Posted by: holdenlee | December 12, 2015

Computer Science for the Masses, Robert Sedgewick

Robert Sedgewick, professor of computer science at Princeton, gave a talk on Thursday on “Computer Science for the Masses,” outlining his vision for education in the future. He used as example his experience in turning the introductory CS course and algorithms course at Princeton into a MOOC and a “blended” course at Princeton. You can browse his courses on Coursera.

Notes I took during the talk are available here.

I’ll summarize some of the key points from his talk, reorganized for clarity. First, I’ll give some background, introduce the benefits of online education, and frame the question of how to blend it with traditional education. Next, I’ll give Sedgewick’s model of what an online course should look like: how to actually go about making one? Finally, I’ll outline his vision for the future of education. To distinguish my points from his, I’ll put my own thoughts in brackets.


When the CS department at Princeton was founded in 1992, the professors needed an introductory course. They decided that it must satisfy the following:

  • It must be geared at the entire student body because computer science was, and is, becoming more important for students outside of CS.
  • It will give students a model for programming, demystify how computers work (how are they built?), and teach fundamental concepts in the theory of computing.
  • Everything will be done in the context of applications. For example, the students’ 4th homework assignment is to write code to simulate a 3-body system, and they can modify it as they want to simulate planetary systems of their own creation. This is a prime example of the use of computing: to understand a system that is intractable to compute analytically.

In 2012, Robert Sedgewick and Kevin Wayne were asked to teach the intro CS course they had been teaching for many years, online. They were initially hesitant but embraced the opportunity.

Enrollment in computer science in Princeton is still rising rapidly. Now 40% of Princeton students enroll in the intro CS class, 25% in Algorithms (the second class), and 40% of computer science students are women. Sedgewick wants at least comparable numbers in universities around the nation, and estimates that the number of students who will need basic knowledge of computer science is 90%.

1 Why online education? Framing the question

The current system is unsustainable. In the last live lecture that Sedgewick gave, so many students were enrolled in the intro class that even the biggest lecture hall on campus was too small, and students in aisles were unable to see the screen. It was impossible to schedule a review session outside of class so that everyone can make it. Textbooks nowadays cost as much as $300+ to buy and $50 to rent, whereas online education is free and accessible.

[Much debate centers around, “should we keep traditional classroom instruction xor switch to online education?” rather than “what is the right way to create and use online education?” The first question is based off the false dilemma that we either have traditional classroom course, or fire our teachers and have an online course, but not both. The third alternative is that we use the best parts of each.]

In response to a New Yorker article saying that the road to knowledge must inevitably lead through dusty library shelves, Sedgewick frames the question this way:

“What is the most effective way to produce and disseminate knowledge with today’s technology? How can we best structure what we know and learn so that students, researchers, and scholars of the future can best understand the work of today’s researchers and scholars?”

Our endgoal is education, and any debates about business models, disruption, etc. must keep that goal in mind.

[There are several important ways in which online education offers order-of-magnitudes improvement. See my previous post and Tabarrok’s article for a list.]

Most important is the fact that online lectures eliminate redundancy. Thousands of professors prepare the same lecture on Quicksort every semester. But once a good lecture on it is put online, all students can watch that lecture online. Moreover, the online lecture will most likely be better because it’s been produced with a lot more preparation and deliberation. Note it is important how the online course is put together – slapping something together and putting it online will confer comparatively little benefit.

What’s good about the traditional university course? Here are some commonly given reasons.

  1. A traditional lecture allows the instructor to control the pace and direction of the class. It gathers all the students together and puts them on the same page with each other.
  2. Students aren’t motivated to learn the material themselves (e.g. from online lectures) if there isn’t a regular lecture. [Making alternative resources available means that students won’t go to lectures and hence won’t learn the material.]
  3. Interaction between teachers and students is important, because it means that students can learn in a more personalized fashion, and that learning is not passive. Students can ask questions during lecture. Teachers can ask students questions, lead class discussions, guide them through activities, etc. Students can pursue specific projects they’re interested in with the teacher’s guidance.
  4. [Note that what’s not a valid reason is that there’s something “sacred” about sitting in a large lecture hall or being in class; this reason has to be broken down into reasons like the ones above.]

Consider how we might achieve all these goals better in a blended model.

  1. A series of lectures online also gives a direction for the class, without locking everyone into the same pace. It’s not desirable to force everyone to go at the same pace: Beginners can play online lectures at a slower speed; advanced students can speed it up until it’s not boring. Students can stop and replay what they don’t understand. In a live lecture, there is only 1 speed, so the professor is only talking to a few people! What is good is to get students together in a room discussing the same topic- and this is a key part of the blended model.
  2. Sedgewick’s experience was that students did watch the online lectures, and enjoyed the flexibility in time that it gave them – some watched it first thing in the morning, or right before bed, or on the bus. The worry that students wouldn’t watch the lectures turned out to be unfounded. In fact, it made students enjoy the class more.
  3. There is actually very little personal interaction in a traditional large lecture class (and these are the kinds of classes that are the most important targets of MOOCification). Students play a very passive role in those classes. A quote attributed to Mark Twain goes,

    College is a place where a professor’s lecture notes go straight to the students’ lecture notes, without passing through the brains of either.

