Social Influence BriefingA daily communication practice for people who influence through ideas

The Curse of Knowledge: Explaining What You Already Know

Published · Last reviewed

Before you explain something you know well, write down what a newcomer would not know: the terms, the skipped steps and the background you assume. Then test the explanation on one real newcomer by asking them to say it back or try the first step. Experiments since 1989 show that knowing something makes it hard to judge what others know, even when money is at stake.

Durable practice (not new). This is established communication practice, not a discovery or a trend.

What it fixes

Experts skip steps, terms and background that have become invisible to them. The learner meets the gap, stops following and rarely says so. The result is a lesson, tutorial or talk that only people who already understood it can follow.

How to run it

  1. Name the newcomer: one real person, or a precise description such as “someone who has finished the first week of the course”.
  2. Write three short lists: terms they have not met, steps you do without thinking, and background you assume they share.
  3. Rewrite your opening so every item on those lists is defined, shown or cut.
  4. Give the explanation to one real newcomer, then ask them to say it back in their own words or to try the first step while you watch.
  5. Fix the exact place where they stalled before you polish anything else.

“I know this topic well, which makes it easy for me to skip things. Tell me back what you would do first, and stop me at any word I have not explained.”

Keep each list short. Three to six items per list is enough to change the first minute of an explanation. The aim is to catch the gaps you would otherwise leave, and a longer list rarely adds more than a few extra gaps.

Say it out loud

Announce what you are doing and why. The line above does it. It still works when the listener knows the purpose, because the purpose is to ask for their help: it invites them to correct you and gives them permission to say “I lost you.” Keep the check about your explanation so that the listener is never the one being tested.

Profile translations

How to tell it worked

Action
Write the three lists before you draft the explanation.
Why
Gaps are easier to see when you are looking for them on paper than when you are already speaking.
Verification
Your lists contain at least one term or step you would have skipped without them.
Action
Ask one newcomer to say it back, in their words, after you have explained.
Why
A nod shows politeness. A restatement shows what actually arrived.
Verification
They name the first step correctly, or their restatement shows you exactly which part did not arrive.
Action
Repair the stall point, then try it on a second newcomer or the same one the next day.
Why
A fix is only a fix if the same gap no longer appears.
Verification
They complete the first step without asking you a question, or they ask about a new point instead of the old one.

Failure signs: stop if you see these

  • The newcomer nods throughout but cannot restate the first step.
  • You test the draft on a colleague in your own field, so nothing stalls.
  • The newcomer says “I know that” more than once, which signals the explanation is now talking down to them.
  • The lists keep growing and the explanation starts to bury the one thing the newcomer needs.

Why it works, and how strong the evidence is

Evidence. Experimental, with one earlier line of work and a later partial replication. The foundational paper is a set of laboratory market experiments; the second study is a laboratory reasoning task with adults. Neither tests the teaching practice described here. The page's steps are practical reasoning from their findings. Effect sizes for explanations that follow this practice: Not reported.

Camerer, Loewenstein and Weber (1989, Journal of Political Economy) tested a common assumption in economics: that better-informed people can reproduce the judgments of less-informed people. In their experiments, informed participants were unable to ignore what they knew when predicting what less-informed forecasters had estimated (the forecasters were 51 students estimating company earnings). More information could hurt the informed person. Trading in a market, with money at stake, reduced the bias by approximately half without removing it, and the authors report that feedback alone had little effect.

Birch and Bloom (2007, Psychological Science) found a version of the same bias in adults reasoning about what another person believes. When adults knew how an event ended, that knowledge compromised their judgment of what someone else believed about it, and how plausible the event seemed affected the size of the bias. A later set of three experiments with a related false-belief task (Farrar and Ostojic, 2018, PLOS ONE) reported that the bias appeared reliably when adults judged the beliefs of someone in their own group.

What the research does not show is that listing a newcomer's gaps improves anyone's understanding. The studies establish that the bias is persistent and that incentives and feedback only partly reduce it. The practice here follows from that: if you cannot reliably judge what a newcomer knows, check against a newcomer.

Worked example

An illustration, not data. A software trainer opens a video on automated testing.

Before

“Today we will set up your pipeline. Make sure your runners are configured and your config file passes validation.”

After

“Today we will set up automatic checks that run each time you save changes to a project. You need two things first: a project stored on GitHub, and a text file called checks.yml in its main folder. I will show both on screen. Pause the video if you are missing either.”

The trainer's three lists produced the changes. The terms “pipeline” and “runners” were replaced with what they do; the step of creating the config file, which the trainer did automatically, was added; and the assumption that the viewer already had a hosted project was stated. A newcomer then watched the first minute and was asked to say what to do first.

Watch for the opposite error: if the newcomer says they already knew the new material, cut it. The lists exist to remove gaps, not to pad the opening.

Frequently asked questions

What is the curse of knowledge?

It is the difficulty that better-informed people have in ignoring what they know when judging what others know. Camerer, Loewenstein and Weber (1989) documented it in laboratory market experiments, and Birch and Bloom (2007) found it in adults' reasoning about another person's beliefs.

Does a financial incentive or feedback remove the curse of knowledge?

Not entirely. In the 1989 experiments, market conditions reduced the effect by approximately half without removing it, and the authors report that feedback alone had little effect.

Is there proof that testing an explanation on a newcomer improves learning?

No study cited on this page tests that exact step. The practice follows from the finding that people cannot reliably judge what others know, so it asks you to check against a real person. Treat it as sound practice with indirect support.

How do I find a newcomer to test on?

Anyone outside your field will do: a colleague from another team, a new hire, a friend or a family member. A single person is enough for a first pass, and the aim is to find the first place they stall.

What if the explanation is for a mixed audience?

Write for the least-informed person who will realistically be there and offer the advanced material as an optional extra. Say clearly where it begins, so experts can skip to it.