未登录,请登录后再发表信息
最新评论 (0)
播放视频

揭秘隐藏的数据悖论

Unmasking the hidden paradox in data

一口心理学
Bite Size Psych
1970年初
The early 1940s,
伯克利大学因这一数据
Berkeley’s University was sued for sexual discrimination
而被起诉性别歧视
based on this statistic.
录取的男性申请者为44%
44% of male applicants got accepted
而被录取的女性申请者只有35%
whereas only 35% of female applicants got accepted.
此外 如果说10%的性别差异是偶然的
Furthermore, there is less than a 1 in 10,000 chance
这种机率不到万分之一
that this 10% gender gap is due to chance.
问题来了:这能否证明性别歧视?
The question is: can you say this is evidence of sexism?
这一切似乎非常令人信服
It all seems quite convincing.
但是如果我告诉你在上面的案例中
But what if I told you that in the above case,
实际上遭到性别歧视的是男性呢
it was actually the males who are getting discriminated against.
为了方便展示 让我们把数据按学院分开
To show you, let’s break it down into the separate departments.
值得注意的是 在大多数学院里
The thing to note is that in most departments,
实际上录取最多的是女性
it was actually women who are more likely to be admitted.
然而奇怪的是 当你平均起来看
And yet strangely when you take the average,
好像被录取的女性少
the opposite seems true.
为什么会这样呢?
What’s going on?
答案其实就在眼前
Well, the answer is actually hidden in plain sight.
暂停视频 看看你能否发现吧
Pause this video to see if you can figure it out.
时间到 所以问题是
Okay. So the problem is:
我们看到了10%的差异
We saw a 10% gap
然后马上就认为是因为性别歧视
and immediately assumed it was due to gender discrimination.
但是真正的原因是因为女性倾向于申请那些
But the real reason was that women tend to apply for courses
很难被录取的课程
which are harder to get into.
而男性倾向于申请容易被录取的课程
Whereas males apply for courses which are easier.
因此 女性被拒绝的就更多一些
As such women were getting rejected more often.
举个例子来说 把A学院与E学院相比较
For example, take a look at department A compared to E.
A学院比E学院的录取率高了很多
A has a much higher acceptance rate than E.
然而与男性相比
Yet compared to men,
大批女性回避A学院而选择E学院
women are shying away from A and flocking to E.
这种不同的两组数据的趋势
This phenomenon in which trends against different groups
一旦平均起来就消失甚至相反
disappear or even reverse when averaged.
这就叫做辛普森悖论
It’s called the Simpsons paradox.
再举个例子
Here’s another example.
看一下佛罗里达州死刑的实施
Look at the start-up on death penalty’s in Florida.
被判死刑的白人嫌疑犯为4.8%
4.8% of white suspects for the death sentence
而黑人嫌犯则是2.5%
compared to 2.5% of black suspects.
那么大家就知道那些不相信种族歧视存在的人们
Now you can imagine people who believe racism doesn’t exist
在辩论时会用这个数据反驳
using this statistic in a debate.
他们可以争辩如果有任何偏见
They could argue that if there’s any bias,
那也是针对白人的偏见 不是吗?
it’s against the white suspects, right?
但是如果我们根据受害者和嫌犯的种族来分析
But let’s break it down based on the victims and suspects race.
在左上角
So in the top left corner
当白人杀死一个白人
when a white suspect kills a white person,
4.9%被判以死刑
4.9% of them got sentenced to death.
然而 在左下角
However, in the bottom left corner,
当黑人杀掉一个白人
when a black suspect kills a white person,
12.6%被判以死刑
12.6% of them got sentenced to death.
另一个值得注意的事情是
Another important thing to notice is
与杀掉黑人相比
that when a white person was killed,
杀掉白人的刑罚更严酷
the sentences were harsher compared to when a black person was killed.
这都说明了在司法界存在种族歧视
This all suggests that there is racial bias within the justice system.
然而一旦总体平均起来 就呈现了相反的效果
Yet once again the overall averages save the complete opposite.
那么这是为什么呢?
So what’s going on here?
是这样的 白人倾向于杀白人
Well, white people tend to kill other white people,
而黑人倾向于杀黑人
whereas black people tend to kill other black people.
因此 因为平均白人被害受的刑罚更严酷
Therefore sense cases with white victims got harsher sentences on average,
所以白人嫌疑犯被判死刑的几率更大
white suspects got a higher percentage of death sentences.
目前 辛普森悖论不是唯一一种显示数据可以误导你的方式
Now the Simpsons paradox is merely one way statistics can mislead you.
还有无数种方式
There are countless more.
因此 很多人会很失望
And at this point, a lot of people get disheartened
认为我们不应该相信任何数据
and take this to mean we shouldn’t trust statistics at all.
但是所有这些真正的意思是
But all this really means is that the general public
在大众被新闻媒体 政府和信息图的数据
needs to get a better grasp of statistics
轰炸之前
before they are bombarded with stats
他们需要对数据有更好的理解
from news outlets, politicians and infographics.
辛普森悖论仅仅是
The Simpsons paradox is simply a demonstration of
数据可以轻易欺骗我们的证明
just how easily numbers can deceive us.
伙伴们 顺便说一句 我最近创建了一个patreon账号
By the way guys, I’ve recently set up a patreon account,
如果喜欢我的视频并想看更多视频 就去逛逛吧
so if you like my stuff and want to see more then go and check it out.

发表评论

译制信息
视频概述

数据悖论的概念 原因和例子

听录译者

波妞妞

翻译译者

波妞妞

审核员

审核员 HZ

视频来源

https://www.youtube.com/watch?v=JJO4J_tJC2s

相关推荐