MOOCs, Robots, and the Secret of Life

For the last two years, MOOCs have dominated the national conversation on technology and the economics of higher education. But for all the talk of whether they’ll usher in a new age of democratized global learning or destroy higher learning as we know it (or possibly both at the same time), it’s been hard to get a handle on MOOCs are, and what they can be. A lot of MOOC journalism has been like this, wherein a general-interest magazine writer signed up for 11 courses, finished one of them (the easiest, apparently), and formed his opinions accordingly. On the theory that to understand an educational experience you should actually experience it, I’ve spent the last four months taking two MOOCs. Now I’m done, and this is what I learned.

The first was Introduction to Philosophy, from Coursera (also the one class MOOC dropout guy finished, coincidentally.) It’s a nice, friendly, seven-week overview of major philosophical concepts, with each week’s lecture led by a different professor from the University of Edinburgh’s philosophy department. It was fun, and I learned some things. It was not, however, the equivalent of a legitimate college course. And to be clear, it didn’t pretend to be. The expected to workload was listed as “1-2 hours/week” and even granting the many problems with equating time and learning, that’s a clear signal the class isn’t something people should be getting three college credits for.

And beyond the brevity, Introduction to Philosophy was missing important things. Grossly simplified, there are two main components of an educational process. One involves making choices about knowledge, ideas, and skills. What do we want students to learn? How do we present that information? There are many ways to go about this, and Introduction to Philosophy chose two time-honored methods: lectures, presented on video, and supplemental reading. But the course was mostly missing the second component: creating a process and environment in which students make meaning out of the information you’ve presented, integrating it into prior knowledge and larger concepts in a way that allows for applications to problem-solving and transfer to other domains. There are plenty of well-known ways to do this, too, and they took a stab at one of them, assigning an optional peer-graded essay at the end of the class. But for the most part, the education was passive.

My second MOOC experience was very different. Having sampled Coursera (and a Udacity statistics class last year), I wanted to try something from the third major MOOC player, edX. I asked edX president Anant Agarwal to recommend one, and he suggested MIT 7.00x, an introductory biology course that was starting in a few weeks. It would show the cool things the edX learning platform can do, he said, plus the professor is great. So I logged on and in roughly two minutes I was enrolled.

The most important thing to understand about 7.00x is that it is a very close translation of a real MIT course. Unlike most colleges and universities, MIT has an actual undergraduate curriculum, called the General Institute Requirements. All students regardless of major are required to take six core courses: Single Variable Calculus, Multivariable Calculus, Classical Mechanics, Electricity and Magnetism, Principles of Chemical Science, and Introduction to Biology. There are a few options in the sense that they offer an extra-hard version of Chemistry for students who spent their high school years knee-deep in test tubes, but otherwise there’s no testing out with AP credits or otherwise avoiding these courses. MIT decides what you need to learn to get an MIT degree. (Because people at MIT are endearingly rational, they number everything, including their buildings and majors. A biology major will tell you they’re taking “Course 7,” a chemical engineering major “Course 10.” Thus, Intro to Biology is course 7.01 (with variants like 7.012 and 7.013), Cell Biology is 7.06, and so on. Thus, they named the 7.012 equivalent 7.00x.) The courses are hard enough that, even given the kind of people who are admitted to MIT, the university has a policy prohibiting freshmen from taking more than four of them in one term.

The educational model for the courses is pretty straightforward: Students attend lectures twice a week along with several hundred peers, and read supplementary materials. That’s the information part. Then they’re assigned problem sets, or “psets,” to complete and turn in for grades. The problem sets are extensive and difficult and that’s where most of the real learning occurs. As one MIT student told me, “learning science is about spending hours banging away at something until you get it right.” There are weekly course sections where students meet with a T.A. and then a couple of midterms and a final. It’s a very traditional approach, pedagogically. Where MIT distinguishes itself is in making the lectures very good, the problem sets challenging, the examination standards high, and the curriculum expansive in scope.