    By being able to control the speed, pause lectures, etc., students learn more actively. They don’t have to worry about copying everything down because they could always replay the material, and can focus on the learning.
    Very few students actually get to ask questions in a large course – but now there’s more time to ask questions with more small-group interactions, or on an online forum, etc.
    A blended course means that lectures are replaced by interactions in smaller groups, with more active learning such as discussions. (Sedgewick’s intro CS class includes 22 undergrad graders, 10 AIs (grad students), 5-6 faculty members.) Without the pressure of prepping and giving lectures, professors have more time to talk to students.
    [When people talk about the benefits of traditional college courses they are mostly referring to the higher-level, smaller-class-size discussion/project-oriented courses – and a blended model allows all courses to become more like this ideal.]

2 How do we do it?

Sedgewick has (co)produced 6 online courses and deployed 4 of them. Together, their online material totals to more than 70 lectures of 60-90 minutes and 1000+ slides, has reached over 1 million students.

Their distribution model is as follows:

  • Offer courses 2 times a year.
  • Each has an associated textbook.
  • Videos are bundled with the book/course.
  • Each textbook has a website.

Creating an online course takes an order of magnitude more effort than teaching a course (but less than writing a textbook). Even starting with lectures given many times before it still takes 50-100 hours of work per lecture.

What’s involved? According to Sedgewick, a good online course (“textbook of the future”) has 3 components, which he also calls “lasting abstractions.”

  1. Lectures: lectures introduce the material and inspire further study. An online lecture can simply be a recording of a live lecture with cameras from the back of the room, but nowadays lectures are “evolving to new standard of excellence.” Sedgewick’s online lectures are not blackboard talks, but “studio-produced” lectures. For example, in his online course on Analytic Combinatorics, Sedgewick has slides with drawings of combinatorial objects, produced using illustration software, which would be time-consuming and error-prone to reproduce on the board. He has special online lecture flows for “stories” and math derivations he calls “builds” [There are certain concepts/derivations with “moving parts” that are especially apt to be introduced through video.]
  2. Web content (a.k.a. booksite): This is a repository of additional resources related to the lesson, with around 10 times as much content as the book. The intent isn’t for students to go through all of it, but everyone can go through some of it depending on their interests. The repository
    • is integrated with web search,
    • always up to date (dynamic),
    • and includes content types (videos, applets, etc.) not available in print.
  3. Books: The abstraction of a book is that it includes what students can reasonably learn about the subject in one semester, and serves as a comprehensive reference for the course material. Textbooks are not obsolete, but are as important as ever. Even today, many students still prefer a paper book. The disadvantage of a book is that it needs to be written by a professor (or expert). The number of academic books has grown tremendously; more and more people are being empowered to write books.

3 What does the future look like, and what are the obstacles?

It’s an ongoing challenge to make education scalable. We need to exploit technology as much as we can, and continually improve things that are working and throw out things that are not working.

If online courses have so many benefits, why aren’t more faculty developing online courses?

Universities are providing only a small fraction of the resources needed. We need to embrace technology, provide training, support and real incentives for teaching and content creation, and invest in research on the interface of education and technology. We should develop a “teaching class” of professors who can teach blended classes.

In reality, many universities are reluctant to swap out traditional lectures for online lectures. Thus, so far individuals have been spearheading such development. The academic leadership needs to actively encourage usage and development of online courses. [The culture is such that a professor who wants to swap out the lectures ey give for online lectures by others may be seen as shirking responsibilities – even if the online lectures are better than those ey could give.]

Developing online education requires a lot of money, which means we need to reevaluate how universities are spending their money. Sedgewick suggests that a proportion of the money invested in the libraries can be used instead for developing online courses. In particular, print libraries are costly, and we can shift to a digital library architecture.

One of the big stumbling blocks has been a misplaced focus on the business aspects of online education.

Institutions are trying to take control and failing. Their focus has incorrectly been on monetization and control: how are universities going to profit from online education? How can they import the evaluation and credential system to this new platform? As a result, bad business models were created prematurely, and the real goal of teaching has taken a back seat. Our primary goal is to teach, not give credentials.

Content creation is the province of individuals. Sedgewick envisions that the future of online education will look like an app store, the main point being that individuals create and publish content onto a central platform and are rewarded. In his words,

“We pay individuals who create the content and the ones that are popular win. And that’s what we do in the internet age. We don’t make deals with institutions that take all the money and not give any money to the content creators. The record industry figured that out, and universities [need to as well].”

How would someone who’s inexperienced in teaching going to produce content in this new environment? The point is that they won’t have to. The online marketplace is itself a place for the next generations of teachers to develop their skills. [The smaller-group professor-student interactions that we have in a blended model are also a friendlier way for professors to gain teaching experience.]

If universities want to play a role in disseminating knowledge in the future, then they can’t afford not to invest in online education.


The future of education is a blended model: all common material will be available as professionally produced online lectures. Freed of the need to prepare and give these lectures, professors can focus energy on better interactions with students.

To achieve this future, we need top-down academic leadership that embraces and invests in online education. In particular, the end goal should always be to give the best education to students – this question is how we should evaluate every question (ex. should we allow professors to use other people’s online lectures?), rather than, for example, attachment to tradition.

One thing we need to realize is that educational content is created by individuals, not by universities as an institution. To end up with the best content, we need to incentivize content creation, and have an online “marketplace” where content can “compete” to students and the best ones can rise to the top.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s


%d bloggers like this: