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

譯者: Adrienne Lin 審譯者: Ying Ru Wu

If you remember that first decade of the web,

如果你還記得網路出現的頭十年，

it was really a static place.

當時是一個很靜態的環境。

You could go online, you could look at pages,

你可以上網、瀏覽網頁，

and they were put up either by organizations

這些網站或許是由一些機構製作，

who had teams to do it

這些機構有自己的團隊，

or by individuals who were really tech-savvy

或是當時很懂科技的人製作的。

for the time.

And with the rise of social media

隨著社交媒體、

and social networks in the early 2000s,

社交網路在 21 世紀初期的興起，

the web was completely changed

網路世界完全改變了。

to a place where now the vast majority of content

現在的網路有很多內容

我們互動的內容是由網路用戶放上網的，

we interact with is put up by average users,

either in YouTube videos or blog posts

不管是 YouTube 上的影片或者部落格，

or product reviews or social media postings.

抑或是商品評價或者社交媒體的文章。

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

除此之外，網路也多了很多互動。

where people are interacting with others,

人們在網絡上互動，

they're commenting, they're sharing,

他們評論、分享，

they're not just reading.

而不僅是看看而已。

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

臉書不是唯一一個 能做這些事的網站，

but it's the biggest,

但它是最大的。

and it serves to illustrate the numbers.

我們可以通過臉書 來判斷使用人數。

Facebook has 1.2 billion users per month.

臉書每個月的用戶高達 12 億。

So half the Earth's Internet population

也就是說全球一半的網民

is using Facebook.

都在使用臉書。

They are a site, along with others,

這個網站，還有其他的網站，

that has allowed people to create an online persona

讓網民能創建網路上的個人形象

with very little technical skill,

而且無需太多的技術即可操作。

and people responded by putting huge amounts

用戶反應熱烈，上傳大量的

of personal data online.

個人訊息到網路上。

So the result is that we have behavioral,

這樣一來我們就有了有關行為、

preference, demographic data

偏好、地理數據，

for hundreds of millions of people,

提供給成千上萬的人，

which is unprecedented in history.

這是史無前例的。

And as a computer scientist, what this means is that

作為電腦科學家，這就意味著

I've been able to build models

我可以建立很多模型

that can predict all sorts of hidden attributes

用來推測各種隱藏特性，

for all of you that you don't even know

而你們自己可能都不知道

你們分享的訊息透露了這些特性。

you're sharing information about.

As scientists, we use that to help

科學家利用這些數據來改善

the way people interact online,

網民在網路上的互動，

but there's less altruistic applications,

但網路也有一些 沒那麼利他主義的應用，

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

我們面臨一個問題， 那就是網路用戶並不真正

understand these techniques and how they work,

了解這些網路技術、它們的運作原理，

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

而且即使他們懂， 也沒什麼辦法控制其影響。

So what I want to talk to you about today

所以我今天想和你們分享的，

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

是我們力所能及、可控制的一些事情，

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

給大家一些想法，看看我們如何發展才能

to move some control back into the hands of users.

把部分控制權交回到網路用戶的手上。

So this is Target, the company.

這個是 Target 公司。

I didn't just put that logo

我不是沒事把 Target 的標誌放在

on this poor, pregnant woman's belly.

這個可憐孕婦的肚子上。

You may have seen this anecdote that was printed

你可能讀過一個小故事，刊登在

in Forbes magazine where Target

富比士雜誌。故事提到 Target

sent a flyer to this 15-year-old girl

發了張傳單給一位 15 歲的女孩。

with advertisements and coupons

上面的廣告和折價卷

for baby bottles and diapers and cribs

都是嬰兒奶瓶、尿布、嬰兒床的。

two weeks before she told her parents

這還是在她告訴她父親

that she was pregnant.

自己懷孕了之前兩週的事。

Yeah, the dad was really upset.

是的，她的父親很難過。

He said, "How did Target figure out

那為什麼 Target 知道

that this high school girl was pregnant

在這高中女生告訴父母她懷孕以前， 就已經先知道了呢？

before she told her parents?"

