用在线课程扩大教育规模

他来了，好的。你上线了。你按了录制键吗？谢谢。

There he is, all right. You're on. Did you push the records? Thank you.

大家早上好，非常感谢邀请我。我本想亲自到场，但我在加州有个先前的约定，大约一小时后还有另一个演讲。所以无法亲自参加。但今天我想和大家分享我们在 Coursera 上所做的一些工作。我将共享屏幕，希望你们能看到我的幻灯片。如果幻灯片消失了，请告诉我。

Good morning everyone, thank you very much for having me. I would have loved to join you in person, but I had a prior obligation here in California, giving another talk here in about an hour. So I couldn't join in person. But what I wanted to do today is share with you some of the work we've been doing on Coursera. So I'm going to share the screen so that hopefully you can see my slides. And if the slides ever go away, please let me know.

Andrew，我们还看不到幻灯片。

Andrew, we can't see the slides yet.

哦，抱歉。让我看看。现在能看到幻灯片吗？

Oh, I'm sorry. Let me see. Are you able to see the slides now?

不，你的视频卡住了。

No, your video is frozen.

我明白了。我想一定是带宽问题，网络问题。实际上，我用了视频。好，现在能看到你的桌面了。太好了，谢谢。有时网络不够快就会这样。但如果幻灯片变得不可见或完全消失，请告诉我。

I see. I think it must be a bandwidth issue, a network issue. Actually, I used the video. Yeah, okay. Now we can see your desktop. Yeah, great. Thank you. Sometimes when the network isn't fast enough, this happens. But if the slides ever become not visible or disappear entirely, I hope you will please let me know.

非常有趣。有人刚刚发了一条关于我刚说的评论的推文，它出现在我的推特通知里。

It's very interesting. Someone just sent a tweet about a comment that I just made, and it appeared in my Twitter notification.

如各位所知，我是斯坦福大学教授，也是 Coursera 的联合创始人。如今大多数人永远无法上标准课程，但我们正在改变这一点。今天，我们与美国顶尖学府如斯坦福、普林斯顿、哥伦比亚、加州理工，以及世界各地的优秀学校如北京大学、耶路撒冷大学、多伦多大学合作，将这些学校的课程免费上线，供任何人学习。我们利用技术大规模提供优质教育。我自豪地说，如今全球前 60 所大学中有 30 所，以及 14 个不同国家的排名第一或第二的大学，都在与我们合作创建在线课程。

As most of you know, I'm a Stanford professor and also co-founder of Coursera. Most people today will never get to take a standard class, but we're changing that. Today, working with such great schools in the US as Stanford, Princeton, Columbia, Caltech, and many great schools around the world like Peking University, University of Jerusalem, University of Toronto, we're taking courses from all of these great schools and putting them online free for anyone to take. We're using technology to offer a great education at scale. I'm proud that today 30 of the top 60 universities worldwide, as well as the number one or two or both in 14 different countries, are working with us to create online courses.

让我分享这个故事是如何开始的。作为斯坦福教授，我通常教一个 400 人的班级。就在两年多前，差不多两年零一个月前，我把我的课程放到网上，吸引了 10 万名学生。为了说明这个数字，要接触到同等规模的受众，我需要在斯坦福大学教 250 年。所以自从从斯坦福推出最初几门课程以来，这场运动和 Coursera 都大幅增长。如今，我们拥有超过 100 个合作伙伴、500 多门课程和 500 多万学生，是全球最大的 MOOC 平台。

Let me share with you my story of how this started. As a Stanford professor, I usually teach a 400-student Stanford class. It was just over two years ago, almost two years and a month ago, that I placed my course online and it reached an audience of 100,000 students. To put that number in context, for me to reach a comparable size audience, I would otherwise have had to teach at Stanford University for 250 years. So since launching the first couple of classes out of Stanford, the movement and Coursera have grown substantially. Today, with over 100 partners, over 500 courses, and over 5 million students, we are by far the largest MOOC platform around the world.

两年内五百万学生。为什么这么多学生想上我们的课？我认为首先，这些是真正的课程。每门课程在特定日期开始，持续固定时间，每周有常规讲座和作业。如果学生完成作业并达到高分标准，他们还可以获得证书，可以放在简历上，用于寻找更好的就业或教育机会。

Five million students over two years. Why do so many students want to take our courses? I think first and foremost, these are real courses. Each course begins on a certain day and runs for a fixed period of time, with regular lectures and homeworks each week. If the students complete the homeworks and meet a high grading bar, they can also receive a certificate that they can put on their resume and use to find better employment or educational opportunities.

我知道你们很多人都听说过 MOOC 运动。让我快速展示一下学生上这些课的体验，以及一些你可能没见过的新发展。像其他平台一样，我们使用视频教学。让我展示一个典型讲座视频的片段，并强调加速播放功能（1.5 倍或 2 倍速）、英语字幕（对非母语者有用）以及志愿者编写的外语字幕。事实证明，允许学生以加速速度播放视频，可以让他们更有效地利用时间。他们可以以 1.5 倍或 2 倍速轻松浏览材料，因为如果错过了什么，可以倒回去。

I know that many of you will have heard of the MOOC movement. Let me quickly show you what the student experience is like taking one of these courses, and also show you some of the newer developments that you might not have seen. Like many others, we use video-based instruction. Let me show you a short clip of what a typical lecture video looks like, and highlight the accelerated playback at 1.5x or 2x speed, English captions which are useful for non-native speakers, as well as foreign language captions written by volunteers. It turns out that allowing students to play the video back at accelerated speed allows students to make much more efficient use of their time. It allows students to breeze through the material at 1.5x or 2x because if they miss something, they can just go back.

