Placeholder Image

字幕列表 影片播放

  • For the last 10 years, I've been spending my time trying to figure out

    過去10年來,我試著了解,

  • how and why human beings

    人們為何形成社交網路,

  • assemble themselves into social networks.

    以及這些網路是如何形成的。

  • And the kind of social network I'm talking about

    我所要談的網路,

  • is not the recent online variety,

    並非現在所謂的網路社群。

  • but rather, the kind of social networks

    而是更原始的社交網路,

  • that human beings have been assembling for hundreds of thousands of years,

    自從人類在非洲大草原出現以來,

  • ever since we emerged from the African savannah.

    已經使用這種連結十幾萬年了。

  • So, I form friendships and co-worker

    我和其他人分享友誼、同事、

  • and sibling and relative relationships with other people

    手足和親戚等等人際關係,

  • who in turn have similar relationships with other people.

    這些人也和其他人有相似的連結。

  • And this spreads on out endlessly into a distance.

    這樣的連結向外擴散,

  • And you get a network that looks like this.

    從而得到的網路看起來會像這樣。

  • Every dot is a person.

    每點代表一個人,

  • Every line between them is a relationship between two people --

    兩點間的線則代表兩個人之間的關係,

  • different kinds of relationships.

    各種不同的關係。

  • And you can get this kind of vast fabric of humanity,

    種種的關係交織成一幅巨大的網路,

  • in which we're all embedded.

    而我們都位於其中。

  • And my colleague, James Fowler and I have been studying for quite sometime

    我的同事James Fowler和我花了滿長時間研究,

  • what are the mathematical, social,

    想找到一個基於數學、社會學、

  • biological and psychological rules

    生物學或是心理學的規則,

  • that govern how these networks are assembled

    能夠主導這些網路的形成。

  • and what are the similar rules

    以及是否有類似的規則

  • that govern how they operate, how they affect our lives.

    主導網路的運作,進而影響我們的生活。

  • But recently, we've been wondering

    直到最近,我們開始思考,

  • whether it might be possible to take advantage of this insight,

    是否有可能利用這些發現,

  • to actually find ways to improve the world,

    來找出增進人類福祉的方法,

  • to do something better,

    改善現況,

  • to actually fix things, not just understand things.

    去導正,而非只是單純理解問題。

  • So one of the first things we thought we would tackle

    我們最先著手研究的議題,

  • would be how we go about predicting epidemics.

    是如何預測流行趨勢。

  • And the current state of the art in predicting an epidemic --

    目前最先進的預測方法—

  • if you're the CDC or some other national body --

    如果你在疾病管制中心(CDC)或類似的政府單位工作—

  • is to sit in the middle where you are

    是待在中央枯等,

  • and collect data

    並收集資料,

  • from physicians and laboratories in the field

    第一線的醫生和實驗室把資料傳進來,

  • that report the prevalence or the incidence of certain conditions.

    報告疾病的流行程度或發生機率。

  • So, so and so patients have been diagnosed with something,

    這邊有某個病患被診斷出來,

  • or other patients have been diagnosed,

    那邊又有別人得病。

  • and all these data are fed into a central repository, with some delay.

    資訊經過一些延遲之後,傳進中央的資料庫裡。

  • And if everything goes smoothly,

    如果一切順利,

  • one to two weeks from now

    一到兩個禮拜之後,

  • you'll know where the epidemic was today.

    我們才會得知當天流行病的狀況。

  • And actually, about a year or so ago,

    事實上一年多以前,

  • there was this promulgation

    有人發表了這樣的概念,

  • of the idea of Google Flu Trends, with respect to the flu,

    使用Google流感趨勢(Flu Trends)來尋找流感。

  • where by looking at people's searching behavior today,

    透過對搜尋行為的分析,

  • we could know where the flu --

    我們能夠得知流感發生的區域,

  • what the status of the epidemic was today,

    得知當天傳染病的狀態,

  • what's the prevalence of the epidemic today.

    以及傳染病的影響程度。

  • But what I'd like to show you today

    不過這次我要介紹的方法,

  • is a means by which we might get

    讓我們不只能夠

  • not just rapid warning about an epidemic,

    得到傳染病的快速預警,

  • but also actually

    更能夠讓我們

  • early detection of an epidemic.

    提早偵測到流行病的發生。

  • And, in fact, this idea can be used

    事實上,這個概念不止能夠

  • not just to predict epidemics of germs,

    用來預測病菌的流行,

  • but also to predict epidemics of all sorts of kinds.

