The human insights missing from big data | Tricia Wang

譯者: Shizumi Ch 審譯者: Wilde Luo

In ancient Greece,

古希臘時期，

when anyone from slaves to soldiers, poets and politicians,

不論是奴隸或士兵，詩人或政治家，

needed to make a big decision on life's most important questions,

當他們人生遇到重大問題時， 需要做出重要的決定，

like, "Should I get married?"

像是「我該結婚嗎？」

or "Should we embark on this voyage?"

或是「我該開始這次的航行嗎？」

or "Should our army advance into this territory?"

或是「我的士兵該進攻這個領地嗎？」

they all consulted the oracle.

他們都會請示先知。

So this is how it worked:

運行模式是這樣的：

you would bring her a question and you would get on your knees,

你把問題告訴她，接著屈膝跪下，

and then she would go into this trance.

然後她就會進入出神狀態。

It would take a couple of days,

這會花上幾天的時間，

and then eventually she would come out of it,

最終她會回神，

giving you her predictions as your answer.

答復你她的預知。

From the oracle bones of ancient China

從古中國的甲骨文， 到古希臘，再到馬雅曆，

to ancient Greece to Mayan calendars,

people have craved for prophecy

人們都渴求著預言，

in order to find out what's going to happen next.

為了知道接下來會發生什麼事。

And that's because we all want to make the right decision.

而這是因為我們都想做正確的決定，

We don't want to miss something.

我們不希望漏掉了什麼。

The future is scary,

未來令人害怕。

so it's much nicer knowing that we can make a decision

所以能在某種程度上 保障決定的結果，是很棒的事。

with some assurance of the outcome.

Well, we have a new oracle,

我們有了新的先知，

and it's name is big data,

名字叫大數據。

or we call it "Watson" or "deep learning" or "neural net."

也可以稱它為「華生」、 「深度學習」或「人工神經網路」。

And these are the kinds of questions we ask of our oracle now,

如今我們會問先知這樣的問題：

like, "What's the most efficient way to ship these phones

「要將這批手機從中國 運到瑞典，怎樣最有效率？」

from China to Sweden?"

Or, "What are the odds

或是「我的小孩出生就有 遺傳疾病的機率是多少？」

of my child being born with a genetic disorder?"

Or, "What are the sales volume we can predict for this product?"

或是「預期這產品的銷售量多少？」

I have a dog. Her name is Elle, and she hates the rain.

我養了隻狗，名叫埃萊，最討厭下雨。

And I have tried everything to untrain her.

我用盡方法來訓練她， 讓她適應下雨。

But because I have failed at this,

但因為我失敗了，

I also have to consult an oracle, called Dark Sky,

我還是得諮詢一位叫 Dark Sky（天氣預報公司）的先知，

every time before we go on a walk,

每次散步之前都會諮詢，

for very accurate weather predictions in the next 10 minutes.

以獲得接下來十分鐘的準確天氣預報。

She's so sweet.

她真的很貼心。

So because of all of this, our oracle is a $122 billion industry.

基於這些理由，我們的「先知」 是個 1220 億美元的產業。

Now, despite the size of this industry,

先不論這個產業的規模，

the returns are surprisingly low.

令人驚訝的是它極低的報酬率。

Investing in big data is easy,

投資大數據很簡單，

but using it is hard.

運用大數據卻很難。

Over 73 percent of big data projects aren't even profitable,

73% 以上的大數據計畫根本不賺錢，

and I have executives coming up to me saying,

有些業務主管跑來跟我說，

"We're experiencing the same thing.

「我們都面臨了同樣的問題。

We invested in some big data system,

我們投資了幾個大數據系統，

and our employees aren't making better decisions.

但我們的員工卻還是不能 做出更優的決定。

And they're certainly not coming up with more breakthrough ideas."

他們當然也沒有想出 更多突破性的點子。」

So this is all really interesting to me,

這些對我來說都很有趣，

because I'm a technology ethnographer.

因為我是個科技人類學家。

I study and I advise companies

我研究並給予公司建議，

on the patterns of how people use technology,

告訴他們人們使用科技的形態，

and one of my interest areas is data.