大约六个月前，我在加州 LinkedIn 的一个派对上遇到了一个上过我机器学习课的学生。他走过来对我说：‘哦，教授，我真的很喜欢机器学习课。我学到了很多机器学习知识，发现它对工作很有用。而且我喜欢能以 2 倍速观看你所有的视频，这样我就能更高效地利用时间。所以我很喜欢你的课，但看了 20 小时你以 2 倍速讲话的视频后，现在终于见到你本人，我真的很惊讶你现实中说话这么慢。’但这确实让学生更有效地利用时间。

It's about 6 months ago that I was at a party at LinkedIn here in California when I met one of the students that had taken my machine learning course. He came up to me and said, 'Oh Professor, I really enjoyed the machine learning class. I learned a lot of machine learning, found it really useful for my job. And I also loved that I could watch all of your videos at 2x accelerated playback because that allowed me to get much more done with my time. So I really enjoyed your class, but having watched 20 hours of videos of you talking at 2x, now that I finally meet you in person, I'm really surprised that in real life you talk so slowly.' But it does let students make more efficient use of their time.

除了在线发布讲座，我们还鼓励教师将概念分解成 10 分钟的小片段。这样教师可以发布可选的先修或高级材料，让学生挑选最感兴趣的部分，从而摆脱一刀切的教育模式。但课程远不止讲座。我们知道学生最好的学习方式不是被动听讲，而是通过实践材料。换句话说，学习往往不是发生在讲座中，而是发生在作业中。因此，我们为学生创造了大量自我测试和实践材料的机会。

Beyond posting lectures online, we also encourage instructors to take concepts and break them down into, say, 10-minute bite-sized pieces. This allows instructors to post optional prerequisite or advanced material and allow students to pick and choose the pieces that interest them most, thus moving away from the one-size-fits-all model of education. But the course is much more than just lectures. We know that students learn best not by passively listening but by practicing with the material. In other words, often the lecture is not where the learning takes place; it is in the homeworks that the learning takes place. Because of that, we build many opportunities for students to test themselves and to practice the material.

让我展示一个我们称为视频内测验的例子。

Let me show you an example of what we call in-video quizzes.

Andrew，屏幕又卡住了。

Andrew, the screen is frozen again.

哦，抱歉。你知道吗，我想也许我不用视频了，但你能看到一张写着‘do the upper left’的幻灯片吗？

Oh, sorry. You know what, I think maybe I'll try to do without the videos, but are you able to see a slide that says 'do the upper left'?

不，我们还在看那个有遗传学内容的视频教学幻灯片。

No, we were still seeing the video instruction slide with the genetics stuff on it.

我明白了。好的，抱歉。我想一定是网络连接慢。也许我继续讲，看看网络能否恢复。希望音频能正常工作，帮助传达我们描述的内容。

I see. Okay, all right. Apologies. I think there must be a slow network connection somewhere. Maybe I'll keep talking and see if the network comes back up. Hopefully the audio will work and help convey what we're describing.

除了视频讲座，我们所有的课程还为学生提供了大量与材料互动的机会。这包括在讲座视频中弹出的问题，以及大多数每周作业。在我们的课程中，学生通常每周需要花少量时间做作业，包括选择题、简答题、数学题。我们还有很多建模题，比如要求学生构建电子表格财务模型并上传到网站由计算机评分，或者让学生构建化学模型或代码模型等等。

In addition to the video lectures, all of our courses also have many opportunities for students to interact with the material. This includes questions that pop up right in the middle of the lecture video, as well as most weekly homeworks. Every week in our classes, students are usually asked to do a small number of hours of homeworks, including multiple choice questions, short answer questions, math questions. We also have a lot of modeling questions where we might ask students to build a spreadsheet financial model and upload that to the website to be graded by computer, or have students build chemical models or code models, and so on.

好吧，这很不幸。屏幕卡住了，这通常是网络连接慢的迹象。

All right, this is unfortunate. The screen is frozen, which is usually a sign of a slow network connection.

是的，画面在变化。我想它快恢复了。稍等一下。好的。嗯，你能看到吗？不，只看到你的脸和一个转圈图标。幻灯片还在加载。好吧，我们继续。如果幻灯片恢复了请告诉我。我试着不用视觉辅助来讲。好的。关于分组和自动批改作业，另一个好处是我们可以让学生多次尝试学习任务。在斯坦福的典型课堂中，学生提交作业，我们批改后发还给他们，如果他们得了 60 分（满分 100），那还能怎么办？很好。好的。为了展示我们作业的多样性，很多老师会设置投票问题。比如这个例子中，Dan Jurafsky 问：“如果你需要撕掉创可贴，你会用什么策略？直接扯掉还是慢慢撕掉？” 学生可以参与回答并看到其他人的想法。我们还有多种类型的题目，通过计算机自动批改，给学生即时反馈。此外，我们开发了一个应用平台，让大学合作伙伴能编写非常复杂的插件软件，主要是交互式练习。比如这里，学生在浏览器中实现一个电脑游戏。其他大学，比如伯克利音乐学院，也加入了组件，让学生用音乐做有趣的事情，或者探索复杂的临床结构。我们意识到，Coursera 不可能实现所有教学上有用的功能，所以我们决定成为一个平台，让大学合作伙伴实现他们的创意和教学愿景，我们帮助他们将这些推广到成千上万的学生。我对网站上交互的多样性和丰富性感到惊讶。到目前为止，我一直在讲计算机批改的作业，它的一个好处是允许学生多次尝试。在斯坦福的典型课堂中，如果学生错过了一次作业，他们写了，但你能怎么办？课程继续，他们只能带着 60 分的成绩继续。但在网站上，我们可以让学生多次尝试。我们提供工具帮助老师随机出题。比如这个例子，老师写一个问题，四个正确和四个错误答案，每次学生尝试测验，他们看到不同的选项。这样，如果学生没答对，我们鼓励他们回家学习，第二天再试。如果还是不对，再学习再试。这给了学生多次机会，在进入下一个主题之前掌握当前内容。我认为 Salman Khan 说过一句很有深意的话：在我们国家的教育体系中，固定的是学生在某个材料上花费的时间，比如我们课程中的 10 周，而变化的是学生掌握的程度。有些学生掌握了，我们叫他们 A 等生，有些掌握得少，叫 B 等或 C 等生。但教育应该帮助每个学生成功。有了自动批改，我们就能让学生多次尝试直到成功。这样，固定的是每个学生都应该掌握、都应该成功，而变化的是学生需要的时间和尝试次数。这种技术叫做掌握学习，它让每个学生有机会掌握材料，在进入下一部分之前打下更坚实的基础。

