Curse: Expertise can make explanation harder. Once you understand something well, it becomes difficult to reconstruct what confusion felt like, which conceptual leaps needed scaffolding, and which terms need defining. This does not mean experts are inherently poor teachers: effective teaching requires skills beyond subject expertise, and experts who have those skills, and who get feedback, teach well.
The curse of knowledge is the difficulty of setting aside what you know when reasoning about what someone else knows. It is a well-documented bias rather than an absolute inability, and its size varies with the task, the audience, the feedback available, and how recently you learned the material yourself.
What the Curse Is, and What It Is Not
Several related ideas get collapsed together. They overlap, but they are not interchangeable, and they call for different responses.
| Concept | What it refers to |
|---|---|
| Curse of knowledge | Your own knowledge biases your estimate of what others know |
| Curse of expertise | Expertise specifically impairs predictions of novice performance or learning time |
| Illusion of transparency | You overestimate how visible your internal states are to others |
| False consensus | You overestimate how widely others share your beliefs or behavior |
| Tacit knowledge | Knowledge that is hard to put into words at all |
| Cognitive load | Processing demands placed on limited working memory |
This article is about the first of these. The others appear where they genuinely apply, rather than being folded into one explanation for every communication failure.
What the Research Actually Shows
The tapping study
In 1990, Stanford doctoral candidate Elizabeth Newton ran an experiment simple enough to conduct in a living room.[1] Participants were divided into tappers and listeners. Tappers were given well-known songs, Happy Birthday and similar standards, and tapped out a rhythm on a table. Listeners tried to name the tune from the taps alone.
Before tapping, each tapper estimated what proportion of listeners would succeed. They predicted roughly 50 percent. The actual identification rate was about 2.5 percent.
The explanation is not that listeners were inattentive. When a tapper knocks out the rhythm of Happy Birthday, she simultaneously hears the full melody in her head: the rise and fall of the notes, the harmony, the words. The taps arrive accompanied by a rich internal soundtrack that makes them feel obviously recognisable. The listener has only the knocks.
This is a vivid illustration of how private knowledge biases predictions about what another person perceives. It is one demonstration, not proof that experts in general cannot imagine a novice's experience.
Imputing your own knowledge to others
Raymond Nickerson's 1999 review in Psychological Bulletin gathers the broader evidence: people have a persistent tendency to impute their own knowledge to others, often incorrectly.[2] People can partially adjust, ask questions, build models of their audience, and improve with feedback. The bias is systematic; it is not a wall.
Economic settings
Camerer, Loewenstein, and Weber introduced the term in a 1989 Journal of Political Economy paper examining economic decisions under asymmetric information.[3] Better-informed participants had difficulty reasoning as though they lacked information they possessed, which affected their predictions about less-informed parties.
The experiment demonstrated the bias in an economic setting and suggests relevance wherever people must reason about differently informed parties. One study does not establish measured effects across negotiation, law, and management, and this article does not claim it does.
Predicting novice difficulty
Pamela Hinds (1999) had people with varying expertise at a task predict how long true novices would take to learn it.[4] Experts underestimated novice learning time and overestimated novice performance, even when explicitly warned to adjust for their own expertise. Debiasing instructions helped less than one might hope.
Her results also did not worsen in a simple straight line with expertise: calibration differed across expertise levels in ways that a single study cannot fully explain. Why that happened warrants caution rather than a confident mechanism.
What this shows is that experts can misestimate novice difficulty, which is one mechanism that may impair instruction. It measured predictions, not the quality of every explanation an expert produces.
Why Expertise Creates Explanation Gaps
Chunking. Chase and Simon's 1973 chess research showed that experienced players perceive meaningful configurations rather than individual pieces, and that their recall advantage largely disappears for randomly arranged boards.[5] A knowledgeable player may recognise a familiar opening pattern where a beginner sees separate pieces on separate squares. The same compression happens in any practised domain.
This creates significant explanatory challenges. The expert often articulates the compressed version, the endpoint of learning rather than its pathway, and the beginner has no way to unpack it.
Automaticity. Steps that once required conscious effort stop being available for inspection. This is what makes expertise efficient and what makes narrating it hard.
Missing prerequisites. Experts project their current mental models onto their audience, assuming background knowledge and inferential connections the audience lacks. This projection can occur automatically, particularly when the communicator has not actively assessed what the audience already knows. The explanation feels clear to its author because, to its author, it is.
Specialised vocabulary. Jargon compresses meaning that the audience may not yet have built. Within a shared community it is genuinely useful for precision; across a knowledge gap it becomes an obstacle.[12]
Where It Shows Up
Teaching and training
Concepts get introduced before their prerequisites, and confusion gets attributed to the learner. When an audience struggles, it is easy to conclude they lack preparation, attention, or aptitude. Any of those may be true. Explanation quality is one possible cause that experts may underweight, alongside audience preparation, material difficulty, and accessibility.
Technical documentation
Documentation can fail when it is organised around implementation knowledge rather than user goals: undefined terms, missing steps the author considers obvious, error messages that describe system state rather than what the user should do, and abstract descriptions without concrete examples.
Many developers write excellent user documentation, and many projects employ technical writers, so this is a failure mode rather than a property of developers. Documentation problems also stem from search difficulty, weak information architecture, missing troubleshooting guidance, and time pressure, not from the curse alone.
Product and interface design
Designers know the system's architecture, state model, and edge cases, which makes it difficult to experience the interface as a first-time user would. They can observe users, run tests, and inspect analytics; the difficulty is that prior knowledge biases interpretation. Don Norman's distinction between a user's mental model and the system's implementation model is the standard framing here, and effective interfaces respect the former even where it diverges from the latter.[11]
What Does Not Work
Adding more detail. Beginners may genuinely need more information, but adding detail without an organising structure often increases confusion rather than reducing it. Working memory is limited, and unnecessary presentation demands make unfamiliar material harder to process.
Only removing technical words. Stripping vocabulary while leaving the conceptual gaps in place produces prose that sounds accessible and still does not land.
Relying on self-assessment. Asking "does that make sense?" is weak evidence. People may not recognise their own gaps, and may hesitate to say so.
Testing with knowledgeable colleagues. Reviewers who share your background will fill your gaps automatically and report that the explanation is clear.
Reaching for an analogy without checking it. Analogies help when the relevant structural correspondence is clear and their limits are stated. Gentner's structure-mapping work distinguishes surface similarity from structural similarity: an expert may pick an analogy for a deep relationship while a beginner takes away the surface features.[6] State what corresponds, say where the analogy breaks, and return to the real mechanism.
How to Reduce It
The curse is hard to overcome by introspection, because the thing you are trying to inspect is exactly what has become invisible to you. Procedural checks work better than trying harder.
- Define the target audience. Not "a beginner" in the abstract, but the actual people: what they already know and what they need to do afterward.
- List assumed prerequisites. Write down every term, model, and causal step your draft assumes. The unexplained items are where the audience will lose the thread.
- Observe a representative reader. Watch someone from the intended audience work through the material and say what they are thinking. Patterns of confusion across several people matter more than one person's reaction.
- Ask for teach-back or application. Have them explain it back in their own words, apply it to a new case, or predict what the model implies.[8] Repeating your words back is weaker evidence than it looks.
- Record recurring confusion. Keep the questions that keep coming up. This is the institutional version of the fix: teaching teams that document common mistakes carry that knowledge into the next revision.
- Revise and retest. One pass rarely resolves it.
Test with people who match the intended audience's actual knowledge and goals. Automatically recruiting complete beginners can mislead when the real audience is intermediate practitioners, managers, or specialists from an adjacent field.
Externalising what you know implicitly
Several practices force implicit knowledge into the open:
- Worked examples with narration: rather than presenting a finished solution, say what you notice and why you try what you try. Sweller and Cooper's research found worked examples helped novices learn algebra more efficiently than unguided practice, though the advantage is clearest for structurally similar problems and can reverse for more advanced learners.[7]
- Error analysis: explaining why a common mistake is tempting requires reconstructing the faulty reasoning, which approximates the beginner's state.
