowledge concerns facts you hold and others lack.
- It is not the false consensus effect, the tendency to overestimate how many people share your opinions and choices. That is about agreement; the curse is about information.
- It is not egocentric bias in general, which covers every way your own perspective is overweighted. The curse of knowledge is the specific case in which what is overweighted is what you know.
Tests and tools that use this
Related biases
Curse of Knowledge FAQ
What is the curse of knowledge?
It is the difficulty that better-informed people have in imagining the perspective of people who know less, so that they assume their knowledge is shared and overestimate how much others understand. Camerer, Loewenstein and Weber named it in 1989 after showing that people told a company's actual earnings could not fully ignore them when predicting what uninformed forecasters had guessed. The tapping study, in which tappers predicted that half of listeners would name a tune and 2.5 percent did, is its best-known illustration.
What was the tappers and listeners experiment?
Elizabeth Newton, in experiments for her 1990 Stanford dissertation, paired people as tappers and listeners. Tappers picked a well-known song from a list of 25, tapped its rhythm on a table, and predicted whether the listener would name it. Tappers predicted success about 50 percent of the time; listeners named about 2.5 percent of the tunes, 3 in 120. The tapper could hear the melody while tapping and could not imagine the taps without it. The results were described by Griffin and Ross in 1991.
How does the demo on this page work?
You estimated what share of listeners would name a tapped tune and what share caught a speaker's intended meaning in the 2002 study, and the page compared each guess with the recorded figure. Guessing above the record on both counts as overestimating understanding in this illustration. Guessing at or below it on either does not. The third slider, your guess at what the tappers predicted, is for comparison and not scored. The demo measures your estimates of other people's understanding, not your own communication.
What is the difference between the curse of knowledge and hindsight bias?
They share a mechanism and differ in the target. Hindsight bias, which Fischhoff demonstrated in 1975, is the tendency to judge a past event as more predictable once you know how it turned out. The curse of knowledge is the tendency to assume that other people know what you know, or to be unable to set your knowledge aside when judging their beliefs. Camerer and colleagues built their 1989 paper on the hindsight literature and extended it from judging the past to judging other minds.
Why can experts be bad at explaining things?
Because expertise widens the gap the curse of knowledge has to bridge. An expert's picture of a subject contains structure, vocabulary and shortcuts that a beginner does not have, and the expert cannot see which parts of the picture are missing from the beginner's, because none of it feels like knowledge any more. Explanations skip the steps that have become automatic. The remedy is to test the explanation on a real novice and to keep records of what you did not know when you started.
Does the curse of knowledge go away when money is at stake?
Not in the paper that named it. Camerer, Loewenstein and Weber paid participants for accurate estimates of what uninformed forecasters had predicted, and gave some of them feedback; neither reduced the bias. Only trading in experimental markets did, cutting the curse by about half without removing it, and the authors traced that to less biased traders acting more often. Their conclusion was that more information is not always better, because better-informed people cannot set aside what they know.
References
- Griffin, D. W., & Ross, L. (1991). Subjective construal, social inference, and human misunderstanding. Advances in Experimental Social Psychology, 24, 319-359. (Reports Newton, E. L., 1990, doctoral dissertation, Stanford University.) doi.org
- Camerer, C., Loewenstein, G., & Weber, M. (1989). The curse of knowledge in economic settings: An experimental analysis. Journal of Political Economy, 97(5), 1232-1254. doi.org
- Keysar, B., & Henly, A. S. (2002). Speakers' overestimation of their effectiveness. Psychological Science, 13(3), 207-212. doi.org
- Birch, S. A. J., & Bloom, P. (2007). The curse of knowledge in reasoning about false beliefs. Psychological Science, 18(5), 382-386. doi.org
- Fischhoff, B. (1975). Hindsight is not equal to foresight: The effect of outcome knowledge on judgment under uncertainty. Journal of Experimental Psychology: Human Perception and Performance, 1(3), 288-299. doi.org
Cite this source
Psychology.com. (2026, October 21). Curse of Knowledge. Psychology.com. https://psychology.com/biases/curse-of-knowledge