The Department of Transport estimated that last year

今天我要談的是科技與社會。

35,000 people died from traffic crashes in the US alone.

交通部估計去年

Worldwide, 1.2 million people die every year in traffic accidents.

在美國有 35,000 人死於交通事故。

If there was a way we could eliminate 90 percent of those accidents,

全球，每年有 120 萬人 死於交通意外。

would you support it?

如果有方法可以避免 90% 的那些意外，

Of course you would.

你們願意支持嗎？

This is what driverless car technology promises to achieve

當然你們會願意。

by eliminating the main source of accidents --

這就是無人駕駛車技術 許諾要達成的目標，

human error.

做法是消除意外事故的主要源頭：

Now picture yourself in a driverless car in the year 2030,

人類錯誤。

sitting back and watching this vintage TEDxCambridge video.

想像一下，2030 年， 你坐在一臺無人駕駛的車內，

(Laughter)

你靠坐著，看著這支極佳的 TEDxCambridge 影片。

All of a sudden,

（笑聲）

the car experiences mechanical failure and is unable to stop.

突然間，

If the car continues,

這臺車發生機械故障，無法停下來。

it will crash into a bunch of pedestrians crossing the street,

如果車繼續前進，

but the car may swerve,

它會撞上過馬路的一群行人，

hitting one bystander,

但可以把車突然轉向，

killing them to save the pedestrians.

只撞死一個旁觀者，

What should the car do, and who should decide?

撞死他就可以救一群行人。

What if instead the car could swerve into a wall,

這臺車會怎麼做？誰來決定？

crashing and killing you, the passenger,

如果改成，這臺車可以 突然轉向去撞牆，

in order to save those pedestrians?

撞毀後只有乘客──也就是你── 會身亡，

This scenario is inspired by the trolley problem,

這樣就能夠救那些行人，如何？

which was invented by philosophers a few decades ago

這個情境的靈感是來自於電車問題，

to think about ethics.

那是幾十年前哲學家所發明的問題，

Now, the way we think about this problem matters.

做道德思考用。

We may for example not think about it at all.

我們思考這個問題的方式 是很重要的。

We may say this scenario is unrealistic,

比如，我們有可能完全不去想它。

incredibly unlikely, or just silly.

我們可能會說，這個情境不實際、

But I think this criticism misses the point

非常不可能發生或很蠢。

because it takes the scenario too literally.

但我認為這種批判是失焦了，

Of course no accident is going to look like this;

因為它看這個情境看得太表面。

no accident has two or three options

當然不會有像這樣子的意外發生；

where everybody dies somehow.

沒有意外會有兩、三個選項，

Instead, the car is going to calculate something

且每個選項都有人死。

like the probability of hitting a certain group of people,

情況是，這臺車去做計算，

if you swerve one direction versus another direction,

比如計算撞到某一群人的機率，

you might slightly increase the risk to passengers or other drivers

如果你向這個方向急轉彎， 跟另一個方向做比較，

versus pedestrians.

你可能會稍微增加乘客 或其他駕駛人的風險，

It's going to be a more complex calculation,

跟行人的風險比較等等。

but it's still going to involve trade-offs,

這會是個比較複雜的計算，

and trade-offs often require ethics.

但它仍然會涉及取捨，

We might say then, "Well, let's not worry about this.

而取捨通常都會需要用到道德。

Let's wait until technology is fully ready and 100 percent safe."

我們可能會說：「別擔心這個。

Suppose that we can indeed eliminate 90 percent of those accidents,

我們可以等到技術完全 準備好且 100% 安全。」

or even 99 percent in the next 10 years.

假設我們的確可以減少 90% 的意外事故，

What if eliminating the last one percent of accidents

或是在接下來十年甚至達到 99%。

requires 50 more years of research?

如果說要把最後 1% 的 意外事故都消除，

Should we not adopt the technology?

會需要再多做 50 年的研究呢？

That's 60 million people dead in car accidents

我們不該採用這個技術嗎？

if we maintain the current rate.

如果我們繼續用目前的方式，

So the point is,

那就會有 6000 萬人死亡。

waiting for full safety is also a choice,

所以重點是，

and it also involves trade-offs.

等到完全安全也是一個選擇，

People online on social media have been coming up with all sorts of ways

而這個選擇也涉及取捨。

to not think about this problem.

人們在網路社會媒體上 提出各式各樣的方式，

One person suggested the car should just swerve somehow

來避免思考這個問題。

in between the passengers --

有個人建議，車應該要急轉彎，

(Laughter)

穿過這些路人──

and the bystander.

（笑聲）

Of course if that's what the car can do, that's what the car should do.

和旁觀者之間。

We're interested in scenarios in which this is not possible.

當然，如果這臺車能這麼做， 那它是應該這麼做。

And my personal favorite was a suggestion by a blogger

我們感興趣的是 沒辦法這樣做的情況。

to have an eject button in the car that you press --

我個人的最愛， 是一個部落客的建議，

(Laughter)

他建議車要裝設一個 彈射按鈕，讓你可以──

just before the car self-destructs.

（笑聲）

(Laughter)

在車自毀之前按下它。

So if we acknowledge that cars will have to make trade-offs on the road,

（笑聲）

how do we think about those trade-offs,

所以，如果我們承認車在 路上行駛時會需要做取捨，

and how do we decide?

我們要如何去想那些取捨？

Well, maybe we should run a survey to find out what society wants,

我們要如何做決定？

because ultimately,

也許我們該做個調查 來了解社會想要什麼，

regulations and the law are a reflection of societal values.

因為，最終，

So this is what we did.

規定和法律都是反應出社會價值。

With my collaborators,

所以我們就這麼做了。

Jean-François Bonnefon and Azim Shariff,

我和我的共同研究者，

we ran a survey

尚方斯華．邦納方和阿辛．夏利夫，

in which we presented people with these types of scenarios.

一起進行了一項調查，

We gave them two options inspired by two philosophers:

調查中，我們給人們看這些情境。

Jeremy Bentham and Immanuel Kant.

我們給予他們兩個選項， 選項靈感來自兩位哲學家：

Bentham says the car should follow utilitarian ethics:

傑瑞米．邊沁及伊曼努爾．康德。

it should take the action that will minimize total harm --

邊沁說，車應該要遵循特定的道德：

even if that action will kill a bystander

它應該要採取 能讓傷害最小的行為──

and even if that action will kill the passenger.

即使那個行為會害死一名旁觀者，

Immanuel Kant says the car should follow duty-bound principles,

即使那個行為會害死一名乘客。

like "Thou shalt not kill."

康德說，車應該要遵循 責無旁貸原則，

So you should not take an action that explicitly harms a human being,

比如「你不該殺人」。

and you should let the car take its course

所以你不能採取很明確 會傷害到人類的行為，

even if that's going to harm more people.

你應該讓車就照它的路去走，

What do you think?

即使那最後會傷害到更多人。

Bentham or Kant?

你們認為如何？

Here's what we found.

邊沁或康德？

Most people sided with Bentham.

我們發現的結果如下。

So it seems that people want cars to be utilitarian,

大部分人站在邊沁這一邊。

minimize total harm,

所以，似乎人們會希望 車是功利主義的，

and that's what we should all do.

把總傷害降到最低，

Problem solved.

我們所有人都該這麼做。

But there is a little catch.

問題解決。

When we asked people whether they would purchase such cars,

但有一個小難處。

they said, "Absolutely not."

當我們問人們， 他們是否會買這樣的車，

(Laughter)

他們說：「絕對不會。」

They would like to buy cars that protect them at all costs,

（笑聲）

but they want everybody else to buy cars that minimize harm.

他們會買不計代價來保護他們的車，

(Laughter)

但他們希望其他人都買 能把傷害減到最低的車。

We've seen this problem before.

（笑聲）

It's called a social dilemma.

我們以前就看過這個問題。

And to understand the social dilemma,

它叫做社會兩難。

we have to go a little bit back in history.

要了解社會兩難，

In the 1800s,

我們得要回溯一下歷史。

English economist William Forster Lloyd published a pamphlet

在 1800 年代，

which describes the following scenario.

英國經濟學家威廉．佛斯特．洛伊 出版了一本小冊子，

You have a group of farmers --

小冊子描述了下列情境：

English farmers --

有一群農夫

who are sharing a common land for their sheep to graze.

──英國農夫──

Now, if each farmer brings a certain number of sheep --

他們在一塊共用地上面 放牧他們的羊。

let's say three sheep --

每個農夫都會帶來一定數量的羊──

the land will be rejuvenated,

比如說三隻羊──

the farmers are happy,

這塊地會更生，

the sheep are happy,

農夫都很快樂，

everything is good.

羊也很快樂，

Now, if one farmer brings one extra sheep,

一切都很好。

that farmer will do slightly better, and no one else will be harmed.

現在，如果一個農夫多帶了一隻羊，

But if every farmer made that individually rational decision,

那位農夫的生計會更好一些， 其他人也沒受害。

the land will be overrun, and it will be depleted

但如果每位農夫都各別 做出了那個理性的決定，

to the detriment of all the farmers,

就會超過這塊地的限度， 資源會被用盡，

and of course, to the detriment of the sheep.

所有農夫都會受害，

We see this problem in many places:

當然，羊也會受害。

in the difficulty of managing overfishing,

我們在許多地方都會看到這類問題：

or in reducing carbon emissions to mitigate climate change.

比如管理過度捕魚的困難，

When it comes to the regulation of driverless cars,

或是減少碳排放來緩和氣候變遷。

the common land now is basically public safety --

當用到規範無人駕駛車的情況時，

that's the common good --

這塊共用地基本上就是公共安全──

and the farmers are the passengers

也就是共善──

or the car owners who are choosing to ride in those cars.

農夫就是乘客，

And by making the individually rational choice

或是選擇坐這種車的車主。

of prioritizing their own safety,

他們各自做出理性的選擇，

they may collectively be diminishing the common good,

把他們自己的安全擺在第一，

which is minimizing total harm.

那麼整體來說， 他們就可能會削減了共善，

It's called the tragedy of the commons,

共善就是把總傷害減到最低。

traditionally,

這就是所謂的「公地悲劇」，

but I think in the case of driverless cars,

傳統是這麼稱呼的。

the problem may be a little bit more insidious

但我認為在無人駕駛車的情況中，

because there is not necessarily an individual human being

問題可能還有涉及其他隱患，

making those decisions.

因為並沒有特別一個人類

So car manufacturers may simply program cars

在做那些決定。

that will maximize safety for their clients,

汽車製造商很可能就會把他們的車

and those cars may learn automatically on their own

設計成以確保客戶安全為第一要務，

that doing so requires slightly increasing risk for pedestrians.

而那些車可能會自己發現，

So to use the sheep metaphor,

保護客戶會稍微增加行人的風險。

it's like we now have electric sheep that have a mind of their own.

若用放牧羊的比喻，

(Laughter)

就像是有電子羊，自己會思考。

And they may go and graze even if the farmer doesn't know it.

（笑聲）

So this is what we may call the tragedy of the algorithmic commons,

他們可能會在農夫不知道的 情況下自己去吃草。

and if offers new types of challenges.

所以我們可以稱之為 「演算法公地悲劇」，

Typically, traditionally,

它會帶來新的挑戰。

we solve these types of social dilemmas using regulation,

通常，傳統上，

so either governments or communities get together,

我們會用規制來解決這些社會兩難，

and they decide collectively what kind of outcome they want

所以政府或是社區會集合起來，

and what sort of constraints on individual behavior

他們共同決定想要什麼樣的結果、

they need to implement.

以及對於個人行為，他們需要實施

And then using monitoring and enforcement,

什麼樣的限制。

they can make sure that the public good is preserved.

接著透過監控和執行，

So why don't we just,

他們就能確保公善能被維護。

as regulators,

身為管理者，

require that all cars minimize harm?

我們為什麼不

After all, this is what people say they want.

要求所有的車都要把 傷害降至最低就好了？

And more importantly,

畢竟，這是人民要的。

I can be sure that as an individual,

更重要的，

if I buy a car that may sacrifice me in a very rare case,

我可以確定，身為個人，

I'm not the only sucker doing that

如果我買了一臺有可能在 非常少數情況下把我犧牲的車，

while everybody else enjoys unconditional protection.

至少我不會是唯一的冤大頭，

In our survey, we did ask people whether they would support regulation

因為人人都享有無條件保護。

and here's what we found.

在我們的調查中， 我們有問人們是否支持規制，

First of all, people said no to regulation;

這是我們的發現。

and second, they said,

首先，人們對規制說不；

"Well if you regulate cars to do this and to minimize total harm,

第二，他們說：

I will not buy those cars."

「如果你們規定車都要這樣做， 並把總傷害降至最低，

So ironically,

我不會去買那些車。」

by regulating cars to minimize harm,

所以，很諷刺，

we may actually end up with more harm

藉由規範汽車來將傷害降至最小，

because people may not opt into the safer technology

結果卻會是更多的傷害，

even if it's much safer than human drivers.

因為人們可能不會選擇 比較安全的科技，

I don't have the final answer to this riddle,

即使科技比人類駕駛還安全。

but I think as a starting point,

關於這個謎，我沒有最終的答案，

we need society to come together

但我認為，做為一個起始點，

to decide what trade-offs we are comfortable with

我們需要社會能團結起來，

and to come up with ways in which we can enforce those trade-offs.

來決定我們對何種取捨 會覺得比較舒服，

As a starting point, my brilliant students,

並提出方法來執行這些取捨。

Edmond Awad and Sohan Dsouza,

做為一個起始點，我的出色學生們，

built the Moral Machine website,

艾德蒙．艾瓦和索漢．達蘇札，

which generates random scenarios at you --

建立了「道德機器」網站，

basically a bunch of random dilemmas in a sequence

它會為你產生隨機的情境，

where you have to choose what the car should do in a given scenario.

基本上是一連串的隨機兩難，

And we vary the ages and even the species of the different victims.

你得要選擇在給定的情境中， 車要怎麼做。

So far we've collected over five million decisions

我們會改變受害者的年齡和物種。

by over one million people worldwide

目前我們已經從 來自全世界超過 100 萬人，

from the website.

收集到了超過 500 萬個決定，

And this is helping us form an early picture

透過網站取得的。

of what trade-offs people are comfortable with

這能協助我們勾勒出初步的狀況，

and what matters to them --

了解人們對怎樣的取捨 感到比較舒服、

even across cultures.

什麼對他們比較重要──

But more importantly,

且是跨文化的。

doing this exercise is helping people recognize

但更重要的，

the difficulty of making those choices

做這個演練，能協助人們認清

and that the regulators are tasked with impossible choices.

做那些決定有多困難、

And maybe this will help us as a society understand the kinds of trade-offs

而管理者的工作就是要 做不可能的選擇。

that will be implemented ultimately in regulation.

也許這能協助我們這個社會 去了解這類的取捨，

And indeed, I was very happy to hear

終究這類的取捨在將來也會 被納入到規制當中。

that the first set of regulations

的確，我非常高興聽到

that came from the Department of Transport --

來自交通部的

announced last week --

第一組規定

included a 15-point checklist for all carmakers to provide,

在上週宣佈，

and number 14 was ethical consideration --

內容有一張清單，上面有 15 個項目， 是要汽車製造商避免的，

how are you going to deal with that.

而第 14 項就是道德考量──

We also have people reflect on their own decisions

你要如何去處理它。

by giving them summaries of what they chose.

也有人會反思他們自己的決定，

I'll give you one example --

我們會把人們所做的決定 歸納給他們。

I'm just going to warn you that this is not your typical example,

我舉一個例子給各位──

your typical user.

我要先警告各位， 這不是典型的例子，

This is the most sacrificed and the most saved character for this person.

使用者也不是典型的。

(Laughter)

這個人最會犧牲的角色（小孩） 及最會拯救的角色（貓）是這些。

Some of you may agree with him,

（笑聲）

or her, we don't know.

有些人可能認同他，

But this person also seems to slightly prefer passengers over pedestrians

或她，性別不詳。

in their choices

但這個人對於乘客的偏好 也略高於行人，

and is very happy to punish jaywalking.

從他們的選擇看得出來，

(Laughter)

且非常樂意懲罰闖紅燈的人。

So let's wrap up.

（笑聲）

We started with the question -- let's call it the ethical dilemma --

來總結收尾一下。

of what the car should do in a specific scenario:

我們一開頭問的問題， 就姑且稱之為道德兩難，

swerve or stay?

在特定的情境中， 車該怎麼做：

But then we realized that the problem was a different one.

急轉彎或保持原路？

It was the problem of how to get society to agree on and enforce

但接著我們了解到， 這個問題有所不同。

the trade-offs they're comfortable with.

問題是要如何讓社會認同，

It's a social dilemma.

並且執行讓他們覺得舒服的取捨。

In the 1940s, Isaac Asimov wrote his famous laws of robotics --

這是個社會兩難。

the three laws of robotics.

在四〇年代，以撒．艾西莫夫 寫了著名的機器人學法則──

A robot may not harm a human being,

機器人學的三大法則。

a robot may not disobey a human being,

機器人不得傷害人類，

and a robot may not allow itself to come to harm --

機器人不得違背人類，