    也能夠應用來預測各種事物的趨勢。

  • For example, anything that spreads by a form of social contagion

    例如,任何能透過社群的方式傳播的事物,

  • could be understood in this way,

    都可以用這種方式理解。

  • from abstract ideas on the left

    從左邊的抽象概念,

  • like patriotism, or altruism, or religion

    像是愛國主義、利他精神,或是宗教,

  • to practices

    到具體的事物,

  • like dieting behavior, or book purchasing,

    像是飲食行為、購買書籍、

  • or drinking, or bicycle-helmet [and] other safety practices,

    酗酒、使用腳踏車安全帽等安全措施,

  • or products that people might buy,

    或是日常用品,

  • purchases of electronic goods,

    電子產品,

  • anything in which there's kind of an interpersonal spread.

    任何透過人與人之間傳遞的事物。

  • A kind of a diffusion of innovation

    這種創新的擴散,

  • could be understood and predicted

    可以透過接下來我將展示的機制,

  • by the mechanism I'm going to show you now.

    來理解並且預測。

  • So, as all of you probably know,

    你們或許知道,

  • the classic way of thinking about this

    最經典的範例,

  • is the diffusion-of-innovation,

    就是創新的擴散,

  • or the adoption curve.

    或是所謂的「普及曲線」。

  • So here on the Y-axis, we have the percent of the people affected,

    Y軸是受影響人數的百分比,

  • and on the X-axis, we have time.

    X軸表示時間的推移。

  • And at the very beginning, not too many people are affected,

    剛開始沒有太多人受到影響,

  • and you get this classic sigmoidal,

    然後你會看到經典的反曲線,

  • or S-shaped, curve.

    或是S型曲線。

  • And the reason for this shape is that at the very beginning,

    形成這種曲線的原因是,

  • let's say one or two people

    一開始只有一兩個人

  • are infected, or affected by the thing

    被影響,或是被「感染」,

  • and then they affect, or infect, two people,

    然後傳遞給另外兩個人,

  • who in turn affect four, eight, 16 and so forth,

    接著4、8、16,以此類推,

  • and you get the epidemic growth phase of the curve.

    這時進入迅速增長的階段。

  • And eventually, you saturate the population.

    最終擴散到整個群體。

  • There are fewer and fewer people

    於是越來越難找到

  • who are still available that you might infect,

    尚未被影響的人,

  • and then you get the plateau of the curve,

    這時候曲線進入高原期,

  • and you get this classic sigmoidal curve.

    形成整條反曲線。

  • And this holds for germs, ideas,

    這個模式在病菌、創意、

  • product adoption, behaviors,

    新產品的普及、行為,

  • and the like.

    以及類似情況都適用。

  • But things don't just diffuse in human populations at random.

    要注意的是,事物並不是隨機在人群中蔓延,

  • They actually diffuse through networks.

    而是隨著網路分布來擴散。

  • Because, as I said, we live our lives in networks,

    因為我們活在網路的世界,

  • and these networks have a particular kind of a structure.

    而這種網路有特定的結構。

  • Now if you look at a network like this --

    觀察這個網路,

  • this is 105 people.

    裡面有105人。

  • And the lines represent -- the dots are the people,

    每個點代表一個人

  • and the lines represent friendship relationships.

    每條線代表彼此間的友誼關係。

  • You might see that people occupy

    人們在這個網路中

  • different locations within the network.

    佔據不同的位置,

  • And there are different kinds of relationships between the people.

    彼此間有不同類型的關係。

  • You could have friendship relationships, sibling relationships,

    可能是朋友、手足、

  • spousal relationships, co-worker relationships,

    配偶、同事、

  • neighbor relationships and the like.

    鄰居等等。

  • And different sorts of things

    不同的事物會

  • spread across different sorts of ties.

    透過不同的關係來傳播。

  • For instance, sexually transmitted diseases

    例如,性傳染病,

  • will spread across sexual ties.

    會藉由性伴侶的聯繫來散佈。

  • Or, for instance, people's smoking behavior

    或者像人們吸菸,

  • might be influenced by their friends.

    可能是受到朋友的影響。

  • Or their altruistic or their charitable giving behavior

    人們的善行或捐助,

  • might be influenced by their coworkers,

    可能是出自同事間的影響,

  • or by their neighbors.

    或是他們鄰居的行為。

  • But not all positions in the network are the same.

    但是網路中的位置並非都一樣。

  • So if you look at this, you might immediately grasp

    這張圖或許能讓你了解,

  • that different people have different numbers of connections.

    不同人有不同數量的連結。

  • Some people have one connection, some have two,

    有的人一個,有人兩個,

  • some have six, some have 10 connections.

    有人六個,有的人擁有十個連結。

  • And this is called the "degree" of a node,

    也就是一個節點的「度數」,

  • or the number of connections that a node has.

    或是一個節點所擁有的連結數。

  • But in addition, there's something else.

    除此之外,

  • So, if you look at nodes A and B,

    如果觀察節點A與B,

  • they both have six connections.

    兩者都擁有六個連結。

  • But if you can see this image [of the network] from a bird's eye view,

    但是如果鳥瞰整個圖像,

  • you can appreciate that there's something very different

    你就會發現兩者之間,

  • about nodes A and B.

    A與B的不同之處

  • So, let me ask you this -- I can cultivate this intuition by asking a question --

    問題來了 -請用直覺回答-

  • who would you rather be

    你比較想當誰:

  • if a deadly germ was spreading through the network, A or B?

    如果致命病菌正在網路中散佈,A或是B?

  • (Audience: B.) Nicholas Christakis: B, it's obvious.

    (觀眾:B)很明顯的是B。

  • B is located on the edge of the network.

    B處在網路的邊緣。

  • Now, who would you rather be

    現在,你比較想當誰:

  • if a juicy piece of gossip were spreading through the network?

    如果網路中流傳著一個天大的八卦?

  • A. And you have an immediate appreciation

    A。而且你馬上能夠理解到,

  • that A is going to be more likely

    A會有更高的機率

  • to get the thing that's spreading and to get it sooner

    趕上流行,而且早先一步。

  • by virtue of their structural location within the network.

    這要歸功於他們在網路中的位置。

  • A, in fact, is more central,

    A比較靠近中央,

  • and this can be formalized mathematically.

    這可以用數學形式來描述。

  • So, if we want to track something

    因此,如果我們希望追蹤某些事物

  • that was spreading through a network,

    在網路中散佈的狀態,

  • what we ideally would like to do is to set up sensors

    理想狀況是佈置感測器,

  • on the central individuals within the network,

    對準網路裡的中央個體,

  • including node A,

    包括節點A。

  • monitor those people that are right there in the middle of the network,

    監視這些位於中心位置的人們,

  • and somehow get an early detection

    以早期的預警到

  • of whatever it is that is spreading through the network.

    正在網路上傳播的事物。

  • So if you saw them contract a germ or a piece of information,

    亦即,如果這些人染病或是獲悉某些資訊,

  • you would know that, soon enough,

    你就可以推斷,要不了多久,

  • everybody was about to contract this germ

    所有人都會被波及,不管是染病,

  • or this piece of information.

    或是得到資訊。

  • And this would be much better

    這樣的作法遠勝於

  • than monitoring six randomly chosen people,

    隨機挑選六個人來監控,

  • without reference to the structure of the population.

    因為該做法並未考慮到群體的結構。

  • And in fact, if you could do that,

    若是真的能夠實行,

  • what you would see is something like this.

    我們會得到類似這樣的情況:

  • On the left-hand panel, again, we have the S-shaped curve of adoption.

    左邊的圖表,是S型的普及曲線。

  • In the dotted red line, we show

    我們用紅色虛線標示出,

  • what the adoption would be in the random people,

    一般人的普及情形,

  • and in the left-hand line, shifted to the left,

    左邊的線段,則向左偏移,

  • we show what the adoption would be

    顯示出網路中的核心個體,

  • in the central individuals within the network.

    他們的普及情形。

  • On the Y-axis is the cumulative instances of contagion,

    Y軸是受到傳染「病例」的累積數量,

  • and on the X-axis is the time.

    X軸則是時間。

  • And on the right-hand side, we show the same data,

    右邊的圖表是相同的資料,

  • but here with daily incidence.

    呈現的是每日的「感染」數字。

  • And what we show here is -- like, here --

    我們想要傳達的是,

  • very few people are affected, more and more and more and up to here,

    一開始少數人受到影響,然後越來越多直到這裡,

  • and here's the peak of the epidemic.

    這裡就是傳播的高峰期。

  • But shifted to the left is what's occurring in the central individuals.

    向左偏的則是在核心個體發生的情形,

  • And this difference in time between the two

    這兩條曲線間的時間差,

  • is the early detection, the early warning we can get,

    就是預測時差,我們可以從中得到預警,

  • about an impending epidemic

    人群中是否有

  • in the human population.

    即將爆發的疫情。

  • The problem, however,

    然而問題在於,

  • is that mapping human social networks

    人際間的社交網路,

  • is not always possible.

    並不容易繪測。

  • It can be expensive, not feasible,

    這樣的計畫可能所費不貲、非常困難、

  • unethical,

    具有道德爭議

  • or, frankly, just not possible to do such a thing.

    說實話,就是不可能。

  • So, how can we figure out

    所以,我們要如何找出,

  • who the central people are in a network

    網路中的核心個體在哪,

  • without actually mapping the network?

    而無需繪出整個網路?