Yeah, it's changing. I think it's coming back. Just give it a second. Okay. Um, are you able to see it? No, it's just your face with a spinner on it. It's still loading the slide. Okay, yeah, let's keep going. If the slides come back, please let me know. I'm going to try to just talk without visuals. Okay. So one other nice thing about groups and auto-grading of assignments is that we can allow students multiple attempts at a piece of learning. Turns out that in a typical Stanford class, on campus at Stanford, I don't think a student submits a piece of work, then we grade it and return it to them, and if they got 60 out of 100, then what can you do? Great. Good. All right. So just to show you some of the diversity of homeworks we have, you know, many instructors pop up polling questions. In this example, Dan Jurafsky asks: "If you need to remove a bandaid, what strategy would you use? You yank it off or would you slowly peel it off?" Long period of maintenance. Students get to respond and participate in the class and see what others think as well. We also have many different types of problems which we've computer-graded so that we can give students instant feedback. And moreover, we developed an app platform that allows our university partners to write very sophisticated plug-in pieces of software, mostly interactions and interactive exercises, such as here we have students implement a computer game right in the browser, as well as other universities, Berkeley School of Music, have plugged in other components for letting students do interesting things with music or letting students play with complex clinical structures. And I think one thing we realized was that there's no way for Coursera to implement everything that is pedagogically useful, and so we decided to take the approach of being a platform, or being an application platform, that allows other university partners to implement creative things and implement their pedagogic vision, and we help them push this out to a scale of tens of thousands of students. But I think I'm amazed at the diversity and the richness of the interactions that are possible on the website. So far I've been talking about computer-graded assignments, and one really nice thing about that is that you can allow students multiple attempts at a piece of work. In a typical Stanford class, if a student misses a piece of work and they write it, what can you do? The class moves on, and they move on with their homework which they scored 60 out of 100. But on the website, we can allow students multiple attempts. We actually provide tools to help instructors randomize questions. So in this example, the instructor writes the question, four right and four wrong answers, and each time the student attempts the quiz, they see a different set of options. So what this does is if a student attempts a quiz and they don't quite get it right, we encourage them to go home, study, come back the next day and try again. And if they still don't get it right, go home, study some more, and come back and try again. And what this does is gives students multiple attempts to achieve mastery of a subject before moving on to the next subject. I think what Salman Khan said something very profound: that in our country, in our educational systems today, what is held constant is the amount of time that someone spends on a piece of material, say 10 weeks in the case of our course, and what is variable is the degree to which the student gets it. Some students get it, we call them A students, and some students understand the material less, and we call them B and C students. But instead, I think education should be about helping every student succeed. And with auto-grading, it gives us the resources to allow a student to attempt a piece of work multiple times and multiple times to succeed, so that now what can be held constant is that every student should get it, every student should succeed, but what can be variable is the amount of time or the number of attempts a student needs to make before mastery. So what mastery learning, which is what this technique is called, gives every student an opportunity to master the material and have a more firm foundation to build on before they move on to the next piece of content.

到目前为止，我只讲了计算机批改的作业，它的优点是：第一，给学生即时反馈；第二，允许我们大规模实施。但并非所有东西都能由计算机批改。我想起另一个故事：大约一年半前，我访问了费城的宾夕法尼亚大学，我了解到，如果你去找一位诗歌教授，向他展示你引以为傲的、闪亮的新选择题随机化自动批改技术，并试图说服他只使用选择题来教他的课，他会请你离开他的办公室。所以我们意识到，并非所有东西都能由计算机批改。为了让老师提供更开放式的作业，比如让学生提交论文，我们需要一种批改方式。一位教授无法批改 10 万学生的论文，但我们可以让学生互相批改。这种技术叫做同伴互评，在小班中已被证明能给出相当准确的分数。这里，学生评分和老师评分高度相关。所以我们把同伴互评的想法大规模应用，涉及数万名学生。它的工作方式是：学生先完成作业，比如写论文，或者在这个例子中，学生被要求建造一个实物，然后用手机拍照，上传到网站由其他学生评分。这张幻灯片展示了“为真实世界设计”课程中三个学生的实际提交。完成作业后，每个学生接受培训，学习如何评分，最后每个学生需要批改其他五名学生的作业，作为交换，你也会收到五名学生的反馈。这很有效。我们和大学合作伙伴已经证明同伴互评相当准确。普林斯顿大学的一项研究显示，同伴评分与老师评分高度相关，误差大多在 4-5 分之内。此外，同伴互评让学生能给出非常有用的定性反馈。我听到很多学生说他们喜欢同伴互评过程，不仅因为能学习和提交作业，还能看到其他学生的作品。

