Will online classes make professors extinct? | Hacker News Reader

Will online classes make professors extinct?

cnn.com

25 pointsesalazar41 comments

Does anyone have numbers indicating the extent to which tuition subsidizes research funding or vice versa? I (and most HNers, I'd assume) wouldn't shed many tears over seeing the extraordinarily exploitative and inefficient institution of college education go through disruption. My experience has been that online lectures, course material, and forum discussions are more than up to the task of replacing their physical counterparts (pause+rewind is a killer feature, and the discussions in online forums are miles above anything I experienced in section/recitation). Hands-on labs are about the only thing that ought to have a marginal cost in this day and age. But I hate to think what disruption could do to academia.

There are a handful of big (sometimes existential!) problems lurking on the horizon which absolutely rely on academia because incremental progress isn't amenable to monetization. I'm most worried about the drug industry. It's imploding (google "inverse moore's law"), leaving the basic science required to get over the slump in the hands of academics, and I don't want to wait 50 years for high-efficiency gene therapy (or let China grab the market) just because the research is tied to an outdated educational system. I know there's very little anyone can do about it, but I need to know the extent to which college tuition subsidizes said research before I can form a solid opinion on disruption.

EDIT: Got off my ass and googled it. The NSF estimated in 1994 that 9.5% of research funding comes from tuition [1]. I would like to formally submit one internet-vote in favor of disruption, along with one internet vote to increase research funding to compensate :)

[1] http://www.nsf.gov/statistics/issuebrf/sib97313.htm

You mention pause+rewind as a killer feature. Another nice feature is the ability to watch at a faster or slower pace (I usually watch 1.5x). These conveniences have made it such that I strongly prefer watching taped lecture to physically attending class. Add to that the fact that you can almost always get a faster response from posting in an online forum (often Piazza for my Stanford CS classes) than attending overcrowded and understaffed office hours. Most of my education has already moved online.

Even if online class doesn't make professors extinct, I think it will certainly make physical class time a thing of the past.

Yes! 1.5x is fantastic! It solves the impedance mismatch inherent in human speech. 90% of what is being said takes 5% brainpower to process while the remaining 10% takes 300% of available brainpower. If you have to listen at 1x speed you are either bored and sleepy or falling further and further behind, sometimes both. The pause button solves the falling-behind problem and 1.5x speed solves the boredom problem.

"90% of what is being said takes 5% brainpower to process while the remaining 10% takes 300% of available brainpower."

I'd argue that's the result of poor communication, not any deficiency in human speech generally. Granted, it's very very hard to construct something with the right level of addressable information at every moment. But when it's done right it is glorious. I think one of the things people implicitly recognize in good communicators (not necessarily good orators, but good communicators) is their ability to tailor what they're saying in this way.

Variance in the backgrounds of your audience and variance in the physical state of your audience (tired, sick, etc) places a strict upper bound on how well you can communicate. If one student isn't experiencing an impedance mismatch, you can probably find another that is, unless the class is extremely small and homogeneous. Pause+rewind is robust to these problems.

I find that the phrase "pause+rewind" doesn't quite give the right impression; sure, sometimes you missed a sentence and need to rewind, and that's great. Other times, I'd like to fire up an editor and type out the code the lecturer's showing (I'm often surprised at how many things my mind missed just skimming the slide), see if it compiles, how it behaves for 10 minutes or so. Kind of like a self-imposed mini lab, if I'm feeling a bit behind on the topic at hand.

Other times, it's great to play it at 1.5x, or even skip sections entirely. E.g. if I'm comfortable with the topic / environment, I don't want to sit through 20 minutes of setting an IDE up, or a walkthrough of "hello world". I wonder if it would be possible to deviate from the strict linear progression of lectures, and allow the learner to choose. The difficulties in prerequisite knowledge might outweigh the benefits, but it might be an interesting idea. More of a library of closely related lectures and information, rather than a single course.