我有興趣的領域之一就是數據。

So why is having more data not helping us make better decisions,

為什麼獲得更多數據 卻沒有幫我們做更好的決定，

especially for companies who have all these resources

特別是那些有資源， 可以投資大數據系統的公司？

to invest in these big data systems?

Why isn't it getting any easier for them?

為什麼他們沒有更好地做決定？

So, I've witnessed the struggle firsthand.

我第一時間就目睹了這項困境。

In 2009, I started a research position with Nokia.

2009 年，我開始了 在諾基亞的研究工作。

And at the time,

當時，諾基亞是世界上 最大的手機公司之一，

Nokia was one of the largest cell phone companies in the world,

dominating emerging markets like China, Mexico and India --

在中國、墨西哥、印度等 新興市場中佔有主要地位──

all places where I had done a lot of research

我在這些地方都做了很多研究，

on how low-income people use technology.

研究低收入的人怎麼使用科技產品。

And I spent a lot of extra time in China

我在中國花了特別多時間

getting to know the informal economy.

來了解地下經濟。

So I did things like working as a street vendor

所以我當過街頭攤販，

selling dumplings to construction workers.

賣水餃給建築工人。

Or I did fieldwork,

我也做過實地調查，

spending nights and days in internet cafés,

在網咖中日日夜夜地待著，

hanging out with Chinese youth, so I could understand

和中國年輕人來往，這樣我才知道

how they were using games and mobile phones

他們怎麼玩遊戲、使用手機，

and using it between moving from the rural areas to the cities.

以及他們從農村地區 移居到城市時的使用情形。

Through all of this qualitative evidence that I was gathering,

透過我收集的定性資料，

I was starting to see so clearly

我開始清楚看見

that a big change was about to happen among low-income Chinese people.

即將發生在低收入中國人身上的巨變。

Even though they were surrounded by advertisements for luxury products

雖然他們身邊圍繞著奢侈品的廣告，

like fancy toilets -- who wouldn't want one? --

像是花俏的馬桶──誰不想要呢──

and apartments and cars,

還有公寓和車，

through my conversations with them,

從和他們的對話中，

I found out that the ads the actually enticed them the most

我發現最吸引他們的廣告，

were the ones for iPhones,

是 iPhone 的廣告，

promising them this entry into this high-tech life.

那些廣告向他們保證了 進入高科技生活的途徑。

And even when I was living with them in urban slums like this one,

即使我和他們一起 住在這樣的城市貧民窟，

I saw people investing over half of their monthly income

我也看到人們將半個月以上的收入

into buying a phone,

拿去買手機，

and increasingly, they were "shanzhai,"

而且越來越多都是「山寨品」，

which are affordable knock-offs of iPhones and other brands.

也就是他們買得起的 iPhone 或其他品牌的仿冒品。

They're very usable.

這些仿冒品很堪使用。

Does the job.

原廠有的功能都能用。

And after years of living with migrants and working with them

我和移民一起住、一起工作了數年，

and just really doing everything that they were doing,

真的是他們做什麼，我就做什麼，

I started piecing all these data points together --

我開始將所有數據拼湊在一起──

from the things that seem random, like me selling dumplings,

不論是看似不相關的事， 像是我賣水餃的事，

to the things that were more obvious,

或是較明顯相關的事，

like tracking how much they were spending on their cell phone bills.

像是追蹤他們花多少錢付手機費。

所以我才有辦法描繪出 這麼多整體畫面

And I was able to create this much more holistic picture

of what was happening.

來說明當時正發生什麼事。

And that's when I started to realize

這時我才開始理解到

that even the poorest in China would want a smartphone,

連中國最窮的人也想要智慧型手機，

and that they would do almost anything to get their hands on one.

且他們幾乎會不擇手段拿到手。

You have to keep in mind,

你們要記得，

iPhones had just come out, it was 2009,

當時是 2009 年，iPhone 才剛出現，

so this was, like, eight years ago,

這是八年前的事，

and Androids had just started looking like iPhones.

安卓手機才剛開始像 iPhone。

And a lot of very smart and realistic people said,

很多聰明又現實的人說，

"Those smartphones -- that's just a fad.

「智慧型手機只是一時的流行。

Who wants to carry around these heavy things

誰會想帶著這麼重的東西到處走，

where batteries drain quickly and they break every time you drop them?"

又很快就沒電，

還會一掉地就壞？」

But I had a lot of data,

但我有很多數據，

and I was very confident about my insights,

我對自己的洞察觀點非常有自信，

so I was very excited to share them with Nokia.

我興奮地把數據告訴諾基亞。

But Nokia was not convinced,

但我沒能說服諾基亞，

because it wasn't big data.

因為那不是大數據。

They said, "We have millions of data points,

他們說：「我們有幾百萬則數據，

and we don't see any indicators of anyone wanting to buy a smartphone,

而我們沒見到任何數據 指出有人想買智慧型手機，

and your data set of 100, as diverse as it is, is too weak

你的 100 組數據太缺乏多樣性，

for us to even take seriously."

我們完全無法重視這項數據。」

And I said, "Nokia, you're right.

我說：「諾基亞，你說的沒錯。

Of course you wouldn't see this,

你當然不會看到有人要買，

because you're sending out surveys assuming that people don't know

因為你所發送問卷的假設前提

是人們不知道智慧型手機是什麼，

what a smartphone is,

so of course you're not going to get any data back

所以你的數據當然不會反映

about people wanting to buy a smartphone in two years.

兩年內想買智慧型手機的人的想法。

Your surveys, your methods have been designed

你問卷、研究方法的設計理念

to optimize an existing business model,

都是想讓現有的業務型態更好，

and I'm looking at these emergent human dynamics

而我關注的是這些正浮現的人類動態，

that haven't happened yet.

那些是過去沒有發生的，

We're looking outside of market dynamics

我們看的是市場動態之外，

so that we can get ahead of it."

這樣我們才能先走一步。」

Well, you know what happened to Nokia?

你們知道諾基亞怎麼樣了嗎？

Their business fell off a cliff.

他們的產業跌落谷底。

This -- this is the cost of missing something.

這就是錯失的代價。

It was unfathomable.

那代價是深不可測的。

But Nokia's not alone.

但不是只有諾基亞這樣。

I see organizations throwing out data all the time

我看到各機構一天到晚丟棄數據，

because it didn't come from a quant model

因為數據並非來自數量大的模型，

or it doesn't fit in one.

或對不上數量大的模型數據。

But it's not big data's fault.

但這不是大數據的錯。

It's the way we use big data; it's our responsibility.

是我們用錯方法，

是我們的責任。

Big data's reputation for success

但一般認為大數據的成功之處

comes from quantifying very specific environments,

在於量化的對象非常的特定，

like electricity power grids or delivery logistics or genetic code,

像是電網、物流運送或遺傳密碼，

when we're quantifying in systems that are more or less contained.

也就是些基本上可操縱的系統。

But not all systems are as neatly contained.

但並非所有的系統 都能被操縱得好好的。

When you're quantifying and systems are more dynamic,

若你在量化的系統是動態的，

especially systems that involve human beings,

特別是那些有人參與其中的系統，

forces are complex and unpredictable,

會產生影響的事物複雜又難以預測，

and these are things that we don't know how to model so well.

我們不太知道怎樣建立這些模型。

Once you predict something about human behavior,

即使你一時預測了人的行動，

new factors emerge,

又會出現新的要素，

because conditions are constantly changing.

因為情況持續在改變。

That's why it's a never-ending cycle.

正因如此，這是個永無止境的迴圈。

You think you know something,

你以為你瞭解了一件事，

and then something unknown enters the picture.

另一件未知的事物便進入了你的視野。

And that's why just relying on big data alone

所以純粹依靠大數據

increases the chance that we'll miss something,

便增加了我們錯失的機率，

while giving us this illusion that we already know everything.

但同時讓我們以為我們無所不知。

And what makes it really hard to see this paradox

為什麼我們很難發現這個矛盾，

and even wrap our brains around it

甚至也很難去理解它，

is that we have this thing that I call the quantification bias,

是因為我們有我所謂的「量化成見」，

which is the unconscious belief of valuing the measurable

也就是無意識地認為可量化的

over the immeasurable.

