The Curse of Knowledge: How Experts Struggle to Explain

What has been corrected on this page?

Every accepted correction to this page is recorded with the exact change, so readers can see how the page improved over time.

1. 5 August 2026 · corrected by Emir Baycan

   A claim-level review found that this article had expanded a well-documented bias into a universal explanation for teaching, documentation, design, medicine and organizational failure, supported by several unsourced or misattributed claims. It opened by asserting that the most qualified person is often the worst teacher and defined the curse as a systematic inability to reconstruct ignorance, where the literature supports a bias rather than an inability. It attributed to Nickerson a specific teaching-feedback cycle he did not propose, cited neuroimaging vaguely as work by Milton and colleagues with no paper in the bibliography, described a physics-teaching experiment and faculty calibration protocols attributed to Carl Wieman with only a general essay cited, and credited Nielsen Norman Group with proving documentation failure across hundreds of studies without naming a report. Case studies asserted that Toyota engineers could not explain their own system and that observation-first training worked substantially better, that Johns Hopkins used real-time reasoning interruption, and that Stripe had an explain-it-to-a-designer culture, none of which were sourced. It also misnamed Pamela Hinds as Patricia, misattributed How Doctors Think, carried unverified Einstein and Mark Twain quotations, and concluded with an invented claim that confidence correlates negatively with actual clarity. Structurally the page ran two near-identical domain sections, placed a research appendix and case studies after its own conclusion and bibliography, and ended with a literal Article Word Count line that was itself inaccurate.

   What the page claimed

   The article opened: 'The most qualified person to explain a concept is often the worst teacher of it... a cognitive bias so fundamental that it affects every domain where knowledge transfers between individuals', and defined the curse as 'the systematic inability of informed parties to reconstruct the mental state of ignorance they once occupied'. It said 'The curse operates invisibly-those afflicted cannot perceive their own affliction' and called it a 'universal constraint' that 'cannot be fully eliminated'. It stated Nickerson 'argued this creates a self-reinforcing cycle: poor explanations - audience confusion - attribution to audience limitations - no modification'. It cited 'work by Milton and colleagues in the 2000s' with no paper in the bibliography. It described physics professors systematically mispredicting question difficulty and said Wieman 'used this finding to develop faculty calibration protocols', citing only a 2007 general essay. It said Nielsen Norman research 'synthesized across hundreds of usability studies' proves documentation failure. It claimed Toyota's most experienced engineers 'were among the least able to explain it', that Toyota's observation-first approach 'worked substantially better than explanation-first methods', that medical training 'including approaches used at institutions such as Johns Hopkins' used real-time reasoning interruption, and that Stripe's culture emphasises 'explain it to a designer'. It attributed How Doctors Think to Kathryn Montgomery, named the researcher 'Patricia Hinds', said the Camerer study 'established that the curse... operates in... negotiation, finance, legal argumentation, and management decision-making', and asserted a grandmaster sees 'Sicilian Defense, Najdorf Variation, move 12' where a novice sees 'thirty-two pieces in thirty-two locations'. It carried unverified Einstein and Mark Twain quotations, ran Suzuki and Polanyi quotes together with body prose inside the same block, and concluded 'The less confident you feel about an explanation's clarity, the more likely it is clear-confidence in one's explanation correlates negatively with its actual transparency to novices' and that some are 'most severely cursed'. It ended with a literal line reading 'Article Word Count: 3,189' on a page of about 4,300 words, after a research appendix and case studies placed below its own conclusion and bibliography.

   What was corrected

   Three near-duplicate WNF articles on this concept were consolidated into this one, with 301 redirects from the two retired URLs and the sitemap regenerated. The opening now states that expertise can make explanation harder while noting explicitly that this does not make experts inherently poor teachers. The definition is a difficulty or bias rather than an inability. A table now distinguishes the curse of knowledge from the curse of expertise, illusion of transparency, false consensus, tacit knowledge and cognitive load. The Newton tapping study is presented as a vivid illustration rather than proof about experts generally; Nickerson is cited for the general tendency without the invented feedback cycle; Camerer is limited to its economic setting with 'suggests' rather than 'established'; and Hinds is correctly named Pamela, bounded to prediction of novice difficulty, with the non-linear calibration result presented without a confident mechanism. The neuroimaging paragraph, the Wieman experiment, the Nielsen Norman claim, and the Toyota, Johns Hopkins, Stripe and Apple attributions were removed, along with the Montgomery misattribution, the Einstein and Twain quotations, the confidence-clarity correlation and the 'most severely cursed' line. Chase and Simon is now described as recognising familiar patterns rather than a specific named variation. Flanagan is credited for the critical incident technique, cognitive task analysis is described as several methods across traditions, and Kalyuga et al. is added for expertise reversal. A new nuance section covers expert-to-expert communication, the modulating role of feedback, when beginners do and do not teach better, jargon as precision, audience mismatch, and the insufficiency of awareness alone. The Feynman technique and the limits material were merged in from the two retired articles. The post-conclusion appendices and the stray word-count line are gone, and the page runs about 2,500 words in place of roughly 13,500 across three URLs.

   Why: WNF had three separately published articles covering this same concept, all live and all in the sitemap, which is a direct search-intent cannibalisation problem on a site recovering from deindexing. Beyond the duplication, the surviving article supported a legitimate and well-documented bias with a layer of unsourced case studies, a misattributed research cycle, an experiment credited to a researcher without the corresponding paper, a misnamed author, and an invented correlation between low confidence and actual clarity. Consolidation preserves the strongest material from all three, and the corrections bound each remaining claim to what its source establishes.

   View the full record →
2. 28 July 2026 · corrected by Emir Baycan

   Qualified six overstated or unverifiable claims: reframed a single Milton et al. (2007) fMRI study as suggestive rather than a settled general neural explanation of teaching difficulty; softened Hinds's 1999 expert-prediction study from a universal expertise gradient to one controlled study consistent with a broader pattern; replaced an unverifiable 2014 Wieman lecture-evaluation study with the well-documented general finding that expert instructors misjudge which concepts students find hard; qualified the Nielsen Norman Group documentation-abandonment figure (users give up within a few minutes, varying by study, not a fixed two-to-three-minute constant); softened the Toyota 'experts were worst at explaining' claim to a tacit-knowledge framing attributed to Toyota Culture (2008); and reframed the Johns Hopkins/Amazon debiasing measures as promising structured practices rather than established structural cures. The illusion-of-transparency attribution to Gilovich, Savitsky, and Medvec (1998) was already correct.

   What the page claimed

   The article presented Milton 2007 as a general neural explanation, Hinds as establishing a universal expertise gradient, a 2014 Wieman study with a specific protocol, a fixed two-to-three-minute documentation-abandonment figure, Toyota experts as the worst at explaining, and Johns Hopkins/Amazon measures as structural cures.

   What was corrected

   Qualified each to what the sources support, generalized the unverifiable Wieman study, and reframed the debiasing measures as promising practices, while keeping the correct illusion-of-transparency citation.

   Why: The sources are largely genuine but were overstated or, for the 2014 Wieman study, unverifiable as cited. Corrections scope the claims rather than adding new sources.

   View the full record →

Who checked this page?

1 contributor has checked "The Curse of Knowledge: How Experts Struggle to Explain" on When Notes Fly. Each name below links to that person's public CitePep profile, where every contribution they have made is listed with the exact change they proposed.