Skip to content

[Anchor Proposal]: Curse of Knowledge #638

Description

@rehsack

Proposed Term

Curse of Knowledge

Context

A cognitive bias: once you know something, you can no longer easily imagine not knowing it, so experts unconsciously assume background the audience lacks — and communicate over their heads. Coined by Colin Camerer, George Loewenstein & Martin Weber (1989, Journal of Political Economy); popularised by Chip & Dan Heath (Made to Stick, 2007) as the central villain of communication, and by Steven Pinker (The Sense of Style, 2014) as the chief cause of bad technical/academic writing. (Cf. Elizabeth Newton's 1990 "tappers and listeners" experiment.)

Captures the communication limit Jens framed as "I can explain it to you, but I can't understand it for you" — the receiver does the understanding; the sender can only calibrate. As an LLM anchor it is especially useful for documentation aimed at an uncertain audience: instructing a model to "avoid the curse of knowledge" makes it surface assumed knowledge, define jargon, and pitch to the reader's level.

Fits communication-presentation; relatives in the catalog: feynman-technique, plain-english-strunk-white, bluf, inverted-pyramid-style, blooms-taxonomy.

Sources: Camerer, Loewenstein & Weber (1989) · Chip & Dan Heath, Made to Stick (2007) · Steven Pinker, The Sense of Style (2014).

LLM Activation Test Result

Model: Claude (Opus 4.8)
Prompt: "What concepts do you associate with the 'Curse of Knowledge'?"
Response: A cognitive bias where experts can't un-know what they know and so overestimate shared background, producing jargon-heavy, over-assuming explanations; coined by Camerer, Loewenstein & Weber (1989); popularised by the Heath brothers (Made to Stick) and Pinker (The Sense of Style); illustrated by Newton's tappers-and-listeners study; remedies include audience empathy, concreteness, defining terms, and feedback. Rich, consistent, strongly attributable activation.

Pre-submission Checklist

  • I have searched existing anchors and this term is not already included
  • This term refers to a well-established concept (not something I invented)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions