Curse of knowledge
Claim assessed: Someone who knows something finds it hard to imagine how someone who doesn't know thinks.
Published · how we work
In short
The curse of knowledge is the difficulty of setting aside what you know when you try to imagine what someone who knows less believes. The term comes from Camerer, Loewenstein and Weber (1989), who showed that people cannot ignore their extra information even when they are paid to do so. The same mechanism makes experts underestimate how hard things are for beginners. An influential experiment with adults, however, replicated with a much smaller effect than the original.
What it says
The economic experiment. Camerer et al. (1989) gave a group of 51 Wharton students financial reports on eight companies, with data up to 1979, and asked them to estimate 1980 earnings. A second group got the same reports plus the actual 1980 earnings, and had to guess what the first group had estimated. Each good guess was paid. In principle, the second group should have ignored the actual earnings, which the first group didn’t know. They couldn’t: their estimates were pulled towards the real figure. Neither payment nor learning the right answer after each attempt reduced the error. In the groups that traded in a market, the error was about half as large, but it did not disappear.
Why it matters. The authors link the effect to two others. The first is hindsight bias: once you know how something turned out, it seems more predictable than it really was. The second is passing on expertise: when you think what you know is common knowledge, it is hard to explain it to others, and it doesn’t even seem necessary.
In experts. Hinds (1999) asked experts, intermediate users and novices to estimate how long a complex task would take novices. The more people knew, the more they underestimated the time. The intermediate users estimated best. The debiasing techniques tried did not reduce the experts’ underestimation.
In adults, taking someone else’s place. Birch and Bloom (2007) used a story like those given to young children, to see whether adults can imagine what someone believes who hasn’t seen a change. Vicki puts her violin in the blue container and leaves. Her sister moves the violin to another container, then rearranges the containers, and the red container ends up exactly where the blue one had been. The participants, 155 Yale students, had to say how likely Vicki was to look first in each container. Those told that the violin was in the red container gave the red container a higher probability than those who didn’t know where it was. Those told that the violin was in a container where Vicki had no reason to look were not influenced. The authors conclude that what you know influences you mostly when you have a plausible justification for it.
The replications. Ryskin and Brown-Schmidt (2014) wanted to use Birch and Bloom’s task as a test, so they repeated it in seven experiments with 3,074 participants, students and people recruited online. In the first experiment they found no effect at all. In the others, the effect went the same way but was much smaller. Combining all the experiments, the knowledge effect was d = 0.20, less than half the estimate for the original study (0.47). The plausibility effect was 0.24, under a third of the original (0.65). The authors calculate that the original study would have had only about a 17% chance of detecting an effect of 0.20, and note that an effect this size may matter little in real life.
Example
An experienced programmer writes the installation instructions for an app. The steps seem obvious, so they skip “open the terminal” and “check the version”. A beginner gets stuck exactly there. The example is an illustration, not a study cited here.
How to apply it
The steps below are a practical way of applying it that we propose, based on the cited studies.
- Don’t rely on the feeling that something is obvious. The feeling comes precisely from what you already know.
- Test it on someone who doesn’t know. In Camerer et al.’s study, neither payment nor learning the right answer corrected the error. It is safer to see directly how someone without your knowledge copes.
- Ask someone at an intermediate level. In Hinds’s study, intermediate users estimated better than experts how long a task takes beginners.
- When judging a decision by its outcome, remember what was known at the time. It is the same error, applied to your own past.
Limits and nuances
- The violin test replicated with a small effect. The knowledge effect in adults exists, but in this version it is small, and the role of plausibility is weaker than was thought.
- The economic study is small and old. The markets had nine participants each, and the groups with payment and feedback had 19 and 13 participants.
- We couldn’t read some well-known studies. The often-cited “tappers” experiment, in which someone taps out the rhythm of a song and estimates how many listeners will guess it, comes from an unpublished doctoral thesis (Newton, 1990). We haven’t read it, so we don’t describe it here.
- Correcting it is hard. Hinds (1999) notes that the experts resisted debiasing techniques. Camerer et al. found that payment and feedback didn’t help.
- What we read. Camerer et al. (1989), in the copy on the Carnegie Mellon site, Birch and Bloom (2007), from the Yale lab’s site, and Ryskin and Brown-Schmidt (2014) we read in full. For Hinds (1999) we read the abstract.
Sources
- Colin Camerer, George Loewenstein, Martin Weber (1989). The curse of knowledge in economic settings: An experimental analysis
- Pamela J. Hinds (1999). The curse of expertise: The effects of expertise and debiasing methods on prediction of novice performance
- Susan A. J. Birch, Paul Bloom (2007). The curse of knowledge in reasoning about false beliefs
- Rachel A. Ryskin, Sarah Brown-Schmidt (2014). Do adults show a curse of knowledge in false-belief reasoning? A robust estimate of the true effect size
See also: Dunning-Kruger effect, Fundamental attribution error, Hindsight bias