It turns out that they have the purchase history

原來，Target 有購物記錄，

for hundreds of thousands of customers

記錄成千上萬網路顧客的購物歷史，

and they compute what they call a pregnancy score,

而且他們還有一個叫做 “懷孕分數”的計算系統，

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

這個系統不只計算一位女性是否懷孕，

but what her due date is.

還有她們的預產期。

And they compute that

另外，他們不僅探討一些很明顯的資訊，

not by looking at the obvious things,

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

比如說購買了一張嬰兒床、嬰兒服，

but things like, she bought more vitamins

還會計算她買了比平時多的維他命，

than she normally had,

or she bought a handbag

或者是她買了一個 大小足夠放下尿布的包包。

that's big enough to hold diapers.

And by themselves, those purchases don't seem

對購買者來說， 他們並不覺得這些購物訊息

like they might reveal a lot,

透露很多隱私，

but it's a pattern of behavior that,

但其實這是一種行為模式，

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

當你把和成千上萬 網友的資料放在一起看，

starts to actually reveal some insights.

其實就能推測出很多東西。

So that's the kind of thing that we do

所以這些就是我們所做的事情，

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

我們在社群網站上 推測與你們相關的東西。

We're looking for little patterns of behavior that,

我們要找的行為模式是，

when you detect them among millions of people,

當你們從上百萬人身上發現這種模式，

lets us find out all kinds of things.

我們就能找到所有相關的事情。

So in my lab and with colleagues,

所以我和實驗室的同事們，

we've developed mechanisms where we can

開發了多種機制，幫助我們

quite accurately predict things

較精確地推斷很多事情，

like your political preference,

像是你的政治傾向、

your personality score, gender, sexual orientation,

性格測試分數、性別、性取向、

religion, age, intelligence,

宗教信仰、年齡、智力，

along with things like

同時還有像是

how much you trust the people you know

你對認識的人有多信任、

and how strong those relationships are.

你們的關係有多緊密等。

We can do all of this really well.

所有這些我們都可以做得很好。

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

而且，這些都不是來自於

think of as obvious information.

你會認為是明顯的訊息。

So my favorite example is from this study

我最喜歡舉的一個例子

that was published this year

是一個今年發表的研究

in the Proceedings of the National Academies.

刊在《美國國家科學院院刊》上。

If you Google this, you'll find it.

Google 一下就能查到。

It's four pages, easy to read.

研究只有四頁紙，很容易讀。

And they looked at just people's Facebook likes,

他們僅是研究了用戶在臉書的點讚，

so just the things you like on Facebook,

只是你在臉書上點讚的內容，

and used that to predict all these attributes,

用這些點讚的內容 來推斷所有這些特性，

along with some other ones.

以及其他的資訊。

And in their paper they listed the five likes

在調查中，他們列出了五類的讚，

that were most indicative of high intelligence.

這些讚最能表明高智商的用戶。

And among those was liking a page

這其中還包括

到炸馬鈴薯圈頁面點讚。（笑聲）

for curly fries. (Laughter)

Curly fries are delicious,

炸馬鈴薯圈是好吃，

but liking them does not necessarily mean

但是到這頁面按讚不表示

that you're smarter than the average person.

你就比一般人聰明。

So how is it that one of the strongest indicators

到底為什麼，

最能體現你智商指數的指標之一

of your intelligence

is liking this page

是到一個頁面按讚，

when the content is totally irrelevant

即使頁面的內容完全無關於

to the attribute that's being predicted?

要推斷的特性？

And it turns out that we have to look at

結論是，我們需要參考

a whole bunch of underlying theories

很多背後的理論

to see why we're able to do this.

來了解為什麼我們能夠做到這點。

One of them is a sociological theory called homophily,

其中一個就是社會學理論，叫同質相吸，

which basically says people are friends with people like them.

指的是人們通常 和與自己相像的人交朋友。

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

所以如果你聰明， 你會和聰明的人交朋友，

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

如果你年輕， 你會和年輕人交朋友，

and this is well established

這個理論是經過驗證的，

for hundreds of years.

多年來大家都肯定。

We also know a lot

我們還知道很多

about how information spreads through networks.

