Inside OKCupid: The math of online dating - Christian Rudder

探秘OKCupid: 网络交友中的数学 -- Christian Rudder

Translator: Andrea McDonough Reviewer: Bedirhan Cinar

翻译人员: Gena Volz 校对人员: Sharon Loh

Hello, my name is Christian Rudder,

大家好，我叫 Christian Rudder,

and I was one of the founders of OkCupid.

我是 OKCupid 网站的创办人之一。

It's now one of the biggest dating sites in the United States.

这个网站现在已经是 全美最大的交友网站。

Like most everyone at the site, I was a math major,

就象这网站上大多数其他人一样,

我是学数学的， 正如你所期待的那样，

As you may expect, we're known for the analytic approach we take to love.

我们擅于分析。

我们把这方法也应用在爱情上。

我们把它叫做“配对算法”。

We call it our matching algorithm.

基本上 OK Cupid 的配对算法

Basically, OkCupid's matching algorithm helps us decide

帮助我们决定 两个人是否应该约会。

whether two people should go on a date.

We built our entire business around it.

我们的整个业务都是基于这一点。

Now, algorithm is a fancy word,

“算法”这个词说起来专业而高级，

and people like to drop it like it's this big thing.

大家喜欢把它想成很大的一件事，

But really, an algorithm is just a systematic,

但其实,算法只不过是一个系统的,

step-by-step way to solve a problem.

一步一步的解决问题的方法。

It doesn't have to be fancy at all.

根本没有那么复杂。

现在，我将为大家解释

Here in this lesson,

I'm going to explain how we arrived at our particular algorithm,

我们怎样得出这一个特殊的算法。

你会在这看到它是怎样成形的。

so you can see how it's done.

Now, why are algorithms even important?

为什么算法如此重要？

Why does this lesson even exist?

为什么我们要有这堂课？

Well, notice one very significant phrase I used above:

请注意我刚才提到的一个很重要的词:

they are a step-by-step way to solve a problem,

它们是一种"逐步"解决问题的方法，

and as you probably know, computers excel at step-by-step processes.

你或许也知道，

电脑擅长于一步一步的运算过程。

A computer without an algorithm

没有算法的电脑，

基本上只是一个昂贵的镇纸。

is basically an expensive paperweight.

And since computers are such a pervasive part of everyday life,

既然电脑已经普及到我们的日常生活，

algorithms are everywhere.

算法是无处不在。

The math behind OkCupid's matching algorithm is surprisingly simple.

OK Cupid 配对算法背后的数学逻辑

是非常简单的。

It's just some addition, multiplication, a little bit of square roots.

就是一些加法，

乘法，

再来一点平方根。

The tricky part in designing it

不过，设计这套算法的关键部分，

was figuring out how to take something mysterious,

在于要找出那些神秘的

human attraction,

人与人之间的相互吸引力，

and break it into components that a computer can work with.

并把它解构成电脑可以工作的部分，

我们要做的第一件事 就把人和数据关联起来,

The first thing we needed to match people up was data,

something for the algorithm to work with.

这样算法才能生效。

The best way to get data quickly from people is to just ask for it.

要最快的从人们那里得到数据，

最好就是直接询问他们。

So we decided that OkCupid should ask users questions,

我们决定 OK Cupid 应该向用户问问题，

stuff like, "Do you want to have kids one day?"

比如说：“你会想要小孩吗？”，

“你多久刷一次牙？“，

"How often do you brush your teeth?"

"Do you like scary movies?"

”你喜欢看恐怖电影么？”。

And big stuff like, "Do you believe in God?"

也有严肃些的问题， 比如：“你相信上帝么？”。

Now, a lot of the questions are good for matching like with like,

目前有很多问题

在进行同类型配对上都很合适，

that is, when both people answer the same way.

就是当双方的答案相同时。

For example, two people who are both into scary movies

比如，两个人都喜欢看恐怖电影

are probably a better match than one person who is and one who isn't.

可能配对得更成功。

而一个人喜欢,

另外一个人不喜欢的情况下, 适配度就差点。

But what about a question like,

但如果碰到下面的问题 ：

"Do you like to be the center of attention?"

“你喜欢成为关注的中心么?”

If both people in a relationship are saying yes to this,

如果交往中的双方都回答是，

那他们可有大问题了。

they're going to have massive problems.

We realized this early on,

我们很早就意识到了这一点，

and so we decided we needed a bit more data from each question.

所以我们觉得需要

在每个问题再收集多一些数据。

We had to ask people to specify not only their own answer,

我们不仅要人们回答自己的看法，

but the answer they wanted from someone else.

也要他们回答 他们期待对方如何回答。

That worked really well.

这方法很有效，

But we needed one more dimension.

不过我们还要再多加一个维度。

Some questions tell you more about a person than others.

有些问题能表达人们的与众不同之处。

For example, a question about politics, something like,

比如，关于政治的问题，

“ 焚烧书籍或者国旗， 哪个更糟糕 ？”

"Which is worse: book burning or flag burning?"

might reveal more about someone than their taste in movies.

这能展露人们电影口味之外的东西

And it doesn't make sense to weigh all things equally,

同时，并不是所有问题都同等重要的，

所以我们最后增加了一个数据点。

so we added one final data point.

For everything that OkCupid asks you,

任何 OK Cupid 的问题，

you have a chance to tell us the role it plays in your life.

你都可以告诉我们

这问题对你的重要性，

And this ranges from irrelevant to mandatory.

它的程度从“无关”到“必要”。

So now, for every question, we have three things for our algorithm:

现在，每一个问题，

我们有三个资讯提供给算法:

first, your answer;

第一,你的答案;

second, how you want someone else -- your potential match -- to answer;

第二,你希望别人怎么回答;

也就是你潜在的对象,

的答案;

and third, how important the question is to you at all.

第三，这个问题对你有多重要？

With all this information,

有了这些信息，

OkCupid can figure out how well two people will get along.

OK Cupid 可以知道 两个人相处和谐程度如何。

The algorithm crunches the numbers and gives us a result.

算法吃进数字,吐出答案。

实际举例来说吧，

As a practical example,

let's look at how we'd match you with another person.

看我们怎样把你和另外一个人进行配对，

Let's call him "B."

暂且称他为 “B”。

你和 B 的适配度是基于

Your match percentage with B is based on questions you've both answered.

你们双方都进行过回答的问题。

Let's call that set of common questions "s."

姑且把这些共同问题称之为 “s”。

简单举例，我们用小样本的 “s”，

As a very simple example, we use a small set "s"

with just two questions in common,

只需两个共同回答过的问题

and compute a match from that.

电脑会根据它算出适配度。

Here are our two example questions.

这里是我们的两道简单问题：

The first one, let's say, is, "How messy are you?"

第一个是,“你有多杂乱无章?”

And the answer possibilities are:

可供选择的答案选项有

very messy, average and very organized.

非常杂乱无章,

一般，

和非常有条理。

And let's say you answered "very organized,"

我们假设你回答的是“非常有条理”,

你期待别人的回答是“非常有条理”,

and you'd like someone else to answer "very organized,"

and the question is very important to you.

并且对你来说，这个问题非常重要。

Basically, you're a neat freak.

基本上你就是个井井有条的怪胎。

You're neat, you want someone else to be neat, and that's it.

你是整洁有条理的人，

你也希望对方同样如此,

就这样。

And let's say B is a little bit different.

我们假设 B 有些不同。

He answered "very organized" for himself,

他的回答是自己非常有条理,

but "average" is OK with him as an answer from someone else,

但是他也接受“一般”，

如果别人是这样回答的话，

and the question is only a little important to him.

这个问题于他而言不太重要。

我们看第二个问题，

Let's look at the second question, from our previous example:

就是我们最开始举例的：

“你喜欢成为关注的中心么?”

"Do you like to be the center of attention?"

答题项只有“是”或者“否”。

The answers are "yes" and "no."

You've answered "no," you want someone else to answer "no,"

现在你的回答是“否”，

你希望别人怎样回答这栏答的是“否”

and the question is only a little important to you.

这个问题对于你不太重要。

而B呢,他自己的回答是“是”,

Now B, he's answered "yes."

He wants someone else to answer "no,"

他希望别人回答“否”，

because he wants the spotlight on him,

因为他希望所有焦点都在他身上，

and the question is somewhat important to him.

而这个问题对他还算重要。

So, let's try to compute all of this.

现在，我们让电脑来处理一切。

Our first step is, since we use computers to do this,

我们的第一步是，

既然我们要用电脑来处理它，

we need to assign numerical values

我们就需要给一些数值

to ideas like "somewhat important" and "very important,"

来定义比如“还算重要”和“非常重要”，

because computers need everything in numbers.

因为电脑需要把所有资料都转化成数字。

We at OkCupid decided on the following scale:

在 OK Cupid 上我们按如下级别：

"Irrelevant" is worth 0.

“无关”是 0，

“不太重要”的值是1,

"A little important" is worth 1.

"Somewhat important" is worth 10.

“还算重要”的值是 10，

"Very important" is 50.

“非常重要”的值是 50，

And "absolutely mandatory" is 250.

“绝对必要”的值是 250.

Next, the algorithm makes two simple calculations.

接下来,算法要做两个简单的计算。

The first is: How much did B's answers satisfy you?

