“The machine does not isolate man from the great problems of nature but plunges him more deeply into them.”
“这台机器并不能将人与自然界的重大问题隔离开来，而是使他更深入地陷入自然界。”
Antoine de Saint-Exupéry
圣安修伯里
Let me begin by saying i’m a technologist at heart, through my career I have coded in a multitude of programming languages and been enthralled to the possibility of what technology can do, and the help and hope it can bring.
首先，我要说我是一名内心的技术专家，在我的职业生涯中，我已经使用多种编程语言进行编码，并且对技术可以做什么以及它带来的帮助和希望深感兴趣。
I have seen first hand how it broadens the world for all, done right technology is incredible, a leveller to access, removing barriers for people with disabilities and creating a marketplace of ideas and opportunity for everyone regardless of cultural background, wealth or location.
我亲眼目睹了它如何扩大所有人的视野，正确的技术是令人难以置信的，可调节的通道，消除了残疾人的障碍并为所有人创造了一个思想和机会的市场，无论其文化背景，财富或地理位置如何。
Code itself can read as beautifully as poetry, or create images in the mind like the most descriptive of prose. A great programmer can explain ideas with a simplistic brilliance that George Orwell — that great advocate of plain English — would admire. And can be a medium for artists that allows them to create new fields of work, pushing the boundaries and understanding of humanity.
代码本身可以像诗歌一样精美地阅读，或者在大脑中像最具描述性的散文一样创建图像。 一位出色的程序员可以用简单的才华来解释想法，而乔治·奥威尔(朴素的英语的伟大拥护者)会钦佩的。 可以作为艺术家的媒介，使他们能够创造新的工作领域，突破界限并理解人类。
“models encoded human prejudice, misunderstanding, and bias into the software systems that increasingly manage our lives… they tended to punish the poor and the oppressed in society”
“模型将人类的偏见，误解和偏见编码到越来越多地管理我们生活的软件系统中……它们倾向于惩罚社会中的穷人和被压迫者”
Cathy O’Neil, Weapons of math destruction
凯茜·奥尼尔(Cathy O'Neil)，数学破坏武器
Sadly though in recent times this craft, this vitality, openness and humanity has been lost. We have seen the rise of the algorithm yet forgotten the ethics that should underpin it.
可悲的是，尽管最近这种手艺失去了这种活力，开放性和人性。 我们已经看到了算法的兴起，但却忘记了应作为其基础的伦理。
The recent scandal in the United Kingdom about the use of AI to give students their A-Level results is a prime example of this overconfidence in the algorithm. Many assume that algorithms and data, and therefore AI are unbiased.
最近在英国发生的有关使用AI为学生提供A级成绩的丑闻就是这种对算法过度自信的典型例子。 许多人认为算法和数据以及AI是无偏见的。
How is it, they ask, that numbers could be biased? Numbers are raw data, they judge no one, instead they tell a history of what’s been captured.
他们问，数字可能会有偏差吗？ 数字是原始数据，他们判断没有人，而是告诉他们所捕获内容的历史。
Yet as with the A-Level scandal the devil lies in the detail. Data sets often contain the bias of those that set them up, or the bias from where they were collected. Take for example facial recognition, almost every set of facial recognition algorithms is trained on a data set that considers gender to be a defining factor.
然而，与A级丑闻一样，细节在于魔鬼。 数据集通常包含建立数据集的偏见或从收集它们时的偏见。 以面部识别为例，几乎每组面部识别算法都在将性别视为决定性因素的数据集上进行训练。
Let’s consider a situation in the not too distant future where toilet facilities are unlocked by facial recognition, and the system is trained on a model that recognises only male and female faces. A person who is transgender, non-binary, or gender non-conforming may well be locked out of using these spaces. We have, in that case, designed them out of society.
让我们考虑一下在不久的将来出现的情况，即通过面部识别来解锁厕所设施，并且该系统在仅识别男性和女性面部的模型上进行训练。 跨性别，非二元性或性别不符合的人很可能被禁止使用这些空间。 在这种情况下，我们是出于社会目的设计它们的。
If the data set that was used to train the system contained either only the source data of two genders, male and female, or the person who created the algorithm themselves held a natural bias or prejudice that only two genders existed, then the system itself is biased from the outset.
如果用于训练系统的数据集仅包含男性和女性两种性别的源数据，或者创建算法的人本身存在自然偏见或偏见，则仅存在两种性别，那么系统本身就是从一开始就有偏见。
“It is always the image someone chose, to photograph is to frame and to frame is to exclude.”
“这始终是某人选择的图像，照相是要构图，而构图是要排除。”
Susan Sontag, Regarding the pain of others
苏珊·桑塔格(Susan Sontag)，关于他人的痛苦
Likewise we can see this algorithmic bias in other areas too, for example socio-economic or race. With the A-Level scandal, one of the factors in ranking a pupil’s result was how well the institution had performed historically.
同样，我们也可以在其他方面看到这种算法偏差，例如社会经济或种族。 对于A级丑闻，对学生成绩进行排名的因素之一是该机构在历史上的表现。
On paper this system would make sense, take the results of students’ predicted grades, the result the teacher ranked them at and look at the historical likelihood of those grades being correct based on the past grades of the school or institution.
从表面上看，该系统很有意义，它可以将学生的预期成绩作为结果，对教师进行排名后的结果进行排名，并根据学校或机构过去的成绩来查看这些成绩正确的历史可能性。
Yet unfortunately this logic doesn’t take into account the inherent socio economic and class bias that exists in the system. Exceptional students, first in a family tree to university, were downgraded due to past results of the state school they studied at, whilst students at private schools from wealthy backgrounds where parents could afford private tutors and class sizes were smaller had their grade scores bumped, proving that algorithms can, and do, contain class bias.
但是，不幸的是，这种逻辑没有考虑到系统中固有的社会经济和阶级偏见。 成绩优异的学生，首先是进入大学的家谱，由于其就读的公立学校的过往成绩而被降级；而来自富裕背景的私立学校的学生则可以负担得起家教的费用，父母的学费却有所下降，班级规模变小，证明算法可以并且确实包含类偏差。
This algorithmic bias is a topic well covered by Cathy O’Neil in her book Weapons of Math Destruction (Penguin books, 2016), one of her explanation being that the underlying data models have bias built into the model in such a form that it is almost impossible for those affected to challenge it.
这个算法偏差是Cathy O'Neil在其《数学毁灭性武器》 (企鹅书，2016年)一书中很好地论述的一个话题，她的解释之一是，底层数据模型在模型中内置了偏差，其形式是对于那些受影响的人来说几乎是不可能的。
Moreover this bias is becoming more important as our reliance — or more to the point over reliance — on data becomes baked into our day to day lives.
此外，随着我们对数据的依赖(或更确切地说是对依赖的依赖)逐渐渗入我们的日常生活中，这种偏见变得越来越重要。
The workplace in all its guises is an area where we are starting to view this overreach of data and algorithms, Amazon, a company whose business model is built around collection and usage of data set about to “solve” the hiring problem in 2014, having realised that the manual task of sorting CV’s for interviews was a costly exercise.
在所有形式的工作场所中，我们都开始看到这种数据和算法的超范围。亚马逊是一家业务模型围绕收集和使用数据集建立的公司，旨在“解决” 2014年的招聘问题，该公司拥有意识到将简历分类以进行面试的手动任务是一项昂贵的工作。
On paper, and looking at the bottom line only, a problem like CV filtering is the “perfect” problem to solve with AI, but a year into the project a problem emerged, the algorithm that was being used was downgrading candidates for software and other technical roles if they were female.
从表面上看，仅看底线，像CV过滤这样的问题是AI可以解决的“完美”问题，但是进入该项目的一年出现了一个问题，所使用的算法正在降低软件和其他软件的候选等级技术角色(如果是女性)。
The reason behind this was simple, the data that the algorithm itself was being trained on was biased, since the majority of CV’s were male, and the majority of hires were male the system predicted that the perfect candidate would be male. This meant the algorithm downgraded candidates that had been to women’s colleges for example.
这背后的原因很简单，由于大多数CV是男性，而大多数雇员是男性，因此该算法本身接受过培训的数据存在偏见，系统预测理想的候选人将是男性。 例如，这意味着该算法将曾经去过女子大学的候选人降级了。
Amazon was forced to abandon the project, in essence they had tried to “solve” a problem that needed human intervention with a piece of code, the assumption being that it would be both logical and free from bias.
亚马逊被迫放弃该项目，从本质上讲，他们试图用一段代码来“解决”一个需要人为干预的问题，但前提是这既合乎逻辑又没有偏见。
However what they failed to take into account was if the data set it was trained on contained the natural biases of those that collated it, or if the system — or historical conditions — from which it was drawn already contained human bias.
但是，他们没有考虑到的是，如果对其进行训练的数据集是否包含进行整理的数据的自然偏差，或者从中提取数据的系统(或历史条件)是否已经包含了人为偏差。
As we see more and more decisions being made in the workplace by algorithms, data and AI then it would feel natural a new rise in digital Luddism will emerge. We are refining people down to past histories of what they’ve done, their data.
随着我们看到越来越多的算法，数据和AI在工作场所做出决策，那么自然就会出现数字Luddism的新增长。 我们正在将人们的工作细化到过去的历史，他们的数据。
Judging or ranking them by algorithms fraught with the historical biases of industries, class systems; race; gender; sexuality, and removing the subtle nuance of human intervention, flawed as it is, in favour of a machine led industrialisation of all of our workplaces.
用充满行业，阶级制度的历史偏见的算法对它们进行判断或排序； 种族; 性别; 性，并消除了人为干预的细微差别，这是有缺陷的，有利于我们所有工作场所的机械化工业化。
No longer is it the dark satanic mills that threaten our crafts and livelihoods, instead giant factories, or more to the point warehouses, of data determining our lives based on pre-selected and pre-determined ideas of what we should look like.
黑暗的撒旦工厂不再威胁我们的手Craft.io和生计，反而威胁着大型工厂，甚至更多地威胁着仓库，这些数据根据预先选择的和预先确定的关于我们的样子的想法来决定我们的生活。
So is that it? Are we doomed? Our agency removed, our ability and creativity stifled, our potential trapped in a past we had no choice in creating and no means to change. Well yes. And no.
就是这样吗？ 我们注定要失败吗？ 我们的代理人搬走了，我们的能力和创造力被扼杀了，我们被困在过去的潜力中，我们别无选择，没有创造力，也没有改变的手段。 嗯，是。 和不。
The good news is that it doesn’t have to be so biased. The solutions already exist to avoid this and it’s up to us to adopt them. Data sets co-created with under-represented groups are slowly becoming adopted. Google’s Project Euphonia, as an example, uses a data set trained on people with speech impediments and other similar disabilities so as to make their voice assistants more accessible and understandable.
好消息是，它不必如此偏颇。 解决方案已经存在，可以避免这种情况，这取决于我们采用这些解决方案。 与代表性不足的群体共同创建的数据集正在逐渐被采用。 以Google的Euphonia项目为例，该项目使用了针对有语言障碍和其他类似残疾的人进行训练的数据集，以使他们的语音助手更易于访问和理解。
There are also projects that are investing time and forming communities from which new and more diverse datasets can emerge to train models, such as Queer In AI and Black in AI. Alongside this we can add in governance to ensure our systems consider the impact and damage that these systems can bring.
