Jennifer Golbeck: The curly fry conundrum: Why social media "likes" say more than you might think

翻译人员: Li Li 校对人员: 杏儀 歐陽

If you remember that first decade of the web,

如果你还记得网络时代的头十年，

it was really a static place.

网络是一个水尽鹅飞的地方。

You could go online, you could look at pages,

你可以上网，你可以浏览网页，

and they were put up either by organizations

当时的网站

who had teams to do it

要么是由某个组织的专门团队建立，

or by individuals who were really tech-savvy

要么就是由真正的技术行家所做，

for the time.

这就是当时情况。

And with the rise of social media

但在二十一世纪初

and social networks in the early 2000s,

随着社交媒体以及社交网络的兴起，

the web was completely changed

网络发生了翻天覆地的变化：

to a place where now the vast majority of content

如今网络上大部分的互动内容

we interact with is put up by average users,

都是由大众网络用户提供，

either in YouTube videos or blog posts

既有Youtube视频，也有博客文章，

or product reviews or social media postings.

既有产品评论，也有社交媒体发布。

And it's also become a much more interactive place,

与此同时，互联网成为了一个有更多互动的地方，

where people are interacting with others,

人们在这里互相交流、

they're commenting, they're sharing,

互相评论、互相分享，

they're not just reading.

而不只是阅读信息。

So Facebook is not the only place you can do this,

面谱网不是唯一一个你可以做这些事情的地方，

but it's the biggest,

但它确实是最大的一个，

and it serves to illustrate the numbers.

并且它用数字来证明这点。

Facebook has 1.2 billion users per month.

面谱网每个月有12亿用户。

So half the Earth's Internet population

由此可见，地球上一半的互联网用户

is using Facebook.

都在使用面谱网。

They are a site, along with others,

这些都是网站，

that has allowed people to create an online persona

允许人们在网上创建不同的角色，

with very little technical skill,

但这些人又不需要有多少计算机技能，

and people responded by putting huge amounts

而人们的反应是

of personal data online.

在网上输入大量的个人信息。

So the result is that we have behavioral,

结果是，我们拥有数以亿计人的

preference, demographic data

行为信息、喜好信息

for hundreds of millions of people,

以及人口数据资料。

which is unprecedented in history.

这在历史上前所未有。

And as a computer scientist, what this means is that

对于作为计算机科学家的我来说，这意味着

I've been able to build models

我能够建立模型

that can predict all sorts of hidden attributes

来预测各种各样的

for all of you that you don't even know

你或许完全没有意识到的

you're sharing information about.

与你所分享的信息相关的隐藏信息。

As scientists, we use that to help

作为科学家，我们利用这些信息

the way people interact online,

来帮助人们在网上交流。

but there's less altruistic applications,

但也有人用此来谋取自己的私欲，

and there's a problem in that users don't really

而问题是，用户并没有真正理解

understand these techniques and how they work,

其中用到的技术和技术的应用方式。

and even if they did, they don't have a lot of control over it.

即便理解了，也不见得他们有话事权。

So what I want to talk to you about today

所以，我今天想谈谈

is some of these things that we're able to do,

我们能够做的一些事情，

and then give us some ideas of how we might go forward

也启发我们

to move some control back into the hands of users.

如何改善情况、让话事权回归用户。

So this is Target, the company.

这是塔吉特百货公司的商标。

I didn't just put that logo

我并不单单把那个商标

on this poor, pregnant woman's belly.

放在这个可怜的孕妇的肚子上。

You may have seen this anecdote that was printed

或许在福布斯杂志上

in Forbes magazine where Target

你看过这么一则趣事：

sent a flyer to this 15-year-old girl

塔吉特百货公司给这个15岁女孩寄了一份传单，

with advertisements and coupons

传单上都是婴儿奶瓶、尿布、

for baby bottles and diapers and cribs

婴儿床的广告和优惠券。

two weeks before she told her parents

这一切发生在

that she was pregnant.

她把怀孕消息告诉父母的两周前。

Yeah, the dad was really upset.

没错，女孩的父亲很生气。

He said, "How did Target figure out

他说：”塔吉特是如何

that this high school girl was pregnant

在连这个高中女生的父母都尚未知情之前

before she told her parents?"

就知道她怀孕了？“

It turns out that they have the purchase history

原来，塔吉特有成千上万的顾客，

for hundreds of thousands of customers

并拥有他们的购买历史记录，

and they compute what they call a pregnancy score,

他们用计算机推算出他们所谓的“怀孕分数”，

which is not just whether or not a woman's pregnant,

不仅能知道一个女性是否怀孕，

but what her due date is.

而且还能计算出她的分娩日期。

And they compute that

他们计算出的结果

not by looking at the obvious things,

不单单是基于一些显而易见的事情，

like, she's buying a crib or baby clothes,

比如说，她准备买个婴儿床或孩子的衣服，

but things like, she bought more vitamins

更是基于其他一些事情，

than she normally had,

例如她比平时多买了维他命，

or she bought a handbag

或她买了一个新的手提包

that's big enough to hold diapers.

大得可以放尿布。

And by themselves, those purchases don't seem

单独来看这些消费记录

like they might reveal a lot,

或许并不能说明什么，

but it's a pattern of behavior that,

但这确是一种行为模式，

when you take it in the context of thousands of other people,

当你有大量人口背景作比较，

starts to actually reveal some insights.

这种行为模式就开始透露一些见解。

So that's the kind of thing that we do

当我们根据社交媒体来预测关于你的一些事情时，

when we're predicting stuff about you on social media.

这便是我们常做的一类事情。

We're looking for little patterns of behavior that,

我们着眼于零星的行为模式，

when you detect them among millions of people,

当你在众人中发现这些行为模式时，

lets us find out all kinds of things.

会帮助我们发现各种各样的事情。

So in my lab and with colleagues,

在我的实验室，在同事们的合作下，

we've developed mechanisms where we can

我们已经开发了一些机制

quite accurately predict things

来较为准确地推测一些事情，

like your political preference,

比如你的政治立场、

your personality score, gender, sexual orientation,

你的性格得分、性别、性取向、

religion, age, intelligence,

宗教信仰、年龄、智商，

along with things like

另外还有：

how much you trust the people you know

你对认识的人的信任程度、

and how strong those relationships are.

你的人际关系程度。

We can do all of this really well.

我们能够很好地完成这些推测。

And again, it doesn't come from what you might

我在这里在强调一遍，这种推测并基于

think of as obvious information.

在你看来显而易见的信息。

So my favorite example is from this study

我最喜欢的例子是来自

that was published this year

今年发表在

in the Proceedings of the National Academies.

美国国家论文集上的一个研究。

If you Google this, you'll find it.

你可以在谷歌搜索找到这篇文章。

It's four pages, easy to read.

这篇文章总共四页，容易阅读。

And they looked at just people's Facebook likes,

他们仅仅研究了人们在面谱网上的“赞”，

so just the things you like on Facebook,

也就是你在面谱网上喜欢的事情。

and used that to predict all these attributes,

他们利用这些数据来预测

along with some other ones.

之前所说的所有特性，还有其他的一些特性。

And in their paper they listed the five likes

在文章中列举了

that were most indicative of high intelligence.

最能够显示高智商的五个“赞”。

And among those was liking a page

在这五项中

for curly fries. (Laughter)

赞“炸扭薯”页面的是其中之一

Curly fries are delicious,

炸扭薯很好吃，

but liking them does not necessarily mean

但喜欢吃炸扭薯

that you're smarter than the average person.

并不一定意味着你比一般人聪明。

So how is it that one of the strongest indicators

那么为什么喜欢某个页面

of your intelligence

就成为显示你智商

is liking this page

的重要因素，

when the content is totally irrelevant

尽管该页面的内容和所预测的属性

to the attribute that's being predicted?

与此毫不相干？

And it turns out that we have to look at

事实是我们必须审视

a whole bunch of underlying theories

大量的基础理论，

to see why we're able to do this.

从而了解我们是如何做到准确推测的。

One of them is a sociological theory called homophily,

其中一个基础理论是社会学的同质性理论，

which basically says people are friends with people like them.

主要意思是人们和自己相似的人交朋友。

So if you're smart, you tend to be friends with smart people,

所以说，如果你很聪明，你倾向于和聪明的人交朋友。

and if you're young, you tend to be friends with young people,

如果你还年轻，你倾向于和年轻人交朋友。

and this is well established

这是数百年来

for hundreds of years.

公认的理论。

We also know a lot

我们很清楚

about how information spreads through networks.

信息在网络上传播的传播途径。

It turns out things like viral videos

结果是，流行的视频、

or Facebook likes or other information

脸书上得到很多“赞”的内容、

spreads in exactly the same way

或者其他信息的传播，

that diseases spread through social networks.