  • What we came up with

    我們所想到的,

  • was an idea to exploit an old fact,

    是利用一個既有的事實

  • or a known fact, about social networks,

    關於社交網路,眾所皆知的事實。

  • which goes like this:

    也就是:

  • Do you know that your friends

    你知道你的朋友,

  • have more friends than you do?

    所擁有的友人數目比你還多嗎?

  • Your friends have more friends than you do,

    朋友的友人數目比自己擁有的還多,

  • and this is known as the friendship paradox.

    通常這種情況被稱做「友誼悖論」。

  • Imagine a very popular person in the social network --

    試想社交網路中的人氣王 -

  • like a party host who has hundreds of friends --

    例如派對的主人,身邊有上百個朋友 --

  • and a misanthrope who has just one friend,

    和孤僻成性,只有一個朋友的人。

  • and you pick someone at random from the population;

    若是你隨便從人群中挑出一位,

  • they were much more likely to know the party host.

    他們就非常有可能認識這位派對主人,

  • And if they nominate the party host as their friend,

    而當他們舉出派對主人是自己的朋友,

  • that party host has a hundred friends,

    由於他有上百個朋友,

  • therefore, has more friends than they do.

    因此遠比自己的朋友數目還多。

  • And this, in essence, is what's known as the friendship paradox.

    在本質上,這就是友誼悖論:

  • The friends of randomly chosen people

    隨機挑選的人,他的朋友,

  • have higher degree, and are more central

    會有較高的連結數目,也較為趨近核心,

  • than the random people themselves.

    因而優於那些隨機挑選的人。

  • And you can get an intuitive appreciation for this

    因此,你可以憑直覺想像,

  • if you imagine just the people at the perimeter of the network.

    如果是那些位於網路邊緣的人,

  • If you pick this person,

    這樣的人,

  • the only friend they have to nominate is this person,

    他的朋友只會有這個人,

  • who, by construction, must have at least two

    而結構上來說,這個人至少會有兩位、

  • and typically more friends.

    甚至更多的朋友。

  • And that happens at every peripheral node.

    在每個外圍的節點都是這樣。

  • And in fact, it happens throughout the network as you move in,

    當你越往網路的中心移動時就越常見,

  • everyone you pick, when they nominate a random --

    每個被你挑到的人,當他們隨意提出一個...

  • when a random person nominates a friend of theirs,

    每當提出一個他們的朋友,

  • you move closer to the center of the network.

    你就越靠近網路的中心。

  • So, we thought we would exploit this idea

    於是我們認為可以利用這個概念,

  • in order to study whether we could predict phenomena within networks.

    來研究我們是否能預測網路中所發生的現象。

  • Because now, with this idea

    因為有了這樣的發現,

  • we can take a random sample of people,

    我們可以從人群中隨機挑選樣本,

  • have them nominate their friends,

    請他們指出他們的朋友,

  • those friends would be more central,

    這些朋友會比較靠近中心,

  • and we could do this without having to map the network.

    而我們就無須標出整個網路的圖像。

  • And we tested this idea with an outbreak of H1N1 flu

    在哈佛大學,我們利用H1N1流感的爆發

  • at Harvard College

    來測試這個概念。

  • in the fall and winter of 2009, just a few months ago.

    在2009年秋冬,只有幾個月前,

  • We took 1,300 randomly selected undergraduates,

    我們隨機挑選了1300位大學生,

  • we had them nominate their friends,

    請這些人提供他們的朋友名單,

  • and we followed both the random students and their friends

    我們同時追蹤了這些人和他們的朋友,

  • daily in time

    每天為間隔,

  • to see whether or not they had the flu epidemic.

    確認他們是否染上流感。

  • And we did this passively by looking at whether or not they'd gone to university health services.

    除了被動觀察他們是否去健康中心報到,

  • And also, we had them [actively] email us a couple of times a week.

    同時也要求每個禮拜Email給我們。

  • Exactly what we predicted happened.

    結果一如我們所預期。

  • So the random group is in the red line.

    隨機挑選的群體用紅線標示,

  • The epidemic in the friends group has shifted to the left, over here.

    他們的朋友則向左邊偏移,在這邊。

  • And the difference in the two is 16 days.

    兩者間的差距是16天。

  • By monitoring the friends group,

    觀察朋友的群體,

  • we could get 16 days advance warning

    能夠讓我們提早16天得到警示,

  • of an impending epidemic in this human population.

    警告人群中即將爆發的傳染病。

  • Now, in addition to that,

    除此之外,

  • if you were an analyst who was trying to study an epidemic

    如果你是研究傳染病的分析師,

  • or to predict the adoption of a product, for example,

    或者想要預測產品的普及情形。

  • what you could do is you could pick a random sample of the population,

    你可以從人群中挑選隨機樣本,

  • also have them nominate their friends and follow the friends

    請他們指出自己的朋友,