So when it came time to put Introduction to Biology on the edX platform, MIT simply took the course it already teaches its freshmen and reproduced it as faithfully as possible.

7.00x is nothing like the AP Biology class I took in high school back in the ‘80s and many students still take today. At MIT they assume you know all of that already and dive right into the good stuff: genetics and the mysteries of life. The course is taught by a professor named Eric Lander, who is like a walking advertisement for the American educational meritocracy: grew up in a working class neighborhood in Brooklyn, tested into Stuyvesant High School, international math olympiad, Princeton valedictorian at 20, Rhodes scholar, went from higher mathematics to business economics to genetics, and ultimately led the Human Genome Project. Now he co-chairs President Obama’s council of science advisors and runs a big interdisciplinary research center at MIT.

Unlike many brilliant research scientists, Lander is also a really good teacher. He has a knack for connecting with students, even in a 100-person room. As a rule, I’m a reader -- I usually find listening to people explain things to be frustrating, unless I can interrupt them and ask questions. But Lander’s lectures are fantastic. He has a terrific sense of narrative, informed perhaps by the legendary Princeton course in nonfiction writing he took from John McPhee. The course is a long, engrossing story of discovery, during which students learn the fundamental principles and intellectual unification of biology, which Lander represents as a triangle, the course’s “coat of arms”:

So we started with four hour-long lectures on biochemistry, beginning with the composition of cells and moving quickly through various types of molecular bonding, i.e. covalent, hydrogen, ionic, etc. and how they explain the formation of lipids, phospholipids, and high-energy molecules. Then we moved to the bottom right-hand corner of the triangle and got into into protein structures, specifically the various interesting and complicated ways that amino acids behave and interact.

Next we covered enzymes and biochemical reactions, how molecules move through transition states of various energy profiles and how enzymes help them along. All of this comes together in a lengthy explanation of biochemical pathways with the elaborate glycolysis process as the main example.

The lectures are presented on the edX LMS (learning management system) and look like this:

You can download the video or stream it, watch it in full-screen HD on your laptop or tablet, or in a smaller window (pictured) with a text version scrolling alongside it. What’s cool is that if you want to pause and go back to listen to a point again, you can scroll back up through the text to wherever you want to start, click there, and the video will back up accordingly.

The videos themselves were of Lander teaching MIT 7.012 to a group of 100 students, mostly freshmen. The 7.00x course schedule ran about a month behind the on-campus class, to allow time for video production and transcription, but otherwise it followed exactly the same schedule. In addition to the lecture videos, every week featured supplemental “deep dive” videos where T.A.’s and grad students explored key concepts in depth.

It all made a lot of sense as Lander explained it, but of course you don’t really know you’ve learned until you hit the problem sets. Many of the big questions surrounding MOOCs lies with assessment, formative and summative. To assess at large scale and low cost -- 7.00x initially enrolled some 40,000 people and like all edX courses, it’s free--you have to assess with machines. This is often taken to mean simple multiple choice and true / false questions, causing people to question how rich the MOOC experience can be.

But as Agarwal promised, 7.00x took great advantage of available technological tools. So the first problem set opens up a little “Molecule Editor” widget in Java, gives you a question like this, and you manipulate the chemical structure accordingly:

The week two problem set involved 3D protein models that you can rotate around, zooming in and out too see how exactly the various amino acids fit together:

There was also an interactive map of the glycolysis pathway, and a set of problems that involved submitting work done with the Foldit protein-folding simulator:

It was fascinating, and kind of exhausting. MIT offers the following advice to its undergraduates:

Chemistry Preparation for Biology.

You will need some chemistry background to be successful in 7.01x. Before enrolling in one of the Biology classes, you should either have a strong high school chemistry background (e.g., score of 5 on AP Chemistry), or you should complete one of the MIT core chemistry subjects (5.111, or 5.112 or 3.091) before enrolling in Introductory Biology.

The last time I took chemistry was in 1986, when I was a sophomore in high school. I got a B-minus and they bumped me off the Honors science track the following year. I did well enough in AP Bio as as senior, but that was the last science class I took until now. (Taking Astronomy 101 pass/fail to fulfill my science distribution requirement in college doesn’t count.)