- Constraint articulation: say what cannot be done and why. Experts internalise these boundaries and forget to mention them.
- Structured elicitation: cognitive task analysis comprises several methods developed across human-factors, instructional-design, and naturalistic-decision-making traditions, including structured interviews, retrospective probing, critical-decision methods, and think-aloud procedures. The critical incident technique, which asks practitioners to walk through specific demanding cases, traces back to John Flanagan's mid-century work.[9] Combine observation with elicitation rather than relying on introspection alone, since experts may omit automatic or tacit steps when describing their own performance.[10]
The Feynman technique
A widely used self-check, named after the physicist and popularised well after his death, turns the problem into a diagnostic:
- Write the concept's name at the top of a blank page.
- Explain it in plain language, without jargon, as if to someone new to it.
- Mark where your explanation goes vague, circular, or jargon-dependent.
- Go back to the source, then explain it plainly again.
The value is that the points where simple explanation fails tend to be the points where understanding is thinner than it felt. It is a practical heuristic rather than a validated intervention.
Progressive disclosure
Reveal complexity gradually: introduce terminology after the concept it names, provide expansion points for readers who want more, and layer material for different levels. Use concrete examples early where they establish meaning, and lead with an organising principle where the audience needs orientation first. Which order works depends on prior knowledge, task, and goal.
Separating "how it works" from "how to use it" is useful, but they are layered information needs rather than a clean split. Some safe and correct use genuinely requires understanding the mechanism.
Important Nuance
A complete account needs its boundaries.
Experts talking to experts are not impaired. When both parties share a knowledge state, the inability to simulate ignorance does not matter. Compressed technical language is efficient precisely because both sides can interpret it. The curse is a problem at the interface between different levels of knowledge.
Feedback changes the picture. People who regularly learn whether they were understood, because questions are common or incomprehension has visible consequences, tend to calibrate better over time. Experts who present into silence do not.
Beginners are not automatically better teachers. Someone who recently learned something may remember the difficulty more vividly, but they may also transmit misconceptions or lack the structure to sequence material well.
Jargon is not automatically a failure. It supports precision within a community that shares it. The failure is using it across a gap without building the concepts first.
Some complexity is inherent. Writing intended for specialists is not defective because a general audience cannot follow it. That is an audience mismatch, which is a different problem from unclear writing.
Awareness alone is not enough. Knowing about the curse does not remove it. What helps is procedural: assumptions listed, explanations tested, confusion recorded. The expert who cannot imagine confusion can still measure it.
In Short
The curse of knowledge is a real and well-demonstrated bias with a narrow definition: your own knowledge distorts your estimate of what others know. It is worth resisting the temptation to expand it into a single explanation for every failure in teaching, documentation, design, and organizational life, because the remedies differ depending on what actually went wrong.
The practical response is not to know less. It is to stop relying on your own sense of clarity and check it against someone who does not share your knowledge. For related material, see five principles for clear communication and signal versus noise in communication.
Sources & Further Reading
- Newton, E. L. (1990). The Rocky Road from Actions to Intentions. Doctoral dissertation, Stanford University. (The tapping study.)
- Nickerson, R. S. (1999). "How We Know - and Sometimes Misjudge - What Others Know: Imputing One's Own Knowledge to Others." Psychological Bulletin, 125(6), 737-759. DOI: 10.1037/0033-2909.125.6.737(The core review.)
- 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: 10.1086/261651(Where the term originates.)
- Hinds, P. J. (1999). "The Curse of Expertise: The Effects of Expertise and Debiasing Methods on Prediction of Novice Performance." Journal of Experimental Psychology: Applied, 5(2), 205-221. DOI: 10.1037/1076-898X.5.2.205(Predicting novice difficulty.)
- Chase, W. G., & Simon, H. A. (1973). "Perception in Chess." Cognitive Psychology, 4(1), 55-81. DOI: 10.1016/0010-0285(73)90004-2(Expert chunking.)
- Gentner, D. (1983). "Structure-Mapping: A Theoretical Framework for Analogy." Cognitive Science, 7(2), 155-170. DOI: 10.1207/s15516709cog0702_3(Why analogies help or mislead.)