So far I've just been talking about computer-graded assignments and the advantage of that: first, it offers the student instant feedback, and second, it allows us to implement massive scale. But not everything can be computer-graded. I guess another story: I learned an interesting lesson about a year and a half ago. I was visiting the University of Pennsylvania in Philadelphia, and I learned that if you go to a poetry professor and if you show this poetry professor your shiny new multiple choice randomization auto-grading technology that you're so proud of, and if you try to convince this poetry professor to teach his class using only multiple choice, I learned that he will invite you to exit his office. So we realized that not everything can be computer-graded. To enable instructors to offer more open-ended work, such as having students submit essays, we needed to provide a way to get that graded as well. One professor can't grade 100,000 students' essays, but what we can do is have the students grade each other's work. This is a technique called peer grading, and it's been shown in small classes that you get fairly accurate grades. Here, the student grades and teacher grades show a high correlation. So what we did was we took the idea of peer grading and applied it at scale, with tens of thousands of students. The way it works is: a student in one of these classes first does the work, say write an essay, or in this case students were asked to build a physical artifact and then use their cell phone camera to take a picture of the physical artifact that they have built, and then upload the picture of what they designed to the website to be graded by other students. So students first do the work, and this slide shows three actual submissions by actual students in the "Design for the Real World" design class. Having done the work, every student is then trained to grade other students' work, and finally every student is required to grade five other students' work, and in exchange you get feedback from five other students. So this works. We and our university partners have shown that peer grading is fairly accurate. This study done by Princeton University shows that the peer-assigned score is highly correlated with the teacher-assigned score, mostly within 4-5 points. Moreover, peer grading allows students to give each other very useful qualitative feedback. I've heard from quite a lot of students that they enjoy the peer grading process, whereby not only do you get to learn and submit homework, you get to see other students' work.

并且也扮演一点老师的角色，所以你给其他学生的作业提供反馈。很多学生说他们非常珍视这一点，认为这是学习体验的一部分，而且成为社区的一员、能够讨论、能够帮助其他学生完成作业也很有趣。所以我认为 Coursera 加上所有关键技术，使我们能够提供广泛的课程，从 STEM 到人文学科、社会科学、工程、商业、法律、金融等。其中一个垂直领域是教师专业发展。今天我们重点关注，我们也花时间研究 K-12 教育，我们意识到我们的技术非常适合服务那些还没学会坐两个小时做作业的七岁学生。相反，我们与一个最大的教师专业发展组织合作，这些组织培训教师。今天我们有超过 35 门教师专业发展课程，教授 K-12 以及世界各地的其他级别的教师如何成为更好的老师。所以举一些我最喜欢的课程例子：如果你们有小孩或孙子孙女，我们有一门关于如何与幼儿创造有意义的互动、情感上有意义的互动的课程。或者如果你的孩子大一点，你希望你的孩子学什么？我们希望孩子好奇、勤奋、有道德。学校教什么？学校教数学、代数、化学。所以我们有一门由 KIPP 创始人之一讲授的关于如何培养品格教师的精彩教师专业发展课程。所以作为老师或家长或祖父母，如果你想教自己的孩子品格，我们也有这门课，由这个星球上最了不起的老师之一、KIPP 的创始人之一讲授。所以让我们看看。从讲座的角度和作业的角度讨论了学生体验。学生体验的第三个也是最后一个组成部分是社区。一个班有两万名学生的一个好处是社区开始围绕材料形成。这也意味着当学生有问题时，教授无法回答十万名学生的问题，但你可以让学生互相回答问题。我们开发了一个讨论论坛，学生可以发帖并回答别人的问题。由于我们班级学生数量庞大，上次我们在我的第一门课上测量时，当一个学生提问时，中位响应时间，即等待回复的时间，是 22 分钟。我认为 22 分钟的响应时间是一个惊人的服务水平。今天我们也继续做很多其他事情来建立社区感。例如，昨天刚刚在我的机器学习课上宣布，这是第五轮了。差不多两年前，我在斯坦福的小团队推出了我的机器学习课。其中一个班级开始了这个新事物。我的机器学习课现在已经是第五次开设了，昨天刚刚宣布，我们还试验了一个辅导项目，我们邀请那些完成了早期版本课程的学生回来担任老师和导师，通过一对一的 Google Hangout 视频聊天，让当前课程的学生可以与往届成功学生交流，获得更多学习帮助。所以我们不断进行新的实验来建立社区。我对辅导感到兴奋的原因是它让最优秀的学生也能站出来当老师。此外，我们现在开始建立一个跨越多个课程开设的社区。今天 Coursera 服务于一个非常多样化的非传统人群。超过 80% 的学生已经拥有学士学位。所以尽管 Coursera 在高等教育中有其用途，但当你想到 Coursera 的学生时，不要想象一个 18 到 21 岁的大学生。想象一个二十多岁或三十多岁的人，通常已经受过教育，他们从我们这里学习更多知识。我认为在当今快速变化的世界中，知识的半衰期正在缩短。我认为旧的教育模式——你上大学四年，然后在接下来的 40 年里只依赖你学到的东西——在当今快速变化的世界中意义不大。所以 Coursera 的主要直接效果是我们让许多在职成年人重新回到教育系统。因为想想看：大多数职场专业人士没有免费的教育替代方案。每周雇两次保姆去社区大学上夜校非常困难。所以我认为 Coursera 提供的免费访问让许多在职成年人回到了系统。我们还有来自世界各地的学生，这增加了多样性，使得课堂讨论最活跃。由于学生群体的国际性，我们也花了很多时间在平台的国际化上。今天如果你从中国访问 Coursera，你会看到这个网站。它看起来非常像一个中文网站。当你浏览不同页面时，你会看到中文网站。所以我们支持中文，但我们正在努力进一步将平台国际化成多种语言。至于课程内容，我们大多数课程目前仍然是英文的，但我们有中文、西班牙语、意大利语、法语、德语、葡萄牙语等语言的课程。最后，我们也在与内容创始人合作，将原本用英语讲授的课程（如密歇根大学的课程）翻译成其他语言并配上字幕，让国际课程内容对世界各地的人们可访问。我们大多数课程是用英语授课，所以我们从英语翻译成多种语言。但未来，我们计划将其他语言的课程翻译成英语，想象一下你拿一门中文或西班牙语课程翻译成英语，这样美国学生也能从优秀大学学习，比如从印度理工学院学习，即使美国学生不会那种语言。所以让我们看看。从学生角度来看，体验的第四个也是最后一个主要组成部分是获得证书的机会。许多学生想要完成这些课程，这些是真正的课程。也许另一个人，你知道，这是我意识到学生对课程认真程度的时刻。所以在我家，每年圣诞节...

And play a bit of a role of a teacher as well, so you give feedback to other students about their work. Many students said that they really value this as part of their learning experience, and that it was also just fun to feel part of the community, to be able to discuss, to be able to help other students with their work. So I think Coursera, together with all the key technologies that have enabled us to offer a broad range of courses from STEM to Humanities, Social Sciences, Engineering, Business, Law, Finance, and other ways. One of the vertical subjects that the course reflected about is actually teacher professional development. Where as today we focus, we also spend some time examining K-12 education, and we realize that our technology today is highly right for serving the seven-year-old students that haven't yet learned to just sit there for two hours and do work. Instead, what we did was work with one of the greatest teacher professional development organizations, the organizations that train teachers. And today we have over 35 teacher professional development courses teaching K through 12th and teachers at other levels all around the world how to be a better teacher. So some examples of my favorite classes: if any of you have young children or grandchildren, we have a class on how to create meaningful interactions, emotionally meaningful interactions with very young children. Or if you have children a little bit older, what do you want your children to learn? We want our children to be curious, hardworking, ethical. What do our schools teach? Our schools teach math, algebra, chemistry. And so we have a fascinating teacher professional development course called by one of the founders of KIPP on how to create teachers of character. So as a teacher or as a parent or grandparent, if you want to know how to teach your own children character, we have the course on that too, taught by one of the most amazing teachers on this planet, one of the founders of KIPP. So let's see. Talked about this student experience from the perspective of the lectures and from the perspective of the works. The third and final component of the student experience is the community. One of the nice things about having 20,000 students in a class is that the community starts to form around the material. This means also that when students have questions, well a professor can't answer 100,000 students' questions, but you can have the students answer each other's questions. We developed a discussion forum where students can post and answer others' questions. Because of the sheer number of students in our classes, the last time we measured it actually in my first class, when a student asked a question, the median response time, that is how long you have to wait to get a response, was 22 minutes. And I think a 22-minute response time is an amazing level of service. Today we're also continuing to do a lot of other things to build up a sense of community. For example, just announced yesterday in my machine learning class, this is the fifth round of it. So it was almost exactly two years ago that my small team at Stanford launched my machine learning class. And one of the classes started this new thing. My machine learning class is running for the fifth time now, and just announced yesterday, we've also experimented with a tutoring program in which we have taken students that had completed earlier versions of my class and we have invited them back to be teachers and mentors, one-on-one Google Hangout video chats, so that students in the current offering of the class that want to go talk to some of the older students and the previous years' successful students can get more help with learning. So we're constantly running new experiments to build up the community. And the reason I'm excited about tutoring is because it allows the best students to step up to be teachers as well. Moreover, we now start to build a community that spans multiple offerings at the same place. Today Coursera serves a very diverse, non-traditional population. Over 80% of our students already have a bachelor's degree. So even though Coursera has its use in higher education, when you think of Coursera students, don't picture an 18 to 21 year old college student. Picture someone in their late 20s or maybe in their 30s, often already educated, that are taking courses from us to learn more. I think that in our world today, in our rapidly changing world, the half-life of knowledge is decreasing. And I think the old model of education where you go to college for four years and for the next 40 years just rely on what you learned, I think that makes little sense in today's rapidly changing world. So the major proximal effect of Coursera is that we're bringing many working adults back into the educational system. Because think about it: most working professionals don't have free alternatives to education. It's very challenging to hire a babysitter twice a week to go take night classes in community college. So I think the free access that Coursera provides brings many working adults back into the system. And we also have students all around the world, which increases the diversity that makes for the most lively discussions in classes. And because of the very international nature of our student population, we've also spent a lot of time on internationalization of the platform. Today if you were to visit Coursera from China, this is the website you will see. It looks very much like a Chinese website. And as you browse through different pages, you see the Chinese website. So we support Chinese, but we're working on further internationalizing the platform in many languages. As for the course content, most of our courses today are still in English, but we have courses taught in Chinese, Spanish, Italian, French, German, Portuguese, and quite a few other languages now. And finally, we're also working with content founders to take classes that were originally taught, such as University of Michigan courses, and translate them into other languages with subtitles to make the international course content accessible to people around the world. Most of our courses are taught in English, so we're translating from English into multiple languages. But in the future, we plan to take courses taught in other languages, imagine you take a Chinese course or Spanish course and translate that to English, so that an American student can also learn from great universities, can learn from Indian Institute of Technology, even if the American student doesn't speak that language. So let's see. From the student perspective, the fourth and final major component of the experience is the opportunity to earn this certificate. Many of our students want to complete these classes, and these are real classes. Maybe another person, you know, this is the moment when I learned just when I got an emotional understanding of the seriousness that students take these classes. So in my family, every Christmas...

