John L. Hennessy: Risk Taker

Stanford University’s president predicts the death of the lecture hall as university education moves online

9 min read
Photo of John L. Hennessy
Photo: Gabriela Hasbun

John L. Hennessy

Photo: Gabriela Hasbun
2012 IEEE Medal of Honor Recipient John L. Hennessy

In the 1980s, John L. Hennessy, then a professor of electrical engineering at Stanford University, shook up the computer industry by taking the concepts of reduced instruction set computing (RISC) to the masses. Hennessy wrote papers, gave talks, designed chips, started companies, and even, literally, wrote the book (a textbook that’s still used today). The RISC architecture, which focused on simpler, lower-cost microprocessors, was then thought to be an academic exercise with little practical use; today it plays a major role in the industry.

Hennessy, now president of Stanford, is once again designing, testing, and advocating a new architecture, this time in the field of university education. He first began rethinking research at universities and recently began reimagining university education itself.

For these efforts, in June Hennessy will receive the 2012 IEEE Medal of Honor “for pioneering the RISC processor architecture and for leadership in computer engineering and higher education.”

IEEE Spectrum profiled Hennessy and his career as a computer architect and entrepreneur in “RISC Maker,” [November 2002]. This year, we checked in on Hennessy’s recent efforts to shake up higher education. Stanford has a long history in distance education, which in the 1990s moved from closed circuit TV to Internet delivery. More recently, the university explored offering online courses to a much larger audience with a programming class for iPhone applications, first available in 2009, that has been downloaded more than one million times. Since then, Stanford has been developing and testing tools for producing, distributing, and enabling social networking for online courses. This past fall, more than 100 000 students around the world took three engineering classes—Machine Learning, Introduction to Artificial Intelligence, and Introduction to Databases. Hennessy says that’s just the beginning. In fact, in his vision of the future, the lecture hall—those ubiquitous tiers of seats with fold-down writing arms, curving around a professor at a podium—will play a much smaller role.

But not everyone has applauded these educational experiments. One Stanford student who took the online machine learning class and blogged about it afterward said it didn’t match the rigor of traditional courses. He also worried that a proliferation of online classes would reduce the value of a Stanford degree. Other critics take the opposite tack and wonder if the courses are too hard for the general public, noting that the vast majority of students who signed up for the online classes dropped out partway through.

Spectrum visited Hennessy at his office on Stanford’s historic quad and asked him about his educational vision. Here’s what he had to say:

I’m a believer in online technology in education. I think we have learned enough about this to understand that it will be transformative. It’s going to change the world, and it’s going to change the way we think about education. Institutions like Stanford should be willing to fund the experiments, to try different things, to think about different models. We can do what other institutions would be strained financially to do, and they can learn from our experience.

It’s going to filter down into high school, too, where we have an even more dramatic problem, considering the shortage of highly qualified high school teachers, particularly in science and math.

For us, it started a few years ago. One of my faculty colleagues, Daphne Koller, said, “You know, I don’t feel very useful when I stand in front of a classroom and give a set of lectures, 85 percent of which are the same as the year before. It’s not very rewarding for the student, and sitting in large lecture halls is not the way students want to learn, particularly this generation.” She pointed out that the large lecture hall is not a good learning environment, and it’s not a good use of her time. And she was right. I agree that physical presence isn’t all it’s cracked up to be.

But it’s not really about what I think. The students are rewriting the rules for us. That large lecture hall with nice banked seating and 300 people sitting with their attention focused on somebody standing in the front of the classroom is a model that lasted for many years, but the students have made it clear that that’s not a model they find particularly attractive anymore.

Instead, this generation is completely comfortable watching a video online; for them, it’s not markedly different than having a person up at the front of the classroom. They are happy using technology. They know how to hit the pause button; they know how to speed it up a little bit, to watch it 20 percent faster and make the process more efficient.