比不可量化的更有價值。

And we often experience this at our work.

我們工作時常有這樣的經驗。

Maybe we work alongside colleagues who are like this,

或許我們和這樣想的同事一起工作，

or even our whole entire company may be like this,

或者整個公司都這樣想，

where people become so fixated on that number,

人們過於迷戀數字，

that they can't see anything outside of it,

以至於看不見除此之外的任何東西，

even when you present them evidence right in front of their face.

即使你將證據貼到他們臉上，給他們看。

And this is a very appealing message,

這是個十分吸引人的訊息，

because there's nothing wrong with quantifying;

因為量化並沒有錯；

it's actually very satisfying.

量化事實上很讓人滿意。

I get a great sense of comfort from looking at an Excel spreadsheet,

我看著 Excel 電子表格就覺得安心，

even very simple ones.

即使是很簡單的也一樣。

(Laughter)

（笑聲）

It's just kind of like,

那種感覺就是，

"Yes! The formula worked. It's all OK. Everything is under control."

「好的！方程式沒問題。 一切都很好。都在掌控之中。」

But the problem is

問題是，

that quantifying is addictive.

量化會使人上癮。

And when we forget that

我們一旦忘記這件事，

and when we don't have something to kind of keep that in check,

若我們沒能做到時時確認是否上癮，

it's very easy to just throw out data

我們很容易直接扔掉這樣的資料：

because it can't be expressed as a numerical value.

僅僅因為它無法用數值量化。

It's very easy just to slip into silver-bullet thinking,

很容易認為會有完美解決一切的絶招，

as if some simple solution existed.

就好像有某種簡單的解決方法一樣。

Because this is a great moment of danger for any organization,

因為這對任何一間機構來說， 都是危機的重要時刻，

because oftentimes, the future we need to predict --

時常，我們要預測的未來，

it isn't in that haystack,

並不是在這安穩的草堆裡，

but it's that tornado that's bearing down on us

而是在它之外， 是即將襲擊我們的暴風中心。

outside of the barn.

There is no greater risk

沒有什麼比對未知 一無所知來得有風險，

than being blind to the unknown.

It can cause you to make the wrong decisions.

那會使你做出錯誤的決定。

It can cause you to miss something big.

那可能使你錯失重要的事物。

But we don't have to go down this path.

但我們不用這樣做。

It turns out that the oracle of ancient Greece

到頭來，是古希臘的先知 握有顯示道路的神秘鑰匙。

holds the secret key that shows us the path forward.

Now, recent geological research has shown

近年的地質研究顯示，

that the Temple of Apollo, where the most famous oracle sat,

最有名的先知所在的阿波羅神廟，

was actually built over two earthquake faults.

事實上座落在兩個地震斷層上。

And these faults would release these petrochemical fumes

這些斷層會從地殼下釋出石油煙氣，

from underneath the Earth's crust,

and the oracle literally sat right above these faults,

而那位先知就直接坐在那些斷層上方，

inhaling enormous amounts of ethylene gas, these fissures.

從縫隙中吸入數不盡的乙烯氣體。

(Laughter)

It's true.

（笑聲）

那是真的。

(Laughter)

（笑聲）

It's all true, and that's what made her babble and hallucinate

那都是真的，那就是為什麼 她講話含糊不清還看到幻覺，

and go into this trance-like state.

並進入類似出神的狀態。

She was high as a kite!

她感覺自己都飛上天了！

(Laughter)

（笑聲）

So how did anyone --

所以大家要怎麼──

How did anyone get any useful advice out of her

大家要怎麼在這個狀態下 得到有用的建議？

in this state?

Well, you see those people surrounding the oracle?

看到那些圍繞先知的人們了嗎？

You see those people holding her up,

你可以看到那些人支撐著她，

because she's, like, a little woozy?

因為她好像有點頭昏眼花？

And you see that guy on your left-hand side

有沒有發現她左邊的男子

holding the orange notebook?

