【Background】
Building Mind-Machine Combinations: Move Past Standard Partnerships, with Andrew McAfee, Associate Director, MIT Center for Digital Business and Author, Machine, Platform, Crowd
The “standard partnership” between mind and machine is terminology Andrew McAfee and Erik Brynjolfsson use to describe the status quo in most companies, as of 2017. Machines are “essentially glorified spreadsheets”, as McAfee puts it. They store, sort, and send data. Humans are the “big brains” of most operations, looking for patterns and solutions in these data sets and sifting through ideas to find the good ones. The trouble is that even with training and experience, our intuition is flawed. So many forward-thinking companies are restructuring the partnership, finding better ways to use human intuition and machine thinking in conjunction.
今天我们请到了安德鲁·麦卡菲。他是麻省理工学院数字商务中心的副主任,同时也是《机器、平台、群众》一书的作者。在今天的课程中,他将从人和机器、管理者的角度为我们讲解什么是“常规合作关系”,以及如何超越“常规合作关系”。
“常规合作关系”这一术语由安德鲁•麦卡菲和埃里克•布林约尔松提出,用以形容至2017年大多数公司所呈现出的人机关系现状。正如麦卡菲说,机器本质上是“美化过的电子表格”。机器负责存储、整理和传输数据,而人类常常是操作数据的“超级大脑”,人类从数据集中寻求有益模式和解决办法,在一堆想法中挑选好主意。但问题是,就算我们受过训练,经验丰富,人类的直觉也存在缺陷。因此许多有远见的公司正调整人机关系,以期更好地结合人类直觉和机器思维。
【Course】
Rethink the balance between minds and machines 重新思考人和机器之间的平衡
In our book, Machine, Platform, Crowd , Eric Brynjolfsson and I use this phrase the “standard partnership” over and over again to talk about the status quo in organizations, and the status quo in two ways. First of all, the status quo between minds and machines. And the standard partnership right now is the machines do all the record keeping, the transaction processing, they zip the information around the world. They’re glorified spreadsheets in a lot of ways. And then the job of the people, the jobs of the minds in the organization, is essentially to look at all that data that’s generated but then apply their good old-fashioned human judgment and intuition and experience on top of that. And there’s just way too much evidence about how buggy and glitchy and biased our human decision-making processes are to let that standard partnership endure. So the companies that we researched and that I came to respect a lot when we were writing the book are trying very hard to move past that standard partnership. And they’re trying not to get human minds totally out of the loop, but if it turns out that the algorithms, and the data, and the evidence are better and are making better decisions, we need to respect that and we need not to try to slap human judgment on top of it just because people feel good about making decisions. So the example that I like to use is, let’s say it turns out that computers become better at medical diagnostics than even highly trained human doctors. In that case, do we want to continue to let human doctors make the diagnosis just so they get to keep feeling good about themselves? I would say absolutely not if we want a healthier population. And if it turns out, like I think it will, that especially in an era of really, really powerful machine learning that the machines are better at that task than the minds, we really need to give a lot more of that work over to the machines.
在《机器、平台、群众》一书中,我和埃里克·布林约尔松用“常规合作关系”一词,从两个方面多次讨论了企业的处境。一是从人机关系的现状来看,现在的常规合作是指机器负责记录数据,处理事务,汇总世界各地的信息。从很多方面来看,机器就是“美化过的电子表格”。而在企业,人的工作,即人脑的任务,其实就是查看所有生成的数据,再凭借其个人判断、直觉和经验,进一步处理数据,这是老一套的办法,但也十分管用。然而,有太多的证据表明,如果常规合作持续下去,人在决策时往往犯各种问题、小毛病,还常常带有偏见。因此,在我们写这本书时,我研究的那些公司以及我由衷敬佩的那些公司,正努力超越常规合作关系。它们努力让人类在决策时保有一席之地,但如果事实证明,算法、数据和依据有利于做出更好的决策,我们就需要尊重事实,而不是因为人的自我感觉良好,就让人类的判断决定最终选择。这里我想举个例子,如果计算机能比训练有素的人类医生做出更为出色的医疗诊断,这个时候,我们还会为了让医生自我感觉良好而继续让医生诊断吗?我想,如果想要大众更健康,就绝不应该这么做。如果事实证明,当然我也希望,在机器学习兴盛的时代,如果机器比人更擅长某个任务,我们就应当把更多工作交给机器来做。
Get away from managers as gatekeepers-of -ideas 不要把管理者当作想法的“守门人”
So part of the standard partnership is rethinking the balance between minds and machines, but there’s another aspect to it that I find really interesting. Another part of the standard arrangement inside companies is that part of the job of management is to act as a gatekeeper for ideas. And what managers at all levels do, or what they think their job is, is to listen to the people that report to them, think about the ideas that get teed up, and say yeah, that sounds right to me, we’re going to go do that, versus that doesn’t sound right to me, we’re actually not going to do that. And the reason that an idea sounds good or bad is because, again, of the manager’s intuition, judgment, experience – those all sound ok. Here are other reasons why one idea sounds good or bad. The implicit and explicit biases that the manager has, the blinders that they have on, the experiences they don’t have, and the glitches in their mental hardware. One thing that I don’t want to see continue is that we let glitches in the mental hardware of managers determine which ideas get to see the light of day and get a fair hearing inside of organizations. I want to get out of that business. And when we were researching the book we came across companies where it was clear to us that they were working really hard – the managers, the leaders of that organization were working very hard not to intervene and be good gatekeepers and say green light red light to different things, but to get out of that business as much as they could, and to say look, we hired you inside this organization because we trust you, we think you’re good at what you do, and because we value your ideas. What we want to do now is let you run with those ideas as long as they’re not going to be financially ruinous if they turn out not to be good, if we can do a smaller-scale, shorter-term experiment to figure out if this is a good idea or not, if you’re not going to trash the company’s brand or its reputation or get us in very serious legal trouble. Ok, put those parameters around it. But within that, the job of us, the managers and leaders of this organization, is not to gatekeep your ideas anymore. What we want to try to do instead is set up an experimental engine inside this company where you feed ideas into it, you iterate, you figure out which of those ideas are good ones or not. You double down on the good ones, you try to learn from the ones that didn’t work out and walk away from them, but you want a lot of different people pouring ideas into that experimental engine. And some of the companies that I came to respect a lot are saying, look, come one come all. Or they’re trying to get closer to that view of the world. And we’re not only going to let people whose job title says “innovator” or who work in the R&D lab to pour good ideas into that experimentation funnel. We’re going to try to open that up fairly broadly. And it’s difficult homework. It’s easy to say and hard to set up, but the companies that are accomplishing that are seeing a higher rate of good ideas, a higher rate of successful experiments, and they’re charging ahead more quickly into a better future, in every sense of the word.
刚才我们提到,常规合作关系涉及重新思考人和机器之间的平衡,其实它还涉及一个很有意思的方面。公司内部通常认为,管理者的职责是扮演想法的“守门人”。因此,各级管理者所做的,或者说他们对自己工作的设想,是倾听汇报者的意见,思考具有启迪性的想法,然后回复汇报人:“对,这想法听起来不错,我们就这样做吧”或者是“在我看来这想法好像不行,我们不打算这么做”。这些想法听上去是好是坏,其评判标准取决于管理者的直觉、判断和经验。但是还会有一些其他原因,包括,管理者可能怀有隐性或显性的偏见,自身存在盲点,经验不足,思考错误。我不愿再看到管理者的错误思考最终决定要落实的想法,并获得公司内部的聆听和认同。我不想再碰到这种事情。在写这本书、做调研的时候,我们发现一些公司显然正努力改善上述情况,这些公司的管理者和领导者尽量不干涉员工想法,他们并不想扮演想法的“守门人”,对不同的想法亮起红绿灯,而是尽可能地减少干涉。他们会说:我们雇佣你是出于信任,我们觉得你能做好自己的工作。因为我们重视你的想法,现在你们要做的是将想法付诸实践,哪怕结果不好,只要不造成财务损失就行。我们可以进行小型、短期的实验来检验想法,只要你不会损害公司品牌及声誉,或者让公司陷入法律纠纷。除了这些要求,公司的管理者和领导者不会再干涉你们的想法。我们要做的是在公司内部建立一个试验引擎,你们可以不断向引擎注入想法,找出其中的好想法。你们可以加倍努力去实现这些好想法,并从失败案例中吸取教训,避免重蹈覆辙,但这也意味着首先要有许多不同的想法注入到这一试验引擎当中。我敬佩的一些公司会说:“欢迎大家踊跃参与”,还有一些公司正向这种做法靠拢。同时,我们不仅要让“创新者”或研发实验室工作人员将好想法注入到该实验引擎,也要让更多的人参与进来。这一任务十分艰巨,说起来容易,但做起来难。然而,践行这一点的公司的确更能发现好想法,更容易取得成功,而且就各方面而言,它们都能更迅速地奔向美好未来。
【Summary】
Building Mind-Machine Combinations: Move Past Standard Partnerships, with Andrew McAfee, Associate Director, MIT Center for Digital Business and Author, Machine, Platform,Crowd.
Rethink the balance between minds and machines
• In status quo, organizations, machines are used for rote information processing. Human intuition, judgment, and experience are then applied on top. But human decision-making processes are too buggy and biased to let the status quo endure.
• When algorithms, data, and evidence work together to make better decisions, allow
them to lead. Avoid layering human judgment on top just because people feel good
about making decisions.
Get away from managers as gatekeepers-of -ideas
• Biases, blind spots, and lack of experience can lead managers to favor certain ideas over others. Don’t let the flawed mental hardware of one person determine which ideas get promoted.
• Allow employees to test their ideas within certain parameters, such that those ideas:
◦ won’t be financially ruinous
◦ lend themselves to small-scale, short-term experiments
◦ won’t degrade the brand or lead to legal trouble
• Set up an experimental engine into which employees feed their ideas. Iterate and double down on the good ideas. Learn from, and walk away from, the bad ones.
在这节课中,安德鲁·麦卡菲从人和机器、管理者的角度为我们讲解了什么是“常规合作关系”,同时也告诉我们如何超越“常规合作关系”。
一、重新思考人和机器之间的平衡
• 当今企业让机器机械式地处理信息,再让人凭直觉、判断、经验做出决策。但人的决策充满了缺陷和偏见,这种状况不应该持续下去。
• 当算法、数据和依据协同起来,能做出更好的决策时,就应该把它们放在决策的首位。不要仅仅因为人们自我感觉良好就让人类的判断决定最终选择。
二、不要把管理者当作想法的“守门人”
·偏见、盲点,经验缺乏会导致管理者倾向某些想法。不要让持有偏见的人决定哪些想法得到落实。
• 允许员工在一定条件内测试他们的想法,条件如下:
◦不会造成经济损失
◦可以进行小型、短期试验
◦不会损害品牌价值或造成法律纠纷
• 建立一个试验性的引擎,让员工测试自己的想法。重复测试,加倍努力去实现好想法。从失败的想法中吸取教训,避免重蹈覆辙。