我和父亲有一个传统，就是庆祝圣诞节，然后圣诞节后的第二天我们总是去逛电子产品。我们已经这样做了十多年。但去年，去年 12 月，有些不同。12 月 26 日，我和岳父按照家庭传统，去了香港一个巨大的电子产品购物集市。和往年一样，我们沉迷于各种电子小玩意。但去年不同。那天下午 2 点，我父亲看了看手表说：‘天哪，都 2 点了，我得去用电脑了。’然后他就把我丢在那里，回酒店房间去花整个下午做这门课了。

My father and I have a tradition of celebrating Christmas, and the day after Christmas we always go shopping for electronics. We've done this for over a decade. But last year, last December, something was different. On December 26, my father-in-law and I, as was our family tradition, went to this huge electronic shopping bazaar in Hong Kong. Same as every year, we were geeking out over electronic gadgets. But last year was different. At 2:00 that day, my father looked at his watch and said, 'Oh my God, it's 2:00, I have to go to my computer.' And he ditched me right there to go back to his hotel room to spend the afternoon working on this course.

很多学生可能会花几个小时完成作业。当他们成功完成作业时，我们认为如果能给他们颁发一个证书来展示他们的成果，那会很好。颁发经过验证的证书的一个挑战是，如果你有成千上万的学生遍布全球，你怎么知道这些人是谁？所以，我最引以为豪的创新之一就是我们开发的系统，它能够大规模地进行身份验证。它的工作原理基于我在斯坦福做的一些旧研究。事实证明，当你在键盘上打字时，你的打字节奏是独一无二的，就像指纹一样。所以在签名追踪中，为了验证程序，当学生注册时，我们首先要求他们出示一张手持身份证的照片，拍下身份证照片，并将他们的照片或真实姓名关联起来。然后每次他们提交作业时，我们结合使用摄像头拍照和测量打字节奏，以确保是同一个人在做作业。在此基础上，我们颁发所谓的经过验证的证书，该证书来自大学，并被雇主和招生委员会视为可信的凭证。

So many students may spend hours trying to complete their homework. When they do the homework and are successful, we thought it would be nice if we could issue them a credential to show what they have done. One of the challenges with issuing a verified credential is that if you have tens of thousands of students around the world, how do you even know who these people are? So one of the innovations that I'm most proud of is the system we developed to help build identity verification at scale. The way it works is based on some old research I did at Stanford. It turns out that when you type on your keyboard, the way your typing rhythm is unique to you, like a fingerprint. So in the Signature Track, to verify the program, when the student signs up, we first ask them to show us a picture holding up their photo ID, take a picture of the photo ID, and associate their photo or their real names. Then every time they submit a piece of work, we use a combination of taking their picture using the webcam as well as measuring your typing rhythm to make sure that the same person is doing the work. On that basis, we issue what we call a verified certificate that comes from the university and is treated as a credential that employers and admissions committees take seriously.

我们来看看。教授一门有这么多学生的课程的一个好处是你可以收集数据。也许因为时间关系，我可能会跳过这个，直接进入下一部分。

Let's see. One of the nice things about teaching a course with so many students is the data you can collect. Maybe because of interest of time, I might skip this and just jump to the next piece.

到目前为止，我主要谈论的是大规模开放在线课程的学生，这些遍布全球的数十万学生。在过去的一年里，有人问我：如果任何人都可以在线免费学习斯坦福的课程，为什么还有人愿意花 2000 美元获得斯坦福的学位？我认为答案是，大学的角色不应该只是广播内容，因为内容很容易。相反，大学的角色是学生来到大学不仅仅是为了内容，而是为了与教授以及同样有动力的同学互动。所以简单来说，我认为头脑不是一个需要被填满的容器。如今，我们的许多大学合作伙伴正在实施所谓的翻转课堂模式。对于那些知道它是如何开始的人来说，推广翻转课堂理念的其实是联合创始人 Daphne 在两年前 3 月 1 日的一个洞察。Daphne 是我招募来教学的另一位学者，但她有一个惊人的洞察：在线内容也可以用来服务在校学生，而这正是我们大多数大学合作伙伴所关心的。正因为如此，我们最终走到一起成立了 Coursera，如今它服务于数十万学生，但同样重要的是，我们的许多大学合作伙伴也在使用我们来实现翻转课堂。它的工作方式是，你可以将讲座内容放在网站上，让学生去 Coursera 网站观看内容，讽刺的是，这比传统大学课程更具互动性，并能给学生更即时的反馈。这样做的好处是，它解放了课堂时间，使得课堂上的学生可以利用时间互相讨论材料，或者与教授进行更深入的互动。需要明确的是，我认为今天的 Coursera 服务于两个不同但同样重要的人群。首先是那些原本无法接触到这些优秀大学课程的数十万学生。对他们来说，我们显然拓宽了教育机会。但第二个重要群体是斯坦福或加州理工班级里的 50 名学生。通过帮助这些大学在课堂上，我们也在提供更广泛的教育机会，并提高校内教育的质量，为学生和教师创造更多深入互动的时间。当然，研究表明，这些更具互动性的参与模式也带来了更好的学生出勤率、参与度和学习效果。这不是学生学习成绩提高一两个百分点，而是提高了 30%。所以我认为，我们帮助大学合作伙伴改善校内教育这一点非常重要。

So far I've been talking mainly about the massive open online course students, these hundreds of thousands of students around the world. Over the last year, I've been asked: if anyone can go online and take a free Stanford course, why on Earth would anyone still pay $2,000 for a Stanford degree? I think the answer is that the role of a university should not be to just broadcast content, because content is easy. Instead, the role of a university is that students come to our universities not just for the content, but for the interactions with professors and interactions with other equally motivated students. So quickly, I think the mind is not a vessel to be filled. So many of our university partners today are implementing what is called the flipped classroom model. For those of you who know how it started, popularizing the flipped classroom philosophy was really the insight that my co-founder Daphne had two years ago, March the first. Daphne was one of the other academics that I recruited to teach, but she had this amazing insight that the online content can also be used to serve the on-campus students, which is really what most of our university partners care about. Because of that, we ended up getting together to form Coursera, which today serves hundreds of thousands of students, which you know about, but equally important, many of our university partners are also using us to implement the flipped classroom. So the way that works is you can put your lecture content on a website and send students to the Coursera site to watch the content, where ironically it can be more interactive and give more immediate feedback to students than a traditional university course. What this does is it now frees up the classroom time so that students in the classroom can use the time to discuss the material with each other or engage in much deeper interactions with the professor. To be clear, I think of Coursera today as serving two distinct but equally important populations. First is the hundreds of thousands of students out there who otherwise would never have access to these great university courses. To them, we clearly broaden access to education. But the second important group is the 50 students in a Stanford class or a Caltech class. By helping these universities in the classroom, we are also giving much greater access to education and improving the quality of on-campus education as well, by creating much more time for students and faculty to engage more deeply. And of course, research studies show that these more interactive modes of engagement result in much better student attendance, engagement, and learning as well. This is not a one or two percent improvement in student learning; this is 30% improvement in student learning. So I think it's important that we are helping our university partners improve their on-campus education.

最后，总结一下。我从事这些工作已经两年多了，我对我们的大学合作伙伴所取得的成就感到谦卑，也对前方的漫漫长路感到谦卑。让我分享一件我非常希望看到的事情。一两年后，我希望有一个低收入的人，也许是印度人，也许是中国人，也许是加州本地人。我希望那个低收入的人能够注册一个为期七周的课程。他们必须非常努力才能完成，因为这些是顶尖大学的低门槛课程。但我希望那个低收入的人能够把它列在简历上，并用它找到一份更好的工作，为自己谋生。因为我们可以大规模地做到这一点。我们可以利用技术为任何人提供教育机会，一个更好的机会。这意味着我们可以给每个人一个更好的生活的机会。我认为这个国家，美国，需要它。我认为其他国家可能更需要它。但如果我们能给地球上的每个人一个更好的生活的机会，我认为我们可以改变这个星球的运作方式。所以我认为，就在大约两年前我们推出这个项目时，我开始到处谈论这个想法：教育不应该只属于特权阶层。教育应该是一项基本人权。凭借教育带来的机会，以及利用技术大规模低成本地提供教育，我希望在未来几年里，我们能让教育成为一项人权，给每个人一个机会。非常感谢。

Finally, just to wrap up. I've been working on these groups for over two years now, and I'm humbled by how far our university partners have come, but also by how long the road ahead is. Let me share with you something that I would quite like to see. A year or two from now, I want there to be a lower-income individual, maybe someone in India, maybe someone in China, maybe someone right here in California. I want that lower-income individual to be able to sign up for a course for seven weeks. They'll have to work really hard to finish it because these are low courses from top universities. But I want that lower-income individual to be able to list it on their resume and use it to find a better job and earn a living for themselves. Because we can do that at scale. We can use technology to offer anyone a shot at education, a better opportunity. What that means is we can give everyone a shot at a better life. I think this country, the United States, needs it. I think other countries probably need it even more. But if we can give every person on this planet a shot at a better life, I think we can change the way the planet works. So I think that it was just about two years ago when we launched this that I started going around talking about this idea that education shouldn't be only for the privileged. Education should be a fundamental human right. With the opportunities that education can bring, and with technology to offer education inexpensively at scale, I hope that in the coming years we can make education a human right and give everyone a shot. Thank you very much.

Andrew 愿意回答问题。我们还有几分钟。你能把摄像头转过来吗？我想看看观众。

Andrew is willing to take questions. We have a few minutes. Could you turn the camera around? I'd love to see the audience.

我不确定线缆是否允许我这样做。我们试试看。好了，大家看到了。

I'm not sure the cabling will allow me to do this. Let's see. So there's everybody.

罗里第一个提问。麦格瑞博士，您提问，我会重复给安德鲁听。

Think Rory's first in line. Dr. McGreal, if you'll ask the question, I'll repeat it to Andrew so he can hear you.

大卫，几年前在一次会议上，你向我们提出了一个观点：我们处在一个新世界，买书却不拥有它，比如电子书。而 Coursera 则有一个新概念：你学习却得不到什么。你学的东西归 Coursera 所有，你不能用所学去你想去的地方。我的问题是：我能理解 Coursera 作为一家私营公司会加入这种条款，但我不理解的是，全世界这么多大学怎么会同意一家私营公司这么做。谢谢。

David, a few years ago at one of these conferences, you brought up the idea to us: we're in a new world where you buy but you don't own it, with ebooks. And with Coursera, you have a new concept where you learn and you don't get. So what you learn is owned by Coursera, and you are not allowed to use what you learn to get where you want to. My question would be: I can understand why Coursera, as a private company, would put that speculation in. What I don't understand is all these universities around the world that would agree with that at a private company. Thank you.

