Interview- Daphne Koller - Building Coursera

Our company’s mission is to educate the world. We believe that education is the great equalizer and that many of the world’s problems could be made a lot better if more people had access to education, and that includes problems like unemployment, like hunger, even extremism, and if you educate people you open doors for them, to make life better for themselves, for their families, and for the communities. For me this project started about four years ago when a group of us who had been award-winning faculty were invited by John Bravman who was then the Vice Provost for undergraduate education at Stanford and he invited us to talk about how to open up more room in the curriculum for a meaningful engagement between faculty and students. And there were a bunch of us in the room and ideas were thrown around. Nothing particularly gelled. And then two months later I happened to be at a Google Faculty Summit, listening to a talk about YouTube. And all of a sudden it came to me that instead of lecturing in the classroom, the same lecture that I’d been giving for fifteen years, telling the same jokes at the same time, maybe we could take that, and preserve it, and in fact make it available to students in a much more engaging format that has more interaction than you get in a classroom. So I went to, back to John Bravman and he got very enthusiastic and he took me to see President John Hennessey of Stanford, who also was very excited and in fact supported this project with a generous seed grant in order to try and build up the technology as well as the pedagogy for making this happen. And so we piloted this in my class and developed some of the ideas that ended up making their way into Coursera, including the short video chunks, the integration of assessments into the video stream, the significant number of autograded assessments that allow the students to test their knowledge of the material. And, and so that was part of the project. And then in parallel to that, my colleague and co-founder Andrew Ng was actually working on a different trajectory, which is let’s teach the world, and ended up developing pieces of the design and technology and pedagogy independently in the context of this other project. And at some point we realized that the projects were remarkably well aligned because the same ideas, the same designs, the same contents could be used for both improving the quality of the instruction for our on-campus students and for offering a meaningful course experience for the world. And so that realization culminated in the fall of 2011 with the three large Stanford classes that Stanford opened to the world, the machine learning class, the database class, and the AI class. And each of those classes had an enrollment of 100,000 students or more, which was just a completely mind-blowing experience because no one expected that to have this level of uptake. And so in the fall, after we saw that this was happening, we realized that we needed to do something with this. We needed to build up on this success and do something even bigger and more impactful. And it seemed clear that in order to let this have as, the largest impact that it could possibly have in terms of teaching all these people around the world who would never have access to this kind of high-quality education, we needed to make this available not just to Stanford, but to a number of top institutions who could offer their great content to all of these people. And so that’s when we decided to spin this out of Stanford and make this, build up this platform that would provide this great experience for students as well as a good experience for faculty to effectively offer these large online classes that they could open to the world. But were you preparing for the fall of 2011 for like a year, or, or did you just? No. It was kind of of a bit of a last-minute decision to try out these classes in this format. And I don’t think anyone anticipated the extent to which this would take off and have this huge impact, and uptake in the world. So, really you kind of were laying down a lot of the early code in the fall. Yeah. So, what happened was in I think August is when the decision was made, and we realized that the platforms that we had been using inside Stanford were just not up to the task of being available, robustly, to hundreds of thousands of students. Or even tens of thousands, which is what we thought we would have at that point. And so we started to build up a platform from scratch using a group of graduate students and three undergraduates, just unbelievably talented individuals, who ended up also being the core group that would form the engineering team of Coursera down the line. One of the things that we tried to build into this project from the very beginning are opportunities for interaction of the students with the material, as well as interaction of the students with their peers. And there’s different ways in which that happens. In terms of interaction with the material, we do a lot of active exercises where the system basically checks the student’s work so that the student is told whether he or she are actually getting the material or not. And it’s important that those tests, those, those assessments, if you will, are deep and meaningful. And on the other hand, they also need to be graded at scale, because when you have a hundred thousand students you can’t have a TA or an instructor in the loop manually checking papers. And so we have all sorts of different things that we can check at scale. You know, multiple choice is the obvious one, but one can also check numerical answers. One can check, check math. One can autograde programming assignments, and by programming assignments that just don’t mean just code for computer sciences. An Excel spreadsheet is a programming assignment in effect too, and you can now grade a financial model or a marketing model as well. But none of those are the kinds of solutions that would allow us to support all of the classes that we would like, specifically when you want classes that require something that’s more critical thinking where a student would need to write down some argument for one thing versus another or discuss the arguments or the causal chain behind some historical event, or do a, a detailed analysis of a legal case or something. You cannot do that using autograding right now. Maybe if, maybe down the line, artificial intelligence will solve that problem for us, but I’m not counting on it in the near future. And so how do you do that? And that’s where, again, the social component of the platform, which is the other big piece of it, has come to our help because we ended up putting in place a peer-grading pipeline where students can robustly and reproducibly grade each other’s work. And it turns out that not only that is an effective way of grading at scale, it’s also a valuable learning experience for the students, and studies have not, not by us, but by educational researchers, have shown that peer rating teaches the students something, as well as provides a scalable solution to grading. One of the things that we try to build in is an opportunity for students to interact with each other in meaningful ways. And have one student help each other through the hard bits and have the students work together to achieve a better outcome for everyone. And so, for example, we had a question and answer forum where students really were the primary ones responsible for answering each other’s questions. And there was a real community built up around that where students felt incredibly motivated to help each other and answer each other’s questions to the point that in the fall quarter of 2011, the median response time for a question posted on the forum was 22 minutes, which is not a level of service that I, as an instructor, have ever offered my students. But because there was such a broad worldwide community of students all working together, there was, even if you were doing something at 3 o’clock in the morning, chances are that somewhere around the world, there would be somebody who was awake and thinking about the same problem. And so that social community was really an essential part, I think, of the student’s experience. And critical to the engagement and retention of students in the, in the course. I just imagine these people living in Zurich getting together in a coffee shop and as a study group. So what, what were some of your experiences, or others’ experiences on study groups? So it turned out that study groups were something that we hadn’t built into the platform originally, it’s something that we will do soon. But it grew organically. So people basically, without even being suggested to do that, ended up constructing study groups on the Q and A forums. So they said that, we have a study group in London, who wants to join us? We have an, a, a study group of, of Arab speakers or Hispanics or different types of groupings depending on, you know, geography or language or, or sometimes other things. We have, in one class, a global study group, which is specifically people who are looking to connect to people who are not part of their local culture and geography. And so this was really a fun thing for the students and some of them met physically, those that had geographical proximity, and others just communicated in the virtual space, but this organic growth was just such a powerful thing for students that we really quickly realized that we needed to build this into the platform. You probably are so busy that you don’t have any time to think too far in the future, but I’m going to just sort of see. Where, what’s your dream? Our dream is that anyone around the world will have an Internet connection, maybe via mobile device, which seems to be the way to go in most developing countries right now because of the way the infrastructure is developed, or not developed until now. And that people will be able to learn the things that they care about. And some of these will be things that are pragmatic and will help them get a better job and make more money so they can support their families. And some of them will be just ways to expand one’s understanding and one’s imagination by learning amazing things that you didn’t know about because I think it’s unfortunate that for most of us learning ends when we finish high school or when we finish college and that’s the last time we learn anything. And wouldn’t it be really cool if we had access to a life-long learning experience that, where you could come and say, I think this is a really cool topic and I’d like to learn more about that. And you would have access to a really amazing course taught by the top scholar in his or her field that could teach that.