Taking 7.00x was a great reminder of the difference between expert and non-expert learning. When you spend years working in a particular field, like higher education policy, it becomes easy to integrate new information into the complicated structures of knowledge and theory you’ve built in your mind over time. With 7.00x, I was starting from scratch. I could almost feel a new part of my brain being partitioned off to hold all the ideas and data. Accessing the partition was always a struggle; every time I sat down to watch a new lecture or work on a problem set it was like slowly pushing open a heavy set of doors.

Biochemistry finished the right side of the triangle. Next we moved to the left side, genetics, into the more familiar territory of Mendel’s laws, and how early 20th century scientists like Alfred Sturtevant developed linkage mapping techniques to locate genes on particular chromosomes. This led to the various dimensions of genetics: dominant, recessive, and X-linked inheritance, using fruit flies and yeast to hunt for mutants and reason toward the particular characteristics of genes. Here again the availability of technological tools made the learning process really interesting. In one problem set we had to breed multiple generations of fruit flies in a simulator and submit “cages” containing, say, 1,000 drosophila showing a particular statistical distribution of characteristics as evidence of the underlying genetic inheritance patterns.

The problem sets were such that you had a limited number of tries to earn points, after which a detailed explanation of the right answers would become available. The first mid-term, by contrast, gave you only one chance at each problem and that was it. This is a little nerve-racking because the questions are scored instantly, so if you’re bombing you know right away. I stumbled a little on the biochemistry but rebounded on genetics and managed to score 90 percent.

Next we moved to the base of the triangle, the third side connecting genetics to biochemistry: molecular biology. This took five lectures, as Lander described the great race to develop what would become the “central dogma”: DNA to RNA to protein, genetic information leading to the creation of biological function. This involved more vexing chemistry as we learned exactly how the amino acids in DNA molecules fit together, how Linus Pauling raced with Crick and Watson to discover the double helix, the intricate dance of transcription and translation, mismatch detection and repair. In these problem sets we used another program called the “Integrated Gene Viewer” to see what happens when mutations alter single base pairs in a sequence that can run hundreds of millions long, resulting in a new set of instructions for protein creation and sometimes-disastrous consequences for the organism in question.

In the 14th lecture there’s a great moment where Lander explains the molecular basis for the recessive nature of a certain sickle-cell anemia phenotype. A particular mutation changes the protein sequence in a way that creates hemoglobin molecules that, per what we learned about molecular bonding, form together into long rods that deform red blood cells into sickle shapes that jam and clog as they flow through veins. If people have the mutation on one allele but not the other, only half the hemoglobin molecules will stick together this way and the rods don’t become long enough to deform. But if both parents contribute the mutation, people inherit the disease.

Then we got to the really cool stuff: recombinant DNA, the circle in the middle of the triangle that connects all three vertices to one another, running in both directions.

This involves the mechanics of cloning, using restriction enzymes to cut out specific sections of DNA, paste them into plasmids, and grow them on petri plates, a kind of biochemical purification of DNA (and also the reason, I now know, that you collect plasmids to get superpowers in “Bioshock.”) Different kinds of cloning techniques allow different ways to find particular genes, and before you know it you’re into electrophoresis, DNA sequencing, and every episode of every police procedural broadcast on CBS since the late 1990s. I make a stupid mistake about the way DNA sequences are read biochemically and tripped up on the very last question, about different methods of regulating the argB gene, but pulled through with an 89% on Midterm 2.

All of which brought us to the climax of the course: Genomics, the completion of the triangle, and the Human Genome Project, which Lander himself led. We learned how in the years since, scientists have been able to map the origin of species by documenting specific genetic sequences common among multiple organisms following a chain of evolution back millions of years. How the process of medical discovery has become increasingly rationalized through large-population genetic analysis, yielding new insights into the nature of things like mental illness, some of which were literally published while the 7.00x class itself was underway. Heart disease, cancer, all being steadily transformed by new knowledge of the secrets of life. Function to Gene to Protein and back through recombinant DNA to mutation and analysis of function. The triangle was complete.

Four days ago, I completed the final exam. I got stuck once again on biochemistry, and finished with an 85 percent. Combined with the two midterms and the problem sets, that yielded a final score of 87 percent, which you can see, along with a complete set of my class notes, the syllabus and course schedule, and some additional problem set examples, at the 7.00x “cert” I’ve built at Accredible.com, here.

So, what does this all mean for MOOCs?

First, the experience was a welcome reminder that real education is hard work. I spent about 15 hours a week for 15 straight weeks on 7.00x. And they were hard hours, full of careful note-taking and intense concentration and more than the occasional furrowed brow. 7.00x required, easily, 100 times more mental effort than Coursera’s Introduction to Philosophy course. With the exception of writing my master’s thesis, I don’t think I’ve worked longer or harder in any “regular” college course I’ve taken in my life.

Which is not surprising. The average amount of time spent per week by full-time undergraduates on academic pursuits fell from 40 hours to 27 hours over the last four decades, according to this research. Only 13 percent of students report spending more than 20 hours a week studying outside of class, and a quarter worked less than five hours. For many students, full-time college has become a part-time job.

This is ultimately the fault of colleges and universities failing to enforce meaningful academic standards. Typical students will figure out the amount of work needed to earn the grade they want in a course and act accordingly. If passing grades can be had with five hours a week of studying for a full-time courseload, colleges are to blame.

Granted, the amount of time I spent on 7.00x was partly by design: I wanted to take a class where my academic and professional experience wouldn’t give me any built-in advantage. I suspect the abnormally smart 19-year olds I watched ask Dr. Lander questions all semester were able to finish more quickly. That said, there’s a reason MIT only lets them take four of these things per semester, despite the fact that they’ve been selected from the tiniest top fraction of all science-adept students, and they have all kinds of time and youthful energy at their disposal.

The point being, there is doubtless a lot of variation in the academic quality and rigor of MOOCs--but not more so, I suspect, than among the population of official, credit-bearing courses at accredited colleges and universities.

The important question is: What’s the best possible MOOC, given the parameters of an entirely technology-driven educational environment with a zero or very low marginal cost of delivery at scale? How good can a free online course be?

The answer, based on my 7.00x experience, is very good -- better, in fact, that almost anyone wants to admit. To understand why, it helps to examine all the different things that go into a course and the extent which they can be translated, adapted, or delivered in a MOOC environment.  

We tend to think of college courses in terms of content and action -- these books required, this lecture delivered, that exam given. But the underlying structure connecting all of those things is educational design. A good course reflects countless decisions, from the shape and sequence of the syllabus to the nature of the assessments to the minute-by-minute flow of the lecture.

The lecture itself seems entirely replicable with technology. Yet I’ve heard a surprising number of people argue this point, suggesting that there’s something about being in the room that mere video can’t replicate. So I decided to see for myself. In April, made the trip up to Cambridge and attended what turned out to be Lecture 24 of 7.00x, live and in person. The room I’d been watching remotely for months turns out to be located inside a huge new Brain and Cognitive Sciences Complex (“Building 46”). I arrived early and settled into a seat in the back left, notebook and pen ready. Based on what followed, I can say this: live and taped lectures really aren’t the same. Live lectures are definitely worse.

It’s a banal point to make, but there’s a lot to be said for the “pause” button. A big part of taking notes for the class involved drawing fairly complex diagrams and stopping to write out certain key points and ideas, which is hard to do if you’re also trying to keep up with the next point the lecturer is making. Lander packs a lot into an hour, with barely a wasted word. The videos are shot by multiple cameramen with professional equipment in great audio and full HD. The lecture hall only seats 100 people, yet sitting near the back it was harder to see, harder to hear, and more distracting than watching at home. The kid with pink hair next to me kept fiddling with his iPhone and was clearly bored. I much prefer sitting down to watch a lecture at a time and place of my choosing, headphones on, notebook in hand.