同疾病在社交网络中蔓延的方式是相同的。

So this is something we've studied for a long time.

我们在这方面已经研究很久了。

We have good models of it.

我们己经建立了很好的模型。

And so you can put those things together

你能够将所有这些事物放在一起，

and start seeing why things like this happen.

看看为什么这样的事情会发生。

So if I were to give you a hypothesis,

如果要我给你一个假说的话，

it would be that a smart guy started this page,

我会猜测一个聪明的人建立了这个页面，

or maybe one of the first people who liked it

或者第一个喜欢这个页面的人

would have scored high on that test.

拥有挺高的智商得分。

And they liked it, and their friends saw it,

他们喜欢了这个页面，然后他们的朋友看到了，

and by homophily, we know that he probably had smart friends,

根据同质性理论，我们知道这些人可能有聪明的朋友，

and so it spread to them, and some of them liked it,

然后他们看到这类信息，他们中的一部分人也喜欢，

and they had smart friends,

他们也有聪明的朋友，

and so it spread to them,

and so it propagated through the network

所以这类信息也传到其他朋友那里，

所以信息就在网络上

to a host of smart people,

在聪明人的圈子里流传开来了，

so that by the end, the action

因此到了最后，

of liking the curly fries page

喜欢炸扭薯的这个页面

is indicative of high intelligence,

就成了高智商的象征，

not because of the content,

而不是因为内容本身，

but because the actual action of liking

而是“喜欢”这一个实际行动

reflects back the common attributes

反映了那些也付诸同样行动的人

of other people who have done it.

的相同特征。

So this is pretty complicated stuff, right?

听起来很复杂，对吧？

It's a hard thing to sit down and explain

对于一般用户来说

to an average user, and even if you do,

它比较难解释清楚，就算你解释清楚了，

what can the average user do about it?

一般用户又能利用它来干嘛呢？

How do you know that you've liked something

你又怎么能知道你喜欢的事情

that indicates a trait for you

反映了你什么特征

that's totally irrelevant to the content of what you've liked?

而且这个特征还和你喜欢的内容毫不相干呢？

There's a lot of power that users don't have

用户其实没有太多的能力

to control how this data is used.

去控制这些数据的使用。

And I see that as a real problem going forward.

我把这个看作将来的真实问题，

So I think there's a couple paths

我认为，要是我们想让用户拥有

that we want to look at

使用这些数据的能力，

if we want to give users some control

那么有几条路径

over how this data is used,

我们需要探究，

because it's not always going to be used

因为这些数据并不总是

for their benefit.

用来为他们谋利益。

An example I often give is that,

这有一个我经常举的例子，

if I ever get bored being a professor,

如果我厌倦了当一名教授，

I'm going to go start a company

我会选择自己开家公司

that predicts all of these attributes

这家公司能预测这些特性和事物

and things like how well you work in teams

例如你在团队里的能力

and if you're a drug user, if you're an alcoholic.

例如你是否是一个吸毒者或酗酒者。

We know how to predict all that.

我们知道如何去预测这些特性。

And I'm going to sell reports

然后我就会把这些报告

to H.R. companies and big businesses

卖给那些人力资源公司

that want to hire you.

和想要雇佣你的大公司。

We totally can do that now.

我们完全可以做到这点。

I could start that business tomorrow,

我明天就能开始这个项目，

and you would have absolutely no control

并且你对我这用使用你的数据

over me using your data like that.

是一点办法也没有的。

That seems to me to be a problem.

这对我来说是一个问题。

So one of the paths we can go down

所以我们可选的其中一条路径

is the policy and law path.

是政策和法律这条途径。

And in some respects, I think that that would be most effective,

某程度上我觉得这可能是最有效的

but the problem is we'd actually have to do it.

但问题是，事实上我们将不得不这么做。

Observing our political process in action

观察我们目前的政治进程

makes me think it's highly unlikely

让我觉得在美国

that we're going to get a bunch of representatives

把一帮代表们聚在一起

to sit down, learn about this,

让他们坐下来理解这个问题，

and then enact sweeping changes

然后颁布有关知识产权法方面的颠覆性条例，

to intellectual property law in the U.S.

让用户掌控自己的数据，

so users control their data.

这似乎是不可能的。

We could go the policy route,

我们可以走政策途径，

where social media companies say,

这样社交媒体公司就会告诉你，

you know what? You own your data.