關於訊息在網路上如何傳播。

It turns out things like viral videos

我們發現病毒影片、

or Facebook likes or other information

臉書按讚或是其他訊息

spreads in exactly the same way

傳播的方式完全和

that diseases spread through social networks.

病毒透過社群網站傳播的方式一樣。

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

這是我們研究了很長時間的東西，

We have good models of it.

我們有很好的模型。

And so you can put those things together

所以如果你們把這些模型都放在一起，

and start seeing why things like this happen.

就能了解為何這樣的事情會發生了。

So if I were to give you a hypothesis,

如果要給各位一個假設，

it would be that a smart guy started this page,

那就是一個聰明的人 建立了一個粉絲頁，

or maybe one of the first people who liked it

或者剛開始幾個去按讚的人

would have scored high on that test.

在智力測試上得了高分，

And they liked it, and their friends saw it,

他們給這個頁面點了讚， 當他們的朋友看見了，

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

根據同質相吸的原理，我們知道 這些人的朋友可能也很聰明，

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

當訊息傳給他們， 有些人也會給這個頁面點讚，

and they had smart friends,

而他們又有聰明的朋友，

and so it spread to them,

and so it propagated through the network

訊息接著傳出去，

這樣一來，就在網路上傳開了，

to a host of smart people,

傳給一群聰明的人，

so that by the end, the action

如此，到最後

of liking the curly fries page

給炸馬鈴薯圈頁面點讚的行為

is indicative of high intelligence,

就成了高智商的指標，

not because of the content,

並不是因為頁面的內容，

but because the actual action of liking

而是因為點讚的這一行為

reflects back the common attributes

反映了做這件事情的人的

of other people who have done it.

共同特性。

So this is pretty complicated stuff, right?

所以這還是挺複雜的，是吧？

It's a hard thing to sit down and explain

要坐下來跟普通用戶解釋是困難的，

to an average user, and even if you do,

而且即使我們分析了，

what can the average user do about it?

對普通用戶們又有什麼用呢？

How do you know that you've liked something

你們怎麼知道到某個粉絲頁按讚

that indicates a trait for you

能夠反映出你的特性，

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

而這特性又和你按讚的內容 完全無關呢？

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

很多的權力用戶都沒有，

to control how this data is used.

他們沒法控制這些數據的使用。

And I see that as a real problem going forward.

我認為這是我們繼續發展 所面臨的真正困難。

So I think there's a couple paths

所以我想到了幾條途徑

that we want to look at

我們可以參考，

if we want to give users some control

看能不能給用戶一些

over how this data is used,

控制這些數據的方法。

because it's not always going to be used

因為這些數據並不總是

for their benefit.

能替用戶帶來益處。

An example I often give is that,

我常舉例說，

if I ever get bored being a professor,

如果我厭倦當教授，

I'm going to go start a company

我要開個公司

that predicts all of these attributes

去推斷所有這些用戶特性，

and things like how well you work in teams

像是你的團隊合作、

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

嗑不嗑藥、是不是酒鬼。

We know how to predict all that.

我們知道如何去推斷這些訊息。

And I'm going to sell reports

接著我就要把這些報告

to H.R. companies and big businesses

賣給人力資源公司或者大企業

that want to hire you.

就是那些將要雇你的人。

We totally can do that now.

我們現在完全可以做到這些。

I could start that business tomorrow,

我明天就可以開始做，

and you would have absolutely no control

而且你完全沒法控制

over me using your data like that.

我這樣使用數據的行為。

That seems to me to be a problem.

這在我看來是一個問題。

So one of the paths we can go down

所以我們能選擇的 其中一條途徑就是

is the policy and law path.

政策和法律的制定。

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

在某種程度上， 我認為這將是最有效的方法，

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

但問題是我們必須得實際執行。

Observing our political process in action

透過觀察我們的政治進程，

makes me think it's highly unlikely

讓我意識到我們很難

that we're going to get a bunch of representatives

集合一群代表，

to sit down, learn about this,

讓他們坐下來了解這件事，

and then enact sweeping changes

然後開始進行大規模改變，

to intellectual property law in the U.S.

修改美國的知識產權法律

so users control their data.

