Why Online Education Works

My essay at Cato Unbound, Why Online Education Works, goes beyond much of the recent discussion to give specific examples of how online teaching increases the productivity and quality of education. Here is one bit:

Dale Carnegie’s advice to “tell the audience what you’re going to say, say it; then tell them what you’ve said” makes sense for a live audience. If 20% of your students aren’t following the lecture, it’s natural to repeat some of the material so that you keep the whole audience involved and following your flow. But if you repeat whenever 20% of the audience doesn’t understand something, that means that 80% of the audience hear something twice that they only needed to hear once. Highly inefficient.

Carnegie’s advice is dead wrong for an online audience. Different medium, different messaging. In an online lecture it pays to be concise. Online, the student is in control and can choose when and what to repeat. The result is a big time-savings as students proceed as fast as their capabilities can take them, repeating only what they need to further their individual understanding.

More at the link including a discussion of how most of my teaching career happened in 15 minutes.

Responses from Siva Vaidhyanatha (Robertson Professor in Media Studies at the University of Virginia), Alan Ryan (former Warden of New College, Oxford) and Kevin Carey (Director of the education policy program at the New America Foundation) follow later this week.


'Highly inefficient.'

As is spoken/written language itself, with its roughly 50% redundancy, which might just suggest that efficiency may not be the best measure of communication, much less teaching.

See the Shannon-Weaver Model -

'The Shannon–Weaver model of communication has been called the "mother of all models."[1] It embodies the concepts of information source, message, transmitter, signal, channel, noise, receiver, information destination, probability of error, encoding, decoding, information rate, channel capacity, etc.

In 1948 Claude Elwood Shannon published A Mathematical Theory of Communication article in two parts in the July and October numbers of the Bell System Technical Journal.[2] In this fundamental work he used tools in probability theory, developed by Norbert Wiener, which were in their nascent stages of being applied to communication theory at that time. Shannon developed information entropy as a measure for the uncertainty in a message while essentially inventing what became known as the dominant form of "information theory."

The book co-authored with Warren Weaver, The Mathematical Theory of Communication, reprints Shannon's 1948 article and Weaver's popularization of it, which is accessible to the non-specialist.[3] Shannon's concepts were also popularized, subject to his own proofreading, in John Robinson Pierce's Symbols, Signals, and Noise.[4]

The term Shannon–Weaver model was widely adopted into social science fields such as education, organizational analysis, psychology, etc. In engineering and mathematics, Shannon's theory is used more literally and is referred to as Shannon theory, or information theory.[5]

Shannon's capacity formula applied to the additive white Gaussian noise channel gives the Shannon–Hartley formula,

C = W \log_{2} (1 + \tfrac{S}{N}),

where C is channel capacity measured in bits/second, W is the bandwidth in Hz, S is the signal level in watts across the bandwidth W, and N is the noise power in watts in the bandwidth W.'


This paper - A new look at phonological 'redundancy' from Gerry Abbot - explores a bit more in a linguistics context -

Redundancy occurs at all linguistic levels, from the phonetic to the syntactic; in total, according to Shannon and Weaver (1949),

the redundancy of English is just about fifty per cent, so that about half of
the letters or words we choose in writing or speaking2 are under our free
choice, and about half (although we are not ordinarily aware of it) are
really controlled by the statistical structure of the language. (p. 13 of
1972 edition)

Unfortunately I think only received about 5% of what you wrote.

read it again?

In the case of MRU, you have the flexibility to determine segment length and the time allotted to each topic. But the move to online education would presumably also affect industries where the length of units of instruction is fixed - for example, CLE and CME, where State Boards set requirements that include 1 hour of instruction to get 1 credit.

It seems this could result in a few outcomes: 1) lawyers and doctors become smarter if "material-per-minute" increases (on the assumption that online material won't cater to the 80-20 model); or 2) online education takes hold slowly in these industries, especially if one of its major potential benefits is the efficiency gain of fitting more material into less time; or 3) online education's general momentum will carry over into these industries, but the resulting material will simply replicate standard models where a talking head lectured for an hour; or 4) other.

