Dunning-Kruger effect條curve

成日見啲人攞下面呢條curve嚟代表DK effect


但實際上嘅份paper所形容嘅DK effect同條curve係兩回事

Perceived logical reasoning ability and test performance as a function of actual test performance, taken from https://www.researchgate.net/publication/12688660_Unskilled_and_Unaware_of_It_How_Difficulties_in_Recognizing_One%27s_Own_Incompetence_Lead_to_Inflated_Self-Assessments

其實條curve係喺邊度嚟，又點解會傳得咁廣？

呢個都係一個問題，psychology嘅實驗成日都被質疑好難replicate
不過我冇咩詳細睇過有關呢方面嘅嘢



我個主要疑問係，成日喺連登同埋外國forums上面見到第一幅圖條curve，suggest冇能力者嘅自信 > 有能力者
但睇返篇文（唔理篇文嘅conclusion係咪solid先），會發現其實有能力者嘅「自信」其實 > 冇能力者
只不過佢哋會underestimate自己，而冇能力者會overestimate自己

所以好奇條curve喺邊度彈出嚟

至於replicability問題，今日晏啲可以搵搵有冇人試過重做呢個實驗
不過我psychology嘅知識貧瘠，可能需要連登巴打絲打提點一下

true
我之前睇咗下個Many Labs Replication Project，發現psy嘅replication crisis真係好嚴重
不過dk effect好似都冇咩特別嘅real world application，啲人都係當meme噉spread，又少咗incentive去replicate


btw
真係見有人re-evaluate過DK effect，話係artefact
https://www.sciencedirect.com/science/article/abs/pii/S0160289620300271

又有人再evaluate個re-evulation，話有effect，但個effect好細
https://www.sciencedirect.com/science/article/abs/pii/S0160289622000988?via%3Dihub
唔想用手機睇文，遲啲先睇

re-evaluation 今日不停打錯字

係，初學者overestimate自己呢個現象喺網上成日見，可能啲人多機會用，好易廣傳
有個effect攞出嚟講又好似頭頭是道噉

但最好奇係幅graph係邊條友整出嚟，重上埋wiki

confidence= perceived score - actual score?
會得到mt. stupid 到 plateau
0 knowledge 到 mt. stupid 就求奇拉直線上

幅圖出現係因為一幅圖一條線易解釋啲掛？

去唔到plateau
top嗰班人係underestimate自己，perceived score - actual score係negative

咁就應該係 y(confidence) = x + (perceived score - actual score)

唔係好明
confidence = competence + perceived ability - actual ability?

定你意思x係個constant term，逼佢上positive

x = y
confidence (y, 0-100) = competence (x, 0-100) + (perceived score - actual score)

佢應該撈埋咗好幾樣想講嘅野

competence唔係practically equivalent to actual score咩
唔係要點理解

咁幅圖就係plot perceived score against actual score

我個諗法係
幅圖base on confidence = competence
with an arbitrary value 0-100/1000 etc.

然後用(perceived score - actual score)表達overconfident/underconfident

sorry for the poor presentation

睇下我有冇理解錯你講咩

如果真係，個plot都幾複雜
正常人會諗到用x = y做base咩

另外如果係噉，個slope of enlightenment應該要打橫 / less steep than plateau of sustainability
plateau of sustainability應該要follow x=y，變slope先make sense

大概係咁
呢啲圖唔洗太講究啦我諗
表達到
1. overconfident
2. back to normal (despair)
3. correctly know yourself (slope of enlighten)
4. underconfident (plateau)
用返D-K 原本幅圖難睇啲

再者咁樣plot 個arbitrary value 細啲幅圖出嚟會juicy 啲
mt. stupid 可以變 mt. stupid Everest

講得通
thx ching

睇過啲質疑嘅文，好多都focus on statistical analyses嘅artefacts
咩regression toward the mean我真心識條撚

冇能力，睇下有冇其他人望下講下

當然條curve本身唔係statistically derived
但有人質疑個effect本身（which the curve attempts to describe）係statistical artefact

btw here's a stackexchange page discussing the same topic
https://psychology.stackexchange.com/questions/17825/what-is-the-primary-source-of-the-mount-stupid-graphic