也有一些项目正在投入时间并形成社区，从中可以涌现出新的，更多样的数据集来训练模型，例如Queer In AI和Black in AI 。 除此之外，我们还可以添加治理以确保我们的系统考虑这些系统可能带来的影响和损害。
Participatory design for AI and Machine Learning is a design methodology that allows for those individuals affected to challenge and contest decisions made on their behalf by code by being involved at the outset, and is becoming a fast emerging sector and tools like IBM’s AI Fairness tool are now emerging on the market.
人工智能和机器学习的参与式设计是一种设计方法，它允许受影响的人们一开始就参与其中，以代码的形式挑战和挑战代表他们做出的决策，并且正在成为一个快速兴起的行业，并且诸如IBM的AI Fairness工具之类的工具正在Swift发展。现在新兴市场。
By challenging those that build the systems to consider the social sciences problems alongside the maths and programming problems and placing empathy at the heart of the system we can avoid exclusion. In fact Gartner have even recently included Responsible AI in their hype cycle as a trigger to innovation.
通过挑战那些构建系统的人员来考虑社会科学问题以及数学和编程问题，并将同情心置于系统的核心，我们可以避免排斥。 实际上，Gartner最近甚至在其炒作周期中加入了负责任的AI，以激发创新。
Let us be honest here AI and the reliance on data is here to stay. It’s an easy argument to make, looks good on spreadsheets and can be proven to make the factory line — physical or intellectual — more efficient (though better? That’s yet to be answered). But with small changes, more inclusion and more participation, we can make it represent more of us than we have before.
坦白地说，人工智能和对数据的依赖将一直存在。 这很容易做出来，在电子表格上看起来不错，并且可以证明它可以使工厂生产线(无论是物理还是智能)更加高效(尽管更好？这尚待解决)。 但是，通过较小的更改，更多的包容性和更多的参与，我们就可以使它比以往拥有更多的人。
Because ultimately with every CV rejected because of someone’s gender, every downgraded mark because of someone’s social status, every poorly chosen career path or allocation of work based on someone’s previous history we aren’t just failing ourselves ethically, we’re alienating potential customers, demotivating our staff and/or damaging our brand reputation.
因为最终每个简历都因为某人的性别而被拒绝，每个降级的分数都是由于某人的社会地位，每一个选择不当的职业道路或基于某人的先前历史而分配的工作，我们不仅仅是在道德上让自己失败，我们正在疏远潜在客户，激励我们的员工和/或损害我们的品牌声誉。
Socially responsible, inclusive, design may just stop us from a new rise of digital Luddites, though on the downside we may also be robbed of a digital Byron.
具有社会责任感，包容性的设计可能只会阻止我们摆脱数字化Luddites的兴起，尽管不利的一面是，我们也可能会抢走数字拜伦。
As the liberty lads o’er the sea
当自由小伙子在海上
Bought their freedom, and cheaply, with blood,
用鲜血廉价地买了他们的自由，
So we, boys, we
所以我们，男孩，我们
Will die fighting, or live free,
会死于战斗或自由生活，
And down with all kings but King Ludd!”
到了除了路德国王以外的所有国王！”
When the web that we weave is complete,
当编织的网完成后，
And the shuttle exchanged for the sword,
航天飞机换了剑，
We will fling the winding sheet
我们将把绕组片扔掉
O’er the despot at our feet,
在我们脚下的霸主
And dye it deep in the gore he has poured.”
并把它染在他倒出的血腥深处。”
Though black as his heart its hue,
虽然他的心是黑色的，
Since his veins are corrupted to mud,
既然他的血管变得泥泞了，
Yet this is the dew
但这是露水
Which the tree shall renew
哪棵树要更新
Of Liberty, planted by Ludd!
自由女神，由路德种下！
Lord Byron
拜伦勋爵
翻译自: https://medium.com/designing-with-empathy/on-becoming-a-modern-day-luddite-3c2a03599949
相关资源：四史答题软件安装包exe