Do you have any predictions about the trajectory of online ed, or about the evolution of the relevant regulatory bodies, in industries like these?

Exactly. To get the full benefits of online education you need to adapt a package. In particular because students can move at their own pace and learn more in less time, online is complementary to competency based testing. Thus,(#4) I expect movements in this direction.

Reports of the revolutionary nature of "online education" are greatly exaggerated -- also, what does an online education signal?

At a bare minimum, getting an online education signals two things.

First, that you're highly self-motivating. And second, that you are ready if an employer demands that you demonstrate any of the knowledge you claim to possess. He or she almost certainly will so test you in the interview, just to see if you are who you say you are.

"First, that you’re highly self-motivating."

Not necessarily. Someone might contrarily infer that you're too lazy/cheap to get to a real classroom.

Depends on the field. If you're a computer programmer, I'm much more impressed by your contribution to open source projects than whether you went to MIT.

In fact, there are companies popping up that make it easy to create a new kind of "resume" that incorporates online properties like Github, LinkedIn and Hacker News.

You don't really have learn anything in a CLE course, whether it is in a classroom or online. You can if you want, but it is not a requirement. I've never taken a test to demonstrate what I learned. Unless that changes, I doubt that they will ever move off the 1 credit/hour standard.

Don't you think a 1-hour CLE that offered superior educational benefits (by fitting more information into less time) would be inherently attractive to lawyers, whether or not testing was a part of the process? If that was the case, online CLEs might have more potential to disrupt the industry by rapidly capturing market share... at which point the industry might evolve to reflect the new platform. Or is the perception of CLEs as perfunctory and unhelpful so ingrained that lawyers wouldn't give two hoots if CLEs offered more bang for the buck (so to speak) through increased information density?

I thought that was a very well written article. I'm skeptical though of Alex's statement that "Online Education Has Already Met the Market Test". Has anyone succeeded in monetizing this model?

If MRU charged even $200-per-year (peanuts in comparison with the offline product) how would it's viewership change?

Yes, the person who has succeeded in monetizing the online models is...the professor who teaches the online course that uses his textbook.

Antioch University announced last month it would pay Coursera to license some courses for one of its bachelor's programs. Essentially, Antioch is buying permission to issue its own credit for the MOOC. The question is how large the market is/will be as people continue to test the pedagogical effectiveness of MOOCs.

As opposed to the move-the-hell-out-of-your-house university that sells their undersupplied housing, cafeterias, parking passes, lab fees, etc.

Don't forget the tailgates and frat parties.

We don't even know if the current model of education works. There is no evidence one way or the other. We don't know how to approve it based on any outcome (benefit to society, earning power, etc . . .) Don't think that I am a detractor of online education. We have come to a point where we need to embrace lifelong learning. In my case however, I found that education mostly got in the way of learning.

1. For students in the 18-22 year range: online has some value, BUT the problem with students in this age group is that they will likely quit...if they do not have some social interaction or network with their peers that offsets the unpleasantness of learning and working at learning. Ask yourself this question: Had you ever thought of dropping out of college but felt that OTHER students around you would think less of you and therefore didn't drop out. Ask yourself: Have you ever persevered on a difficult problem because OTHERS around you seemed to be able to do it. Did you ever meet with friends in a study group, not just studying, but also having a beer and talking. Now, ask yourself: Will a student, living at home, taking an online course, have the same reactions without the social support, or risk of social castigation, of his/her peers.

Before you commit to the all in online model, perhaps you want to test drop out rates, persistency, etc. I think people are missing the social unless they believe students are some bin that gets filled with something called knowledge that is dispensed by some guy/gal standing up on a stage.

2. Online has potential for doing what in person teaching does not, but so far has not offered that option. Ask yourself this question: how many online courses have you taken which follow the semester schedule, or mimic the amount and pace of coverage that would be in a semester course. I bet every one has.

That is, the online course has the same number of lectures--or covers the same amount of material--as a semester course. Why????