下一个问题。好的，我认为问题实际上是关于 Coursera 使用条款中的一项规定，禁止你将 Coursera 的材料与别处的收费或学分项目结合使用。问题是：你如何让那些怀有免费教育愿景的大学合作伙伴，参与一个使用条款相当严格的计划？罗里，这样概括公平吗？

Next question. Okay, so I think the question is really about a provision in the Coursera terms of use that prohibits you from using Coursera material in conjunction with a tuition-based or credit-bearing program somewhere else. And the question was: how do you get the university partners, who have this vision of free education for everyone, to participate in a program with terms of use that are really rather restrictive? Is that a fair summary, Rory?

我认为这是问题的关键。

I think that's the key of the question.

当然，这是个公平的问题。老实说，我们仍在思考如何处理。让我直接说明：有人问我们为什么不把内容以知识共享许可发布？现实是，我们的大学合作伙伴以及 Coursera 目前都承担着大量的内容制作成本。大学制作一个在线内容需要花费 2 万到 6 万美元。我认为我们所有人都首先致力于教育使命，但我们也需要找到一种方式来实现可持续的参与。所以我们的课程内容是免费的——世界上任何学生都可以完全从我们的课程内容中学习——但我们也试图找出一种方法，既能带来合理的收入，用于大学合作伙伴和我们自己，以确保整个努力是可持续的。所以我们考虑的一种商业模式是收费许可，尽管尚未决定投入多少。我们的想法是，如果宾夕法尼亚大学投资了 6 万美元制作一个在线内容，而另一所收费大学想用这个内容来服务他们的付费学生，那么收取一小笔费用似乎是公平的，因为这可能帮助那所大学提高质量并降低成本。但无论我们是否以及如何在近期或远期实际推行这一点，或者是否可能将服务条款改为不同的内容，我们仍在讨论中。我们的意图当然是帮助尽可能多的人将其用于自己的校园教育，但我们仍在研究具体操作。

Sure, it's a fair question. I'll be honest, we are still thinking through what to do about that. Let me just put this up right there: something we ask is why don't we put the content out under a Creative Commons license? And I think the reality is our university partners, as well as Coursera today, are running up substantial content production costs. It costs a university anywhere from $20,000 to $60,000 to create an online piece of content. And I think all of us are first and foremost committed to the education mission, but we also need a way to find a sustainable involvement. So our course content is free—any student in the world can learn entirely from our course content—but we're also trying to figure out if there is a way to bring in reasonable amounts of revenue, both for the university partners and for ourselves, to make sure the whole effort is sustainable. So one of the business models we were contemplating, though undecided how much to pursue, would be fee licensing. Our sense was that if the University of Pennsylvania has invested $60,000 into creating an online piece of content, and a different tuition-charging university wants to use that content to serve their tuition-paying students, it seemed fair to me to request a small fee for that, because it may help that other university raise quality and lower costs. But whether and how we are actually going to pursue this in the near future or distant future, or whether we might change the terms of service to something different, is something we're still discussing. Our intent is certainly to help as many as possible use this for their own campus education, but we're still trying to figure out the logistics.

好的，请继续。

Sure, go ahead.

好的，这是一个关于同伴互评如何运作的问题。如果有五个学生批改同一份作业，你是取平均值、中位数，还是取最高分、最低分？教师是否会抽查？请谈谈这是如何运作的。

Okay, so it's a question about the way that peer grading works. If you have five students grading one piece of student work, do you take a mean, a median, do you take the high, do you take the low? Are those ever spot-checked by faculty? Talk about how that works.

当然。通常我们取中位数——这比平均值好。当我们有 10 万名学生时，有时教师会进去查看一些，以了解同伴互评的运作情况。我们有 10 万份提交；教师可能会看一小部分样本，但现实是绝大多数提交永远不会被教师亲自查看。但学生仍然能得到有用的反馈。我应该说，作为我们正在开发的技术的一个例子，我们正在开发的一项即将推出的技术解决了这个问题。我们有大量数据，因此我们对评分了解得惊人地好。事实证明，如果你运气不好，碰巧遇到特别严格的评分者，我们可以测量出来，因为我们看到他们对其他人也给了低分。或者我们可以测量评分者是否随机打分，分数分布杂乱——我们也能检测到。我们可以进行数学修正，以调整你的五个评分者都很苛刻的情况。在我们的内部试点中——尚未推向网站，只是内部试点——我们将展示通过进行这些数学调整，了解谁严格、谁宽松、谁随机，我们可以得到更准确的分数。这还没有部署，但这是我们可以利用数据做的事情，比如一个班级有 5 万名学生，产生 20 万份提交。

Sure. Usually we take the median—that's better than the mean. And when we have 100,000 students, sometimes the faculty will go in and take a look at a few just to get a sense for themselves of how peer grading is working or not working. We have 100,000 submissions; the faculty might look at a small sample, but the reality is the vast majority of them will never be looked at personally by the faculty. But students still get useful feedback. I should say, as an example of the sort of technology we work on developing, one piece of upcoming technology that we're working on addresses this. We have so much data that we understand grading surprisingly well. It turns out that if you are unlucky and you happen to have graders who were unusually tough, we can measure that because we saw that they gave low grades to everyone else too. Or we can measure if a grader is just random and gives grades all over the place—we can detect that too. And we can make a mathematical correction to probably grade a bit to adjust the fact that all five of your graders are harsh. In our internal pilots—so not yet pushed to the website, just internal pilots—we will show that by making these mathematical adjustments, understanding who's tough, who's easy, who's random, we can get even more accurate grades. This is not deployed yet, but it's the sort of thing we can do with data from, say, 50,000 students in a class that generate 200,000 submissions.

好了，安德鲁现在得去上课了。请大家和我一起再次感谢他。谢谢。再见。

All right, Andrew has to run to teach a class now. So please join me and thank him once again. Thank you. See you later.

好的。

Okay.