A friend of mine recently described the opposite experience -- after attending a live lecture for most of a semester, she missed class, watched the tape, and kind of hated it. It just didn’t seem the same. Which seems perfectly plausible to me. I got used to watching the lecture one way, she another. The point is that neither is inherently better and, again, don’t undervalue the “pause.”

(Side note: The class involved no textbooks, other than a free online version of a textbook that could only be used as a searchable reference, like Wikipedia. But of course Wikipedia already exists. Textbooks are doomed.)

The problem sets are easier to compare, since there’s no comparison -- the MIT undergrads were given exactly the same problems to solve. The tests, too, were in close parallel, although their exams were on paper while ours were on the computer. So: syllabus, lecture, problem sets, exam--that is, design, information, knowledge formation, and assessment--all exactly the same, and all of those components entirely translatable to the online platform. What, exactly, is left?

I spent some time on the phone with a half-dozen MIT students to find out. I was worried that I might be missing something, a crucial distinction between their experience in 7.012 as MIT undergrads and mine as an online student. But for the most part, there wasn’t. They went about the course just like I did: watch the lectures, do the problem sets, take the tests. To be sure, there were some differences. When they got stuck on problems, some would ask their fellow students for help. There was also the weekly discussion section where they could talk to their T.A.

Of course, the single most important thing to happen on the Internet over the last decade is the connection of people to other people. Two 7.00x Facebook groups spontaneously appeared as soon as the class started, and the LMS included discussion forums where students could pose questions to one another that were sometimes answered by TAs. I didn’t use the forums all that much. More interpersonal interaction might have saved me some frustration, but also the hard work of going back through concepts until I really understood the problems to be solved. I was reminded of Arum and Roksa’s findings on the value of solo studying. Sometimes there’s nothing to be done but bang away until you have it right.

Finally, there was Dr. Lander himself. At the end of the lecture I attended in person, a group of ten or so students gathered around him in the well of the lecture hall to chat and ask more questions. Lander is an optimistic, garrulous person, and a number of his students clearly had an intellectual crush. The same was true for the online community. From a student in Prague: “I feel almost like prof. Lander is a part of my family because I see him often on my computer.”  From Baguio City in the Philippines: “I want Prof. LANDER to teach me anything every week for the rest of my life...”

I understood the feeling, and part of me was a little jealous that I couldn’t join the after-lecture colloquy. But not so much that I would have paid $5,000 in tuition (roughly the MIT sticker price) for the privilege. Whatever the value of interpersonal relationships with TAs and the professor might have been, I was able to score 87 percent without them. And if even if I had been willing to pay $5,000, MIT can’t and won’t take my money, partly for reasons of sorting and selectivity and institutional prestige, but also because of pure logistics. There’s only person who led the Human Genome Project and the number of people who can form a group around him after class can’t ever be more than ten.

All of which makes me think that much of the MOOC discussion is badly mis-framed. I was debating Arizona State University president Michael Crow the other day and he said he didn’t want a future where “the rich get face-to-face with professors and everyone else will be taught by some type of robot.” It’s a good line and produced a lot of nodding heads. But stop for a moment and think about how much of Eric Lander made the translation from 7.012 to 7.00x. All of his educational design is the online class, all of his great lectures, all of the wisdom he brought to bear on constructing the tests and problem sets and supervising those who helped him. I believe the vast majority of what Lander ultimately brings to Introduction to Biology - The Secrets of Life is represented in 7.00x. If that makes it a Robot Lander, so be it -- Robot Lander is pretty great, and Robot Lander didn’t charge the 40,000 students who enrolled a dime.

Indeed, much of the university-based discussion of information technology suffers from a false assumption: people in universities assume the university, and then proceed accordingly. So it’s become popular to characterize MOOCs as nothing more than souped-up textbooks, another resource that professors can use to enhance the learning experience they provide to students while everything else stays the same, except for tuition, which keeps going up.  

That’s the wrong way to look at things. Now that we can build Robot Lander, the single most important question facing higher education is not: How can technology improve the cost and quality of the education that people provide to students? It’s: How can people improve the cost and quality of the education that Robot Lander and his ilk provide to students? The burden of proof is no longer on technology to show that it can make traditional higher education better in a way that’s worth the price to students and taxpayers. It’s the other way around.

Once you look at that way, some of the latest research appears in a different light. Consider the well-known Bowen, Chingos et al study in which:

“...we measure the effect on learning outcomes of a prototypical interactive learning online (ILO) statistics course by randomly assigning students on six public university campuses to take the course in a hybrid format (with machine-guided instruction accompanied by one hour of face-to-face instruction each week) or a traditional format (as it is usually offered by their campus, typically with 3-4 hours of face-to-face instruction each week). We find that learning outcomes are essentially the same—that students in the hybrid format "pay no price” for this mode of instruction in terms of pass rates, final exam scores, and performance on a standardized assessment of statistical literacy. These zero-difference coefficients are precisely estimated.”

One way of seeing these results is in terms of loss--the amount of face-to-face instruction per week was reduced from 3-4 hours to one hour, with no price paid in terms of pass rates, final exams scores, and an external assessment of statistical literacy. In other words, the treatment was less instruction, and it had no effect.

But that’s not right -- the students in the treatment group were learning with the Carnegie Mellon Open Learning Initiative (OLI) statistics course, a highly-designed online educational experience based on decades of research in cognitive science. State-of-the-art educational robotics, circa 2013. One hour a week of instruction was then added on top of that, and the results were indistinguishable from a traditional on-campus course. To know this for sure, the study would have needed another treatment group comprised of students who took the OLI course with zero hours of additional instruction. But I suspect what we’re actually seeing is a case where the treatment of adding live human instruction to the robot had no effect. Because OLI has in fact compared the robot-only approach to live instruction in statistics, and found no difference.

It’s also likely that the quality of the best future online courses will improve faster than the typical human instructor--that comparison being the appropriate one because robot courses are much more replicable and scalable than in-person courses. In his final lecture, Lander noted that 7.00x is just a first experiment: “I’m sure that when we look back on this five years from now, we’ll laugh at how primitive it was,” he said. “But that’s what science is. You run experiments, and you get better.”

As tens of thousands of students participate in discussion forums, for example, one can imagine identifying the most commonly-asked questions and recording different variants of responses from Lander himself. Instead of a single problem set, the problems could change based on whether people get previous questions right or wrong. The OLI courses already do this through a process of “adaptive learning.” Future Robot Lander will move closer to passing the Turing test. The T-1000 model, in other words, is going to knock the heck out of the T-800. And five years is not so far away.

Does this mean that live human instruction will eventually become obsolete? Of course not. Educational needs and contexts vary tremendously. Some courses will be more easily translatable than other, some intellectual pursuits more amenable to technology-based learning. The question is: how much of what’s currently done by people can be done better for less money using technology, in what circumstances, and for who? The smart defenders of traditional higher education understand this well, which is why we already see them retreating to the most defensible possible ground. Take for example Andrew Delbanco’s recent MOOC critique in The New Republic, in which he concedes pretty much all the relevant points before concluding:

“Back in the mid-twentieth century, the Ford Foundation report on “telecourses” asked the key question about technology and education: “How effective is this instruction?” When I came upon that sentence, it put me in mind of something Ralph Waldo Emerson wrote a long time ago. “Truly speaking,” he said, “it is not instruction, but provocation, that I can receive from another soul.”I first understood this distinction during my own student days, while struggling with the theologian Jonathan Edwards’s predestinarian view of life. Toward the end of the course, my teacher, the scholar of American religion Alan Heimert, looked me in the eye and asked: “What is it that bothers you about Edwards? Is it that he’s so hard on self-deception?” This was more than instruction; it was a true provocation. It came from a teacher who listened closely to his students and tried to grasp who they were and who they were trying to become. He knew the difference between knowledge and information. He understood education in the Socratic sense, as a quest for self-knowledge.