以讓用戶有權控制他們的數據。

We could go the policy route,

我們可以走政策道路，

where social media companies say,

讓社群公司表態，

you know what? You own your data.

「好，你們擁有自己的數據。

You have total control over how it's used.

你們能完全地控制對它們的使用。」

The problem is that the revenue models

問題在於

多數社交媒體的收益模式

for most social media companies

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

某種程度上仰賴 分享或利用用戶的數據。

It's sometimes said of Facebook that the users

有人說臉書的用戶

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

不是顧客，而是產品。

And so how do you get a company

所以你怎麼可能讓一間公司

to cede control of their main asset

放棄對他們主要收入的控制

back to the users?

把控制權還給用戶呢？

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

這是有可能的，但我不認為

that we're going to see change quickly.

我們能很快看到這一改變。

So I think the other path

所以我認為另外一條途徑

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

一條更有效的途徑，

is one of more science.

是更科學的途徑。

It's doing science that allowed us to develop

正是透過科學，我們才能開發

all these mechanisms for computing

所有的這些機制首先用於計算個人數據

this personal data in the first place.

And it's actually very similar research

事實上，有個很類似的研究，

that we'd have to do

if we want to develop mechanisms

如果我們要發明一些機制

that can say to a user,

是可以對用戶說

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

「這是你剛才所做的行為 要面臨的風險。」

By liking that Facebook page,

藉由臉書按讚，

or by sharing this piece of personal information,

或者是分享私人資訊，

you've now improved my ability

你現在給了我更多能力

to predict whether or not you're using drugs

去推斷你是否嗑藥

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

或者你是否和同事相處融洽。

And that, I think, can affect whether or not

我認為這些會影響

people want to share something,

人們是否願意分享事情、

keep it private, or just keep it offline altogether.

還是設為隱私，或者是完全不放上網絡。

We can also look at things like

我們還可以研究一些像是

allowing people to encrypt data that they upload,

讓用戶可以加密他們上傳的數據，

so it's kind of invisible and worthless

所以對像是臉書的網站， 這是隱形而且無用的，

to sites like Facebook

or third party services that access it,

或者是第三方服務網站也是如此。

but that select users who the person who posted it

但是用戶可選擇上傳的東西

want to see it have access to see it.

要讓誰有權可以看到。

This is all super exciting research

如果我們從知識的角度去看，

from an intellectual perspective,

這些都是非常令人興奮的研究，

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

所以說科學家會願意做相關的研究。

So that gives us an advantage over the law side.

這比起法律的途徑， 給了我們更多的好處。

One of the problems that people bring up

當我談到這個的時候，

when I talk about this is, they say,

人們常會提出一個疑問，

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

你知道，如果人們開始把這些數據都保密了，

all those methods that you've been developing

你們一直在開發的這些

to predict their traits are going to fail.

用來推斷他們特性的方法都將失效，

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

我回答說，完全正確，

但對我來說，那就是成功。

because as a scientist,

因為身為一名科學家，

my goal is not to infer information about users,

我的目標不是要推斷用戶的資訊，

it's to improve the way people interact online.

而是要改進人們在網路互動的方式。

And sometimes that involves inferring things about them,

有時候這包括推斷關於他們的事情，

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

但如果用戶不想要我使用這些數據，

I think they should have the right to do that.

我認為他們有權利這麼做。

I want users to be informed and consenting

我希望用戶們可以知道且同意

users of the tools that we develop.

我們一直開發這些工具。

And so I think encouraging this kind of science

所以，我認為推廣這類科學、

and supporting researchers

支持研究者，

who want to cede some of that control back to users

支持那些希望把控制權 交回到用戶手中，

and away from the social media companies

從社群媒體公司 拿回這些權利的研究者，

means that going forward, as these tools evolve

意味著隨著這些工具進化和發展， 我們是向前發展的。

and advance,

means that we're going to have an educated

我們將有一組教育程度更高、 更有力的用戶數據，

and empowered user base,

and I think all of us can agree

我相信大家都會認同

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

朝此理想的發展方式前進。

Thank you.

謝謝。

(Applause)

（掌聲）