type
Post
Created date
Jun 15, 2022 06:01 AM
category
Filosofia
tags
Logics
Mental Model
status
Published
Language
Chinese
From
Resonance Calendar
summary
slug
password
Author
Priority
Featured
Featured
Cover
Type
URL
Youtube
Youtube
icon
 

Base Rate Fallacy:

Imagine you have a big box of red and blue marbles. Even if most marbles in the box are blue, just because you picked a red one once doesn't mean most of the marbles are red. The "base rate" is like knowing how many of each color marble is in the box before you start guessing.

Three Questions to Ask:
  1. Did I remember how many of each color marble is in the box? i.e., Am I looking at the big picture?
      • Consider all available information, not just the most obvious or immediate details.
      • This is like making sure you remember the big picture, not just one thing you saw.
  1. Am I only thinking about the one red marble I picked and forgetting about all the other marbles? i.e., Am I focusing too much on one piece of information?
      • Even if one detail seems really important, it's essential to weigh it against all other information.
      • This reminds us to think about everything, not just one thing that might seem special.
  1. If I forget about the number of each color in the box, will my guess be way off? i.e., How wrong could I be if I ignore the overall rate?
      • This helps us understand the potential mistake we might make if we don't consider the broader context or general rate of occurrence.
      • This helps us realize that if we don't think about the whole box, we might make a big mistake in our guess.
 

看看以下兩個情境:
1. 有路人在咳嗽,從她和別人的談話得知,她染上了 X 和 Y 其中一種疾病。已知 X 的患者有 75% 會咳嗽,而 Y 的患者只有 45% 會咳嗽,你會推論……? 2. 你擔心自己被傳染,走了去做 X 病的檢測。你用的試劑可以百分百有效地偵測出 X 病患者,但同時也會誤將 10% 沒有患病的人診成患者。換句說話,如果你有病,檢測結果必定呈陽性 (true positive) ;如果你沒有病,檢測結果有 10% 會呈陽性 (false positive) 。你收到檢測結果,天啊竟然是陽性,由此可見……?
多數人的推論是這樣的。
  • 在第一個情境,既然患上 X 而咳嗽的比例高過患上 Y 而咳嗽的比例,那個路人似是患上 X 。有些人甚至會推論,那個路人有 75% 機率患上 X 。
  • 在第二個情境,既然檢測的犯錯率只有 10% ,我應該有 90% 機會患上 X 。
這兩個常見的推論都是錯的,因為它們都忽略了基本比率 (base rate) 。
...

第一個情境

假設 X 病的患者有 1,000 人, Y 病的患者有 9,000 人。
notion image

情境一

用已知的百分比可以計算出,患 X 病而咳嗽的人有 750 個,患 Y 病而有咳嗽的人有 4,050 個,這代表總共有 4,800 個患者有咳嗽的症狀。
  • 咳嗽人士之中, X 病患者佔的比例是: 750 / 4,800 = 15.6%
  • 咳嗽人士之中, Y 病患者佔的比例是: 4,050 / 4,800 = 84.4%
由此可見,如果 X 病患者和 Y 病患者的比例是 1:9 ,咳嗽的人是 X 病患者的可能性便遠低於是 Y 病患者的可能性。這個 1:9 是基本比率。不知基本比率,即使知道兩病患者的咳嗽比例,依然無法單從咳嗽推論出對方更可能患哪種病。
notion image
...

第二個情境

假設 X 病的患者有 1,000 人,沒有患 X 病的人 ── 這當中包括患了其他病和健康的人 ── 有 99,000 人。
notion image

情境二

從已知的百分比可以計算出,患 X 病而檢測出陽性的人數是 1,000 ,沒有患 X 病但同樣檢測出陽性的人數是 9,900 ,所以檢測呈陽性的總人數是 10,900 。
  • 測得陽性的人之中,患有 X 病的比例是: 1,000 / 10,900 = 9.2%
  • 測得陽性的人之中,沒有患 X 病的比例是: 9,900 / 10,900 = 90.8%
假如患 X 病和沒有患 X 病的比例是 1:99 ,檢測呈陽性的人之中,確實有患 X 病的比例,其實連一成也不到。這個 1:99 是基本比率。換句話說,假如不知道基本比率,即使知道檢測可以百分百偵測出 X 病患者,而只有 10% 的非患者會被誤診,依然無法單從陽性結果推論對方患病的概率。
notion image
...

備註

1. 情境一改自 Tracy Bowell 與 Gary Kemp 合著的 Critical Thinking: A Concise Guide (2010, 3rd ed., pp. 206–7) 。 2. 情境二改自 David Papineau 的 Philosophical Devices (2012, p. 113) 。 3. 講「基本比率謬誤 (Base rate fallacy)」的文獻主要以情境二為例,但其實情境一的例子原理相通。
Mental Models: The Best Way to Make Intelligent Decisions (109 Models Explained)PIXAR - Storytelling framework

Jason Siu
A warm welcome! I am a tech enthusiast who loves sharing my passion for learning and self-discovery through my website.
Statistics
Number of posts:
229