你知道吗？你的确拥有你的数据。

You have total control over how it's used.

你绝对能自己决定要怎么去用。

The problem is that the revenue models

但问题在于大部分的社交媒体公司

for most social media companies

他们的盈利模式

rely on sharing or exploiting users' data in some way.

在某方面取决于分享或挖掘用户的数据资料。

It's sometimes said of Facebook that the users

所以有时会说面谱网的用户并不是顾客，

aren't the customer, they're the product.

而是产品。

And so how do you get a company

那么你要怎样让一个公司

to cede control of their main asset

将他们的主要资产控制权

back to the users?

双手拱让给用户呢？

It's possible, but I don't think it's something

这是可能的，但我不觉得

that we're going to see change quickly.

我们能很快见证这种改变。

So I think the other path

所以我认为我们得走另一条途径

that we can go down that's going to be more effective

一条更有效的途径，

is one of more science.

一条更加科学的途径。

It's doing science that allowed us to develop

这途径是开发一种技术

all these mechanisms for computing

让我们能够发展所有这些机制

this personal data in the first place.

来首先处理自己的个人信息资料。

And it's actually very similar research

而这很接近

that we'd have to do

我们必须做的研究，

if we want to develop mechanisms

要是我们想要发展这些机制

that can say to a user,

跟用户说明，

"Here's the risk of that action you just took."

“这样做你需要承担那样的风险。”

By liking that Facebook page,

你在面谱网上点“赞”

or by sharing this piece of personal information,

或者分享一些私人信息，

you've now improved my ability

就相当于增强了我的能力

to predict whether or not you're using drugs

去预测你是不是在吸毒

or whether or not you get along well in the workplace.

或者你在工作中是否顺利。

And that, I think, can affect whether or not

我觉得，这样做

people want to share something,

能够影响人们分享的决定：

keep it private, or just keep it offline altogether.

是要保持私隐，还是在网上只字不提。

We can also look at things like

我们也可以探究一些别的，例如

allowing people to encrypt data that they upload,

让人们去给上传的东西加密，

so it's kind of invisible and worthless

那么像面谱网这样的网站

to sites like Facebook

或其他能获取信息的第三方来说

or third party services that access it,

这些信息就隐秘很多，也少了很多意义，

but that select users who the person who posted it

而且只有上传人指定的用户

want to see it have access to see it.

才有浏览的权限。

This is all super exciting research

从智能的角度来看，

from an intellectual perspective,

这是一个非常振奋人心的研究，

and so scientists are going to be willing to do it.

而且科学家们也会乐意去做这样的事。

So that gives us an advantage over the law side.

这样在法律方面，我们就有优势了。

One of the problems that people bring up

当我谈论到这个话题时，

when I talk about this is, they say,

人们提到的其中一个问题，就是

you know, if people start keeping all this data private,

如果当人们开始把这些数据进行保密，

all those methods that you've been developing

那些你研发的用来预测

to predict their traits are going to fail.

人们特性的手段都会作废。

And I say, absolutely, and for me, that's success,

我会说，绝对会作废，但对我来说，这是成功，

because as a scientist,

因为作为一个科学家，

my goal is not to infer information about users,

我的目标不是去推测出用户的信息，

it's to improve the way people interact online.

而是提高人们在网上互动的方式。

And sometimes that involves inferring things about them,

虽然有时涉及到推测用户的资料，

but if users don't want me to use that data,

但如果用户不希望我们用他们的数据，

I think they should have the right to do that.

我觉得他们应该有权去拒绝。

I want users to be informed and consenting

我希望用户能被告知

users of the tools that we develop.

并且赞同我们开发的这种工具。

And so I think encouraging this kind of science

所以我认为，鼓励这类科学，

and supporting researchers

支持这些研究者们

who want to cede some of that control back to users

这些愿意放弃部分控制，退还给用户们，

and away from the social media companies

并且不让社交媒体公司接触数据的研究者们

means that going forward, as these tools evolve

随着这些工具的进化和提高

and advance,

这一切意味着向前的发展，

means that we're going to have an educated

意味着我们将会拥有一个

and empowered user base,

有素质有权力的用户基础，

and I think all of us can agree

我觉得我们都会同意

that that's a pretty ideal way to go forward.

这是一个理想的前进目标。

Thank you.

谢谢。

(Applause)

（掌声）