正拿著橘色小冊子？

Well, those were the temple guides,

那些是神廟的引導人員，

and they worked hand in hand with the oracle.

他們與先知密切合作。

When inquisitors would come and get on their knees,

當有人來下跪詢問時，

that's when the temple guides would get to work,

神廟的引導人員就開始工作了，

because after they asked her questions,

在來者向先知詢問一些問題後，

they would observe their emotional state,

他們會觀察來者的精神狀態，

and then they would ask them follow-up questions,

然後他們會問來者一些後續問題，

like, "Why do you want to know this prophecy? Who are you?

像是：「為什麼你想知道 這個預言？你是誰？

What are you going to do with this information?"

你會怎麼運用這個資訊？」

And then the temple guides would take this more ethnographic,

接著神廟的引導人員會 用人類學的角度來看，

this more qualitative information,

用質性資訊的角度來看，

and interpret the oracle's babblings.

然後翻譯先知含糊不清的話。

So the oracle didn't stand alone,

所以先知並非自己承攬一切任務，

and neither should our big data systems.

我們的大數據系統同樣也不該如此。

Now to be clear,

我要澄清一下，

I'm not saying that big data systems are huffing ethylene gas,

我並非在說大數據系統 在呼吸着乙烯氣體，

or that they're even giving invalid predictions.

甚至給予沒用的預測。

The total opposite.

完全相反。

But what I am saying

我想說的是，

is that in the same way that the oracle needed her temple guides,

就像先知需要神廟的引導人員那樣，

our big data systems need them, too.

大數據系統同樣也需要。

They need people like ethnographers and user researchers

大數據需要人類學家以及用戶研究人員

who can gather what I call thick data.

來收集我所謂的「厚數據」──

This is precious data from humans,

來自於人們的寶貴數據，

like stories, emotions and interactions that cannot be quantified.

像是故事、情緒和互動， 這些無法計量的事物。

It's the kind of data that I collected for Nokia

就像我收集給諾基亞的那種數據，

that comes in in the form of a very small sample size,

數據樣本規模非常小，

but delivers incredible depth of meaning.

但傳達的涵義卻極其的深。

And what makes it so thick and meaty

它如此厚重、內容豐富的原因是

is the experience of understanding the human narrative.

那些從人們的話語中 明白更多信息的經驗。

And that's what helps to see what's missing in our models.

這才能幫助我們看到 模型裡缺少了什麼東西。

Thick data grounds our business questions in human questions,

厚數據以人類問題為根基 來說明經濟問題，

and that's why integrating big and thick data

這就是為什麼結合大數據和厚數據

forms a more complete picture.

能讓我們得到的訊息更加完整。

Big data is able to offer insights at scale

大數據能在一定程度上洞悉問題，

and leverage the best of machine intelligence,

並最大程度發揮機器智能，

whereas thick data can help us rescue the context loss

而厚數據能幫我們找到 那缺失的背景資訊，

that comes from making big data usable,

能讓大數據便於使用，

and leverage the best of human intelligence.

並最大程度發揮人類智能。

And when you actually integrate the two, that's when things get really fun,

若你真的把這兩個結合在一起 事情就會變得非常有趣，

because then you're no longer just working with data

如此一來，運用的就不只是 你早就收集的數據。

you've already collected.

你還可以運用尚未收集的數據。

You get to also work with data that hasn't been collected.

你就可以知道「為什麼」：

You get to ask questions about why:

Why is this happening?

為什麼會變成這樣？

Now, when Netflix did this,

所以說，網飛這樣做

they unlocked a whole new way to transform their business.

就開啟了轉換商業模式的全新方式。

Netflix is known for their really great recommendation algorithm,

網飛以擁有優秀的推薦演算法而聞名，

and they had this $1 million prize for anyone who could improve it.

且發給任何能改善系統的人 一百萬美元獎金。

And there were winners.

有人贏了獎金。

But Netflix discovered the improvements were only incremental.

但網飛發現效能提升還是不夠明顯。

So to really find out what was going on,

為了知道發生了什麼事，

they hired an ethnographer, Grant McCracken,

他們僱用了人類學家， 格蘭特．麥克拉肯，

to gather thick data insights.