第一个是你对B的回答给多少分,

另外一个是,你给对方答题的满分是多少?

That is, how many possible points did B score on your scale?

Well, you indicated that B's answer to the first question,

你可以指定 B 的答案

在第一个有关条理性的问题上，

about messiness,

对你是非常重要。

was very important to you.

It's worth 50 points and B got that right.

它值50分，B 答对了。

The second question is worth only 1,

第二个问题只有1分，

因为你说这问题对你不太重要,

because you said it was only a little important.

B 答错了。

B got that wrong,

so B's answers were 50 out of 51 possible points.

所以B的回答在51分满分里拿到了50分。

That's 98% satisfactory. Pretty good.

适配满意度是 98%。

非常好。

The second question the algorithm looks at is: How much did you satisfy B?

算法的第二个问题是看

B 对你的满意程度。

B给对于你有关条理性的回答

Well, B placed 1 point on your answer to the messiness question

给1分，

and 10 on your answer to the second.

对于第二个问题的答案给10分。

Of those 11, that's 1 plus 10, you earned 10 --

满分11分，就是1+10.

你得到了10分，

you guys satisfied each other on the second question.

在第二个问题上，你俩彼此都满意。

So your answers were 10 out of 11 equals 91 percent satisfactory to B.

你的回答在B的满意度分数是10/11,

百分比是91%。

That's not bad.

还不错。

The final step is to take these two match percentages

最后一步是把两个适配度百分比放在一起，

and get one number for the both of you.

为你们两打一个分数。

To do this, the algorithm multiplies your scores,

为得到这点， 算法把你们两人的得分相乘，

然后开n次方根,

then takes the nth root,

where "n" is the number of questions.

n 就是问题的数目。

因为“s”-- 也就是问题的数目，

Because s, which is the number of questions in this sample,

在这个例子里,只是“2”,

is only 2,

we have: match percentage equals the square root

我们得到的适配度百分比等于

98% 乘以 91% 再开平方根。

of 98 percent times 91 percent.

That equals 94 percent.

结果等于94%。

That 94 percent is your match percentage with B.

94%就是你和 B 之间的适配度百分比。

It's a mathematical expression of how happy you'd be with each other,

这是通过数学方法来表达--

你们彼此之间相处的愉快程度是怎样。

based on what we know.

基于我们所知道的信息。

为什么算法要相乘，而不是除？

Now, why does the algorithm multiply,

as opposed to, say, average the two match scores together,

比如，把两个分数求平均值以后

and do the square-root business?

再开平方根？

In general, this formula is called the geometric mean.

总的来说，这个公式叫几何平均数，

It's a great way to combine values that have wide ranges

它很适合处理

差异很大的数据，

and represent very different properties.

以及代表不同属性的数据。

换句话说，它能完美的 计算出浪漫爱情适配度。

In other words, it's perfect for romantic matching.

You've got wide ranges and you've got tons of different data points,

你有大范围的,

数不清的数据值,

就像刚说过的，有关电影的，

like I said, about movies, politics, religion -- everything.

有关政治的，

有关宗教的，

有关所有的一切。

Intuitively, too, this makes sense.

凭直觉讲，以下情况很有道理。

Two people satisfying each other 50 percent

两个人彼此的满意度是50%，

会好过

should be a better match than two others who satisfy 0 and 100,

那些两个人彼此满意度是0或者100的。

because affection needs to be mutual.

因为爱慕应该是互相的。

在增加了对误差幅度的小修改后 --

After adding a little correction for margin of error,

in the case where we have a small number of questions,

这种情况在问题量很小的时候会出现,

就像我们刚举的运算实例一样－－

like we do in this example,

we're good to go.

这套算法就可以运作了。

Any time OkCupid matches two people,

任何时候当 OK Cupid 将两个人配对时，

it goes through the steps we just outlined.

它按照我们刚介绍的步骤来運作，

First it collects data about your answers,

首先它收集你的答题的数据，

然后它比较你的选项和 你期待的对方选项，

then it compares your choices and preferences to other people's

以简单的，数学的方法来进行。

in simple, mathematical ways.

这种能将现实世界的现象，

This, the ability to take real-world phenomena

and make them something a microchip can understand,

转化为电脑芯片能读取的数据的能力，

is, I think, the most important skill anyone can have these days.

我认为，

是现代最重要的一种技术。

Like you use sentences to tell a story to a person,

就像你用话语来给一个人讲故事，

你是用算法来跟电脑讲故事。

you use algorithms to tell a story to a computer.

如果你学会了这种语言，

If you learn the language, you can go out and tell your stories.

你就可以去讲故事了。

我希望我刚才的介绍能帮助你做到这点。

I hope this will help you do that.