I would like to see, for example, a very SLOW linear algebra and optimization class...one which takes a year to cover, with examples up the wazoo, and with real world experiments in modeling taken from real world cases...and working it through. Because of time constraints, when I have taught antitrust, I cannot take the time to go through, for example, all the picky little details involved in defining and proving a market. There is no time to fashion questionairres, show how to use Census data, show how to develop tests, etc. There isn't time...but, with online, there is infinite time...but, not one which a student would take, but which a professional would take, over a long time.

3. Online should be tried for high school students for them to take refresher courses during the summer before they get to college. gee, maybe you can refresh yourself on math or learn a little on how to research and write a term paper....before you get to class.

Following up on point 1 above, that students will drop out, this is from the head of Coursera:

'Koller: This is what we see: Enrolling is easy. It's a matter of just clicking a button, and it's free. A lot of students do that. Then when it comes time for the course to actually begin, about 70% of students show up, possibly because those who do not attend have found their life has moved on, and they're now busy doing something else.

Of the ones who show up, we see a bifurcation of the population. There are students who come in and they are primarily there to watch the videos. Of those who start watching videos, about 30% to 40% watch the course to its completion, with a fairly constant drop off rate week-on-week as students get busy.

Then there is the second population: These are students who really do intend to take the course for real. We gauge that by seeing who submits the first assignment. If you submit the first assignment, you probably intend to take the class for real. The retention rate for this population is comparable: About 30% who start the first assignment will complete the last one.

In terms of what you might traditionally call "retention" -- that would be the number of students who submit the final assignment relative to the population who initially enrolled for the course. In those terms, we have a retention rate of 7% to 9%, depending on the course. But that's really the wrong way of looking at it because many of these students never really intended to seriously take the class in the first place."

see interview at http://knowledge.wharton.upenn.edu/article.cfm?articleid=3109

What's the benefit of pre-recorded lectures over just reading the textbook? In other words, why not do away with lectures all together?

Shh - we don't like talking about free public libraries on the Carnegie model here.

'A Carnegie library is a library built with money donated by Scottish-American businessman and philanthropist Andrew Carnegie. 2,509 Carnegie libraries were built between 1883 and 1929, including some belonging to public and university library systems. 1,689 were built in the United States, 660 in Britain and Ireland, 125 in Canada, and others in Australia, New Zealand, Serbia, the Caribbean, and Fiji.

Few towns that requested a grant and agreed to his terms were refused. When the last grant was made in 1919, there were 3,500 libraries in the United States, nearly half of them built with construction grants paid by Carnegie.'


That was a man that put his own fortune to use, and was proud to have his name attached to his attempt to proviode education. These days, the donations are as anonymous as possible, but then, the goals are a bit different than Carnegie's.

(Though one should note this fact in connection with women finally winning the right to vote in the U.S. -

'This coincided with the rise of women's clubs in the post-Civil War period, which were most responsible for organizing efforts to establish libraries, including long-term fundraising and lobbying within their communities to support operations and collections.[2] They led the establishment of 75-80 percent of the libraries in communities across the country.' )

My hometown in rural Georgia had a Carnegie library. Carnegie and his wife both attended the dedication. I believe it was one of the few she attended.

I have very fond memories of that library. Thanks for reminding me of it.

I have taken a number of Coursera courses, and I almost never watch the lectures. I read the lecture slides and then (very infrequently) watch the videos if I need clarification. But I hate (and have always hated) sitting through lectures, it's possible other people enjoy them more.

What about the externality of free online teaching: enhacing the quality of the course offered by offline prof who follow the online course ?

Case in point: Khan Academy

Online education can only "work" if we assume that giving people knowledge is useful. Arguably, somebody's knowledge is irrelevant to their career.

This is my first time I have visited this site. I found a lot of interesting stuff in your blog. From the tons of comments on your posts, I guess I am not the only one! keep up the great work., digital detox, 09318,

As you and Tyler are now in the business of putting out of business many (most?) of PhD-trained economists from GMU (who end up at teaching institutions, if I am not mistaken), would you kindly tell all of your current graduate students to go home and find something else to do? I don't see them doing so well in this brave new world you are creating.

In my mind the "threat" of online education is more of a threat to faculty at research universities than those at teaching institutions. A MOOC a good substitute for a 500 person lecture course, and much less of a substitute for a 30 person class. If online courses supplant physical courses, it will happen first to large lecture courses, and disrupt the cross subsidization of research at research universities.

What percentage of graduate students end up teaching college level?

I'd say about 50% (of PhD's) if it's a top school.

I don't like the analogy to movies and plays. On the one hand you can get the best actors and best directors relative to the average play but the experiences are different. Think of the problem of the symphony. No televised performance of a symphony no matter how great the conductor and performance comes close to the quality of a good symphony performance heard live. The decline of the symphony orchestra is partly **because** it's so relatively expensive, but that doesn't change the fact that the experience is BETTER for the live performance except one or two margins.

In the same way, I bet an above average classroom performance with Alex Tabarrok in which you can react to the questions or even facial expressions of the people in class is much more impressive than the best TED talk you might give. However, that might still not make it cost effective. But don't pretend that ubiquity of symphony recordings heard digitally on the average stereo is anywhere near comparable to the effect of a live orchestra. It's just that the cost differences are so overwhelming.

You could take the redundancy argument one step further of course; why the need for professors at all when there are textbooks from which the professor simply parrots? To me, this gets at the critical role of a teacher, which is not to be brutally efficient in information transmission, but to teach.

It might be good if it online teach lead us to separate education into teaching, tutoring and testing.

The last line in your article -- "not the least of which was the printed book" -- begs what I consider the biggest question. How is on-line learning different from or superior to reading? Is this a dramatic improvement or on the order of "more chalk"? For example, the difference cannot be "flipping" the classroom since this option has always been available: Read this chapter, article, etc. at home and we will discuss it in class.

Your play vs. movie analogy is telling. Certainly movies have the advantage of scale, and comparing star actors and professors is apt. But I think many Broadway aficionados will argue that even the best movie is a poor substitute for a good live performance. Given that tickers may cost 10 times as much or more, they clearly have some appeal. Nevertheless, I hope you are right about the potential.

What blink said.

Online education will be "revolutionary" to the extent, and ONLY to the extent, that it is a revolutionary advance beyond reading books. I think there are two groups of people for whom it is indeed: 1) people in the developing world, for whom books are too expensive, and 2) people whose only motive is certification/accreditation, which they cannot get from books alone.

But as for the 500 students crammed into the Chemistry 101 lecture hall at Big State U? Online education is not going to change that situation much more than textbooks ever changed that situation. Sure, online education offers some features that are useful to these students that can't go into a textbook (which is why it'll shake up Chem 101 a little bit), but nothing revolutionary. Same goes for any other class at Big State U, Liberal Arts C, or Fancy Private U --- again, with the above two exceptions.

Regarding #1, in practice what has really helped the developing world population are innovations like scribd.com and Bittorrent that facilitate the mass piracy of expensive books.

IMHO the first paragraph of the article gives a clue to the nature of the "offline vs. online" debate. It seems like a purely religious question of "evangelical vs. liturgical" worship. I guess the future offline universities will get more and more religious.

As regards:

"Until late college, physics is mostly teaching knowledge known since Newton."

Uh, no. At a good college, you get past Newton in the first semester of freshman physics. This doesn't alter your main point -- that physics doesn't change fast -- but it's just not right, as a statement of fact. If you want to set a date for that statement, it shouldn't be the late 1600s (when Newton lived), but maybe 1950 or so. Even when it comes to teaching stuff Newton knew, we have better ways of explaining it than Newton did.

Nor is it correct that:

"Most of the mathematics known or needed by most people has not advanced much beyond Euclid and Pythagoras, let alone Euler."

The ancient Greeks hadn't even invented the zero. (Euler, on the other hand, knew a lot.) For that matter, it's only the nanny state that keeps people from needing a lot more math than they presently know, and the nanny state is going bankrupt.