Nearly 40 years later, in my own course on American literature, one of my gifted teaching assistants received an e-mail from a student after a discussion on Emerson:

Hi, I just wanted to let you know that our section meeting tonight had a really profound effect on me. ... [T]he way you spoke and the energy our class had really moved me. ...I walked the whole way home staring at the sky, a probably unsafe decision, but a worthwhile one nonetheless. I actually cannot wait for next week's class just so I can dive even further into this. So I just wanted to send you a quick message thanking you, letting you know that this fifty minutes of class has undeniably affected the rest of my life. ... [S]ome fire was lit within me tonight, and I guess I'm blowing the smoke towards you a little bit.

No matter how anxious today’s students may be about gaining this or that competence in a ferociously competitive world, many still crave the enlargement of heart as well as mind that is the gift of true education. It’s hard for me to believe that this kind of experience can happen without face-to-face teaching and the physical presence of other students.”

Nathan Heller’s recent MOOC piece in The New Yorker ends much the same way:

Their discussion left an energetic silence in the room, a feeling of wet paint being laid on canvas. Sitting there, I thought of similarly fragile, unexpected moments that together helped define my college education. Once, during a bout of warm midwinter weather, a teacher of Baroque chorale harmony pranced into the room and spent half of class analyzing a song lingering in his mind; the song was “June in January,” and today it puts me in mind of open windows and warm Fridays. I can still see the faculty office—small, dimly lighted, chilly—where I sat as a freshman, having come with a question, as the professor, a charcoal-gray scarf looped around her neck, mentioned a document trove that became the basis for my senior thesis. I recall being in a survey-course lecture so slick and wrong-headed that, at one point, the woman sitting next to me reached over and wrote ugh-get-me-out-of-here comments in the margin of my notebook. And that she used a blue-ink Uni-ball Vision. And that the seconds while she wrote each note were bliss. I remember those moments, and I remember more. I’ve seen things you people wouldn’t believe.

In both cases I say: Sure! But if Ivy League humanities classes are all the defenders of traditional higher education have to offer (both writers are Harvard graduates) that’s conceding an awful lot. I myself had a few of those magical interpersonal experiences in college, and remember them well. I don’t remember most of the rest, because they didn’t involve ineffable moments of inspiration. They involved days and weeks and months of courses that mostly involved professors talking and me listening, taking notes. The guy who taught me Emerson, my 19th Century American lit professor, didn’t look me in the eye and ask me anything. Instead, he told my class that it would be inappropriate for us to ask questions while he talked. The used copies of Hawthorne and Melville I bought at the campus bookstore already had the relevant passages from the lecture underlined. The problem wasn’t that he was just a flesh-and-blood-robot. It’s that he was a bad robot. And he is not alone.

The question, then, is how much of the vast expanse of what currently comprises higher education can be taught using a technological foundation, at a higher level of quality than what students currently experience, for less money. Not all of it, certainly. But a lot more than people realize or want to admit, and the percentage is only going to grow over time. The more it happens, the more the underlying economics of higher learning will change.

That’s going to mean disruption and dislocation and pain for a lot of people. But the upside really is as exciting as the most unabashed MOOC enthusiasts like to say. The 7.00x students roster is a cross-section of humanity that exceeds a diversity-focused admissions officer’s wildest dreams: doctors and medical students from South America, a group of high schoolers in Greece, a 72-year-old retired chemist living the Netherlands, a full-time homemaker in India, a Sri Lankan college dropout, a high school biology teacher in Oregon. A young woman who says, “My dad is letting me take this instead of my regular 8th grade science. I am 13 years old.” A Ukrainian software engineer, a nurse in the Philippines.

A year ago, Eric Lander could teach no more than the number of people who could fit inside a lecture hall in Cambridge, Massachusetts. Now he can teach anyone in the world with a computer, and Internet connection, and the will to learn. That’s what MOOCs mean.