John L. Hennessy

Photo: Gabriela Hasbun


Stanford’s endowment grew from US $8.9 billion in 2000 to $16.5 billion in 2011. Stanford received $7 billion in gifts during the fiscal years 2001 through 2011, including $1.1 billion in support of undergraduate education and $6.2 billion as part of the Stanford Challenge. These gifts fund more than 130 new endowed faculty appointments and more than 366 graduate fellowships and also provide $27 million in research seed grants.

During the past 12 years, Stanford has completed more than 100 major construction projects, including academic buildings, housing, and athletic facilities. Notable is the new Science and Engineering Quad, a group of buildings that to date includes the Jerry Yang and Akiko Yamazaki Environment and Energy Building ($118.8 million), the Jen‑Hsun Huang Engineering Center ($84.3 million), the Center for Nanoscale Science and Engineering ($77.5 million), and the Bioengineering and Chemical Engineering Building ($197.3 million). In 2011, the trustees approved more projects, including a $41.2 million computing facility and $438 million in improvements for campus energy systems.

We started using technology to improve what was happening on campus. And we learned. We learned that if a video is simply a talking head with a set of PowerPoint slides, it can be just as stupefying as sitting in a big classroom listening to a lecturer. So what you need is more interactive learning. When Daphne Koller started her experiment, she broke her lectures into chunks of 10 to 20 minutes and had a little miniquiz in between. And the miniquiz was able to identify that the student must have been asleep during that section or that there’s one little thing to brush up on before moving on—hit this link and you’ll go back to that place in the video.

We also learned that students like an online technology-mediated learning environment because it allows them to balance their lives. So if a student has a big project due for class XYZ this week, he may choose to fall behind one week of lectures in another course and then catch up on the weekend. Things like this can help with stress reduction, which is a big issue for many students.

And I think we learned that there is fun to be had with this online experiment. Last fall, a group of our faculty said, Well, why don’t we just put these courses online, let anybody who wants take them. And who would have guessed you’d have more than 100 000 people sign up? Nobody. Courses like Introduction to Machine Learning, Introduction to Artificial Intelligence, and Introduction to Database Systems are not of such obvious use as the iPhone programming course we put up a few years ago. That time, we could expect a lot of downloads because we knew people thought they’d soon be selling a widget in the App Store for a dollar a copy and making a lot of money. But Introduction to Artificial Intelligence isn’t a skills-based course like that; it’s a fundamental course. We were surprised how many students were willing to do the assignments and have them graded.

Of course, that creates a challenge, because it means that all assignments have to be graded automatically. There has been a lot of progress in grading technology. SAT writing exams [a standardized university admissions test], for example, are partially graded by computer right now. Each essay gets one grade from a computer and one from a person, with a second person looking at a test only if the scores differ significantly; logic courses have proof checking online. But at this point we don’t know how far we can push grading technology and what is the right mixture of online and off-line grading.

We found that we were able to handle a lot of the Q&A through social networking. It’s amazing that when you have 10 000 students in a class and a student puts up a question, the group quickly converges on the right answer: Several students put up an answer, other students come in and vote for what they think is the best answer, and there’s a high probability that you’ll converge to a pretty good answer in less than an hour. And if at the end of the two days there are three questions where the answers are not really clear, the instructor can come in and say, “This is the way you should think about this problem.”

Our experiments so far have raised questions we don’t know the answers to. How much should we invest in production values? Take Sal Khan, of the Khan Academy online learning effort. His videos are pretty simple in terms of the production qualities; there’s not a big investment. On the other hand, you go to anything that’s done on TV, even something like “Myth Busters,” and the investment in production is gigantic. Where on that spectrum do we really want to be?

One of the more interesting problems came up around problem sets. If you’re teaching an on-campus course with 40 or 50 students or 100 students, you can probably live with a little ambiguity in the questions because there’s usually a teaching assistant around. But if you’re teaching a course online with 10 000 students and there’s ambiguity in a homework question, it’s a nightmare.

Video courtesy of

While we have experience using online technologies for teaching classes of up to a few hundred Stanford students, the experience of teaching a course with thousands or tens of thousands of students is very new. We are still very much in the early experiments. Teaching a smaller class of students who have been screened by our admissions procedures tends to ensure greater consistency in their abilities and allows more adaptation and feedback from instructors. Clearly, a class with 10 000 students allows almost no individual interaction and presents new learning challenges. There are also many experiences at Stanford that we do not see how we can re-create online, like small seminar courses and hands-on project courses. For this reason, blended learning models may be very attractive.

It’s important to keep in mind the difference between learning and credentialing. Universities—and high schools, for that matter—do both: They assist students in learning, and they provide a credential as evidence that a student has mastered certain material. I’m convinced that online learning will be widely used: The millions of downloads a day experienced by the Khan Academy and the broad interest in Stanford’s online courses are clear evidence. But online credentialing is a new space and one that has much greater uncertainty. We have to consider a range of issues, from how credentials received online will be viewed by employers to how good online credentialing can be.

This can all add up to big changes in how we think about a university education. Could you imagine a situation where students do some of their degree work remotely and then do some on campus, particularly the small, experiential classes that don’t transfer well to online?

This could be a way of keeping costs down, because you’re living at home for a while or you’re working part-time while you’re doing the remote courses. So it could begin to address this issue of the cost of higher education. It’s hard to predict how much you could reduce the cost of instruction, but it’s important to understand that the cost of instructors is the No. 1 cost factor in education. If you could double the student–faculty ratio by reusing online material without reducing student learning, you could significantly reduce the cost of education.

The biggest way this could go wrong is if we assume that we can fully automate this, if we think that we can throw students in front of terminals, that we don’t need any live instructors anywhere, that students can be totally successful without ever talking to anybody. I think this is a model that will leave many students behind.

In a university, there is always a very small fraction of students who probably never need to come to class. They could just sit in their rooms, read the textbook, and they’re capable enough, focused enough, disciplined enough, and driven enough that they could be successful. But that’s a very small minority.

Likewise, there’s a small minority of students who could watch everything online, never talk to anybody else, never engage with an instructor, never engage with teaching assistants, and learn just fine. But again, that’s a very tiny minority.

The other potential problem is that we could see a lot of schlock out there. Because the Internet enables everybody to be a publisher, we could have a lot of things out there that are of lesser quality. Will people be able to distinguish quality in online education? This isn’t resolved.

We have to remember that the undergraduate experience is a whole experience. Part of that, particularly in the American higher education system, is the process of living and working with people of your own age. It involves learning how to work in teams, learning how to work with people with different backgrounds and different experiences. You certainly don’t want to dismiss that part in a pursuit of doing it all online.

In spite of these hazards, I think online education and the role of technology in education are going to be transformative. And I’d like to think that what we’ve started here will not only continue at Stanford but that other universities will see it as a way to organize education and to play a larger role in the world.

To do this, however, universities have to be willing to change. Universities build on tradition and history, but they also have to be dynamic. And I think that struggle to balance those two opposing forces—to not become too attached to the past in such a way that you can’t do something new, or to become too faddish in such a way that you lose your core values—is an ongoing challenge for all institutions.

But online education is going to happen; it’s not going to wipe everything else out, but it is going to happen. We have to embrace it.

A correction to this article was made on 27 April 2012.

This article is for IEEE members only. Join IEEE to access our full archive.

Join the world’s largest professional organization devoted to engineering and applied sciences and get access to all of Spectrum’s articles, podcasts, and special reports. Learn more →

If you're already an IEEE member, please sign in to continue reading.

Membership includes:

  • Get unlimited access to IEEE Spectrum content
  • Follow your favorite topics to create a personalized feed of IEEE Spectrum content
  • Save Spectrum articles to read later
  • Network with other technology professionals
  • Establish a professional profile
  • Create a group to share and collaborate on projects
  • Discover IEEE events and activities
  • Join and participate in discussions