- Sweller, J., & Cooper, G. A. (1985). "The Use of Worked Examples as a Substitute for Problem Solving in Learning Algebra." Cognition and Instruction, 2(1), 59-89. DOI: 10.1207/s1532690xci0201_3(Worked examples, and their limits.)
- Chi, M. T. H., Bassok, M., Lewis, M. W., Reimann, P., & Glaser, R. (1989). "Self-Explanations: How Students Study and Use Examples in Learning to Solve Problems." Cognitive Science, 13(2), 145-182. DOI: 10.1207/s15516709cog1302_1(Self-explanation in learning.)
- Flanagan, J. C. (1954). "The Critical Incident Technique." Psychological Bulletin, 51(4), 327-358. DOI: 10.1037/h0061470(Origin of the technique.)
- Polanyi, M. (1966). The Tacit Dimension. University of Chicago Press. (Knowledge that resists articulation.)
- Norman, D. A. (2013). The Design of Everyday Things, revised edition. Basic Books. (Mental models versus implementation models.)
- Pinker, S. (2014). The Sense of Style. Viking. (The curse of knowledge in writing practice.)
Further Reading
- Kalyuga, S., Chandler, P., Tuovinen, J., & Sweller, J. (2001). "When Problem Solving Is Superior to Studying Worked Examples." Journal of Educational Psychology, 93(3), 579-588. DOI: 10.1037/0022-0663.93.3.579(Expertise reversal.)
Frequently Asked Questions
Is the curse of knowledge the same as the curse of expertise?
They are closely related but not identical. The curse of knowledge is the general bias in which what you know distorts your estimate of what others know. The curse of expertise, studied by Pamela Hinds, refers specifically to expertise impairing predictions of novice performance and learning time, and her work found that warning people to adjust for their own expertise helped less than expected. Both are distinct again from the illusion of transparency, which is overestimating how visible your internal states are, and from false consensus, which is overestimating how widely others share your beliefs.
Can beginners sometimes teach better than experts?
Sometimes, but not automatically. Someone who recently learned a subject may remember the specific difficulty more vividly and anticipate where a newcomer will get stuck. They may also pass on misconceptions or lack the structure to sequence material well. The reverse assumption, that expertise makes someone a poor teacher, is equally wrong: experts who have teaching skill and who receive real feedback about whether they were understood tend to calibrate well. Subject knowledge and teaching skill are separate capabilities, and the strongest teaching usually needs both.
Does jargon always indicate poor explanation?
No. Jargon is compression, and within a community that shares its meaning it increases precision and saves repeated explanation. It becomes a problem when used across a knowledge gap, before the audience has built the concepts being compressed. The related trap is assuming that removing technical words fixes an explanation: stripping the vocabulary while leaving the conceptual gaps produces prose that reads as accessible and still fails to land. Introduce terminology after the concept it names rather than avoiding it.
How can I test whether someone actually understood?
Not by asking whether it made sense. Self-assessment is weak evidence, because people may not recognise their own gaps or may hesitate to admit confusion, and repeating your words back can happen without understanding. Ask them to explain it in their own words, apply it to a new case, predict what the model implies, or find the error in a sample application. Watch for patterns of confusion across several people rather than reacting to one person. Test with people who match your intended audience, since recruiting complete beginners can mislead when the real readers are intermediate practitioners.
Does knowing about the curse of knowledge fix it?
Awareness alone is generally not sufficient, because the thing you are trying to inspect is exactly what has become invisible to you. What helps is procedural rather than introspective: define the target audience, write down every term and step your draft assumes, watch a representative reader work through the material, ask for teach-back or application, record recurring confusion, then revise and retest. The expert who cannot imagine confusion can still measure it.
Do analogies help or make things worse?
Both, depending on how they are used. Gentner research on structure-mapping distinguishes surface similarity from structural similarity, and the risk is that an expert selects an analogy for a deep relationship the audience cannot yet see, while the audience takes away the surface features instead. Three rules keep analogies useful: state exactly what corresponds, say where the analogy breaks down, and return to the real mechanism once it has provided orientation.