Can AI predict someone's breakup? - Thomas Hofweber

翻译人员: Kate Tang 校对人员: Sue Lu

You and your partner Alex have been in a strong, loving relationship for years,

你和你的伴侣亚历克斯 多年来一直保持着牢固的恋爱关系，

and lately you're considering getting engaged.

而且最近你正在考虑订婚。

Alex is enthusiastic about the idea, but you can’t get over the statistics.

亚历克斯对此兴高采烈， 但你无法克服对统计数据的不安。

You know a lot of marriages end in divorce, often not amicably.

你知道很多婚姻都以离婚告终， 而且往往不是和平的分手。

And over 10% of couples in their first marriage get divorced

超过10％的初婚夫妇

within the first five years.

在刚结婚的五年内离婚。

If your marriage wouldn’t even last five years,

如果你的婚姻连五年都维持不了，

you feel like tying the knot would be a mistake.

那喜结连理就是个错误。

But you live in the near future,

设想一下，你生活在不久的将来，

where a brand-new company just released an AI-based model

一家新公司刚刚发布了 一种基于人工智能（AI）的模型，

that can predict your likelihood of divorce.

可以预测你离婚的可能性。

The model is trained on data sets

该模型使用数据集进行训练，

containing individuals’ social media activity,

数据包括个人社交媒体活动、

online search histories, spending habits, and history of marriage and divorce.

在线搜索历史、消费习惯 以及结婚和离婚史。

And using this information,

利用这些信息，

the AI can predict if a couple will divorce

AI 可预测一对夫妇是否在婚后的

within the first five years of marriage with 95% accuracy.

头五年内离婚，准确率为 95 ％。

The only catch is the model doesn’t offer any reasons for its results—

问题是 AI 不为其结论提供原因,

it simply predicts that you will or won’t divorce without saying why.

它只是预测你会否离婚而已。

So, should you decide whether or not to get married

那么，你应该依据 AI 的预测

based on this AI’s prediction?

来决定是否结婚吗？

Suppose the model predicts you and Alex

假设模型预测你和亚历克斯

would divorce within five years of getting married.

将在结婚后的五年内离婚。

At this point, you'd have three options.

此时，你有三个选择。

You could get married anyway and hope the prediction is wrong.

你可以结婚并指望预测是错的。

You could break up now,

你可以现在分手，

though there’s no way to know if ending your currently happy relationship

尽管没法知道结束目前幸福的关系

would cause more harm than letting the prediction run its course.

是否会比顺其自然造成更大的伤害。

Or, you could stay together and remain unmarried,

或者，你们可以在一起但保持未婚状态，

on the off-chance marriage itself would be the problem.

以防万一婚姻本身才是问题所在。

Though without understanding the reasons for your predicted divorce,

不了解为什么被预测离婚，

you’d never know if those mystery issues would still emerge

你就永远不知道 那些神秘的破坏因素是否还会出现，

to ruin your relationship.

从而导致你的关系破裂。

The uncertainty undermining all these options

这种不确定性会危及以上所有选择，

stems from a well known issue with AI around explainability and transparency.

它源于 AI 在可解释性和透明度方面 一个众所周知的问题。

This problem plagues tons of potentially useful predictive models,

这个问题困扰着 大量很有潜力的预测模型，

such as those that could be used to predict which bank customers

本来可以用来预测，例如，哪些银行客户

are most likely to repay a loan,

最有可能偿还贷款；

or which prisoners are most likely to reoffend if granted parole.

或者哪些囚犯假释后最有可能再犯。

Without knowing why AI systems reach their decisions,

因为不清楚 AI 系统如何做出决定，

many worry we can’t think critically about how to follow their advice.

许多人担心我们会 无法判断如何遵循其建议。

But the transparency problem doesn’t just prevent us

但是，透明度问题不仅阻碍我们

from understanding these models,

理解这些模型，

it also impacts the user’s accountability.

它还会影响用户对责任的担当。

For example, if the AI's prediction led you to break up with Alex,

例如，如果 AI 的预测 导致你与亚历克斯分手，

what explanation could you reasonably offer them?

你能合理地为他提供什么解释？

That you want to end your happy relationship

你会说你想结束这段幸福的恋情，

because some mysterious machine predicted its demise?

因为某个神秘的机器预言了它的消亡？

That hardly seems fair to Alex.

对亚历克斯来说，这似乎不公平。

We don’t always owe people an explanation for our actions,

我们并不总是需要解释我们的行为，

but when we do,

但是当我们必须解释时，

AI’s lack of transparency can create ethically challenging situations.

AI 缺乏的透明度会造成 挑战道德底线的情况。

And accountability is just one of the tradeoffs we make

责任担当只是我们 将重要决定外包给 AI 时

by outsourcing important decisions to AI.

所做出的权衡之一。

If you’re comfortable deferring your agency to an AI model

如果你乐意将决定权让给 AI 模型，

it’s likely because you’re focused on the accuracy of the prediction.

那很可能是因为你专注于预测的准确性。

In this mindset, it doesn’t really matter why you and Alex might break up—

在这种心态下，你和亚历克斯 分手的原因并不重要 --

simply that you likely will.

因为高分手几率才是罪魁祸首。

But if you prioritize authenticity over accuracy,

但是，如果你把真实性 看得比准确性更重要，

then you'll need to understand and appreciate the reasons

那么你需要先充分了解

for your future divorce before ending things today.

未来离婚的原因，然后再结束现在的关系。

Authentic decision making like this is essential for maintaining accountability,

这种真实的决定 对于维持责任担当至关重要，

and it might be your best chance to prove the prediction wrong.

而且这可能是证明预测错误的良机。

On the other hand, it’s also possible the model

另一方面，该模型也可能

already accounted for your attempts to defy it,

设置了相应程序来对付你的违抗，

and you’re just setting yourself up for failure.

所以你可能只是在为自己的失败设局。

95% accuracy is high, but it’s not perfect—

95% 的准确率很高，但并不完美——

that figure means 1 in 20 couples will receive a false prediction.

意味着每 20 对夫妇中 就有 1 对被误测。

And as more people use this service,

随着越来越多的人使用这项服务，

the likelihood increases that someone who was predicted to divorce

越来越多被预测离婚的人

will do so just because the AI predicted they would.

将为了 AI 的预测而去结束婚姻。

If that happens to enough newlyweds,

如果很多新婚夫妇都这么做，

the AI's success rate could be artificially maintained

这些自我实现的预测就可以人为地维持

or even increased by these self-fulfilling predictions.

甚至提高 AI 的成功预测率。

Of course, no matter what the AI might tell you,

当然，不管 AI 告诉你什么，

whether you even ask for its prediction is still up to you.

甚至你是否求助于它的预测， 一切都取决于你。