來收集厚數據以準確洞察理解。

And what he discovered was something that they hadn't seen initially

他發現了網飛最初未能 從量化數據中看出來的，

in the quantitative data.

He discovered that people loved to binge-watch.

他發現人們喜歡刷劇。 （註：短時間內狂看電視劇）

In fact, people didn't even feel guilty about it.

事實上，人們甚至不覺得有什麼不對。

They enjoyed it.

他們非常享受這個過程。

(Laughter)

（笑聲）

So Netflix was like, "Oh. This is a new insight."

網飛覺得：「噢，這是個新洞見。」

So they went to their data science team,

於是叫他們的數據科學組

and they were able to scale this big data insight

把這洞察放大到 量化數據的規模來衡量。

in with their quantitative data.

And once they verified it and validated it,

一旦他們再次確認了它的準確性，

Netflix decided to do something very simple but impactful.

網飛便決定做一件簡單 卻影響很大的事情。

They said, instead of offering the same show from different genres

他們說：

「與其提供不同類型但相似的影集，

or more of the different shows from similar users,

或是給類似的觀眾 欣賞更多不同的影集，

we'll just offer more of the same show.

只要同一影集提供更多集就好了。

We'll make it easier for you to binge-watch.

我們讓你更容易刷劇。」

And they didn't stop there.

而他們並沒有止步於此。

They did all these things

他們用一樣的方式，

to redesign their entire viewer experience,

重新設計了整個觀眾體驗，

to really encourage binge-watching.

來真正地鼓勵大家刷劇。

It's why people and friends disappear for whole weekends at a time,

這就是為什麼朋友會消失整個星期，

catching up on shows like "Master of None."

追上「無為大師」等戲劇的進度。

By integrating big data and thick data, they not only improved their business,

結合大數據與厚數據，

不只讓產業進步，

but they transformed how we consume media.

也轉變了我們使用媒體的型態。

And now their stocks are projected to double in the next few years.

預期他們的股票 會在接下來幾年內翻倍。

But this isn't just about watching more videos

這不只是關於看了更多影片，

or selling more smartphones.

或賣了更多智慧型手機，等等。

For some, integrating thick data insights into the algorithm

對於一些公司來說，

結合厚數據洞察和演算法，

could mean life or death,

可能讓他們起死回生，

especially for the marginalized.

特別是那些已被邊緣化的公司。

All around the country, police departments are using big data

全國的警察局都用大數據來防止犯罪，

for predictive policing,

to set bond amounts and sentencing recommendations

來設定保證金金額，

並用加劇偏見的方式來建議判刑。

in ways that reinforce existing biases.

NSA's Skynet machine learning algorithm

美國國家安全局的天網學習演算法

has possibly aided in the deaths of thousands of civilians in Pakistan

可能致使幾千名巴基斯坦平民死亡，

from misreading cellular device metadata.

肇因於錯誤判讀了行動電話的數據。

As all of our lives become more automated,

當我們的生活變得更加自動化，

from automobiles to health insurance or to employment,

從汽車、健康保險或者就業，

it is likely that all of us

很可能我們所有人

will be impacted by the quantification bias.

都會受量化偏見的影響。

Now, the good news is that we've come a long way

好消息是

我們從吸入乙烯氣體到做出預測

from huffing ethylene gas to make predictions.

已有長足的進步。

We have better tools, so let's just use them better.

我們有了更好的工具， 那麽讓我們更好地利用它。

Let's integrate the big data with the thick data.

讓我們將大數據與厚數據結合。

Let's bring our temple guides with the oracles,

讓我們使神廟的引導人員 與先知一起合作，

and whether this work happens in companies or nonprofits

不論做這項工作的是

公司、非營利組織、

or government or even in the software,

政府，甚至軟體，

all of it matters,

全部都有其意義，

because that means we're collectively committed

因為這代表我們全體一起努力

to making better data,

來得到更好的數據，

better algorithms, better outputs

更好的演算法、更好的產品，

and better decisions.

以及更好的決定。

This is how we'll avoid missing that something.

這就是避免錯失的方法。

(Applause)

（掌聲）