As for the overall prospects of online education, it'll be quite difficult to come up with an automated way of figuring out why a student has screwed up an answer, and telling him so in terms that he'll understand. Siri isn't anywhere close to being able to do that. But one could imagine a system where that task was farmed out in an automated way, as in something like Amazon's Mechanical Turk, except with students able to complain to the professor if a "Turk"'s correction/explanation was unsatisfactory. Or to take load off the professor, the complaints could first go to some other "Turk" who was known for good answers, before a possible final appeal to the professor himself.

As for the social aspects, something lacking online is the peer pressure of 'everyone else in the room understands this, and you know them and know they aren't much better than you, so buckle down and study it until you understand it yourself.' Perhaps integration with Facebook could provide that sort of peer pressure. Perhaps not.

What's the connection between the nanny state and math? I'm puzzled.

The hazards that the nanny state protects people from are real hazards. Estimating the magnitude of those hazards requires math, which is why regulatory agencies do quite a lot of it. If people had to judge those hazards themselves, they would need to know that math. Of course almost everyone would trust experts rather than doing the math personally, but even to choose between competing experts requires some knowledge of what the experts are supposed to be doing. When the government makes the decision for everyone, and enforces it via regulations, there is much less reason to know math.

(Unless, of course, one anticipates a collapse. Then one wants to know not only the math of ordinary hazards, but the math of ballistics.)

"When the government makes the decision for everyone, and enforces it via regulations, there is much less reason to know math."

... unless the government is making bad decisions because it is in someone's political interest to vastly under- or overestimate a risk. Then there is much more reason to know math, not only to evaluate risk for yourself, but to be prepared to convince everyone that the government's estimate is wrong.

The government is made of people, not omniscient automatons. Laws are voted on by more people. Those people are elected by people. The opportunities to ignore math and choose the outcome you want are numerous. Ask Greece. Go through an airport security screening. Look at the US defense budget. Understand why the ICC had the power to set minimum rates for rail traffic. Read about the Palmer Raids.

Oh, I wasn't trying to suggest that the government always gets things right. Quite often they don't. But right or wrong, when the government makes a decision and enforces it, they reduce people's options, making the knowledge of how to choose between those options less valuable. As for the ability to persuade others that the government has messed up, that's hard when those others don't know math and can't easily understand your arguments. While fighting the government can be quite beneficial to society as a whole, to the people doing the fighting it's mostly just a hassle.

Or, in older language, "...great empires, while they stand, do enervate and destroy the forces of the natives which they have subdued, resting upon their own protecting forces; and then when they fail also, all goes to ruin, and they become a prey." Speaking of things that haven't changed much, that quote is from Francis Bacon, four centuries ago.

The bottom line is that if you don't care about learning something, you won't learn it beyond the basics anyway.

But if you care, then waiting until you can attend lectures on it and only learning about it 2 hours per week is ridiculously insane and will likely remove your interest, as opposed to just reading information and watching video lectures right now 10-16 hours per day until you fully learned it.

Did anybody ever take an online course seriously? I did recently my first course at Coursera.org, a class on Gamification, hold by Kevin Werbach, together with about 70.000 students of which around 10.000 completed the course. Short summary:

The video lectures and materials were excellent and I agree with everybody that it is great to take your own pace, view lectures whenever wanted and view them again, which I did partially for work on the written assignments.

The written assignments were fun to write but peer review was a bit more difficult for two reasons:

1. There have been so many submissions of bad quality that I got annoyed by having to rate them - people simply didn't understand or possibly read the assignments.
2. I was even more annoyed by bad ratings given by students obviously didn't carefully listen to the videos or read the materials.

Based on this experience my opinion is that online eduction must get the "right" students into the virtual classes - the smartest ones into a class, the not so smart ones into other classes and so on. Let me summarize it bluntly - I am not interested in online eduction when I know that I have to attend together with a bunch of morons. Comments and suggestion how to overcome this very welcome!

Not everyone needs to sit through lectures. Lectures, honestly, are becoming out of date. I used to just read the book in class instead of listening and it saved me heaps of time.

Comments for this post are closed