Another feature that I have found to be quite useful is when I don't quite grok something, I can often quickly jump over to a lecture from a completely different person on the same subject. That alternative perspective can make it instantly clear.

For my part, I don't consider college education to be exploitative or inefficient.

People pay for the time and effort of professors who are very knowledgeable and have extraordinary teaching ability.

I also think MOOCs have great potential, not because college education is bad, but because MOOCs allow the best educators to reach even more people. I don't expect the top-end of education to change much, but I hope that MOOCs can raise the overall quality of education.

Only a small amount of the tuition goes to pay for professors. Most money goes to feed administrative bloat.

http://www.bloomberg.com/news/2012-11-14/bureaucrats-paid-25...

Friendly reminder that in something like 48/50 states the highest paid public employee is the football or basketball coach, and the hockey coach in NH or VT.

I read the article, but I'm not convinced that there really is "bloat". Higher education is bigger and more complex than it once was, and the people best able to make strategic decisions might also be expensive. Coming from private industry, those salaries don't seem particularly high.

They may "bring in" more than their salaries (I suspect they don't, which is a separate issue -- the incompetent bumbling cited in the article seems to be the rule rather than the exception) but even if they do "bring in" more than they cost, it doesn't mean they aren't bloat. Competition for students is a zero-sum game. Luring students to one university takes them away from another. The value of a strategic school administrator to society (the extent to which they better inform students of their options minus the extent to which they convince students to act against their interests) is tiny compared to their value to the particular school which employs them (some number of student tuitions). The difference between the two values can reasonably be called 'bloat'.

This is a known flaw of free markets that we usually write off as a cost of doing business. If you present the invisible hand with a prisoner's dilemma, it falls for it every goddamn time. The 'bloat' is then the difference between the cooperative solution and the nash equilibrium. In this case, the cost is egregious enough (it is very high and it is imposed upon a very vulnerable population) that we might want to pay attention. I doubt we will, leaving disruption as the only hope for the education market.

Many universities are utilizing MOOCs and/or online material in order to implement a reverse classroom. In a reverse classroom, students watch lectures and do reading at home and class time is used to extend and reinforce that material usually by answering questions and working through more examples.

Students take to it pretty well when video lectures are available (my experience is that book only reverse classrooms don't work as well). It also generally leads to better retention, since more time is spent on hands-on experience.

Instructors also take it pretty well. Using an existing set of lecture materials reduces the preparation workload, especially in the first year teaching a course.

Whatever the failings/inefficiencies of college education, I'm not convinced that online courses are the answer. Given the non-credit nature of current MOOCs, they attract mostly bright and self-motivated learners, which leads to high-quality discussions on the forums. Every school and every company in the world has a few such students/employees, and there are enough schools and companies in the world to make current MOOCs shine even if a small fraction of these people subscribe.

SJSU's 'disastrous' experience with 'regular' students may indicate that most students need the discipline/peer presence of a physical college:

http://www.commentarymagazine.com/2013/11/23/education-revol...

Betteridge's Law of Headlines says NO.

Katy Jordan did a study of MOOCs recently [0], and it was found that the completion rates are in fact quite abysmal. I suspect that the reason for this is because there isn't much incentive to continue studying.

So this is very much in line with the guy saying at the end of the article: Here's a library, now go get knowledge.

Learning, IMO doesn't happen like this. MOOCs will just perform the functions of a library, albeit far more efficiently. The function of a classroom (not necessarily a professor) is still quite necessary for structured learning. And of course, with a classroom comes TAs and professors... which makes the answer to the question, No

[0]http://www.katyjordan.com/MOOCproject.html

> Katy Jordan did a study of MOOCs recently [0], and it was found that the completion rates are in fact quite abysmal. I suspect that the reason for this is because there isn't much incentive to continue studying.

No, it's because there is no burden to sign up, nor disincentive to dropping out. Thus, you get lots of people who sign up for them just to watch a few videos or to see what the course is all about, but they have no real buy-in.

According to a talk I saw by Andrew Ng, 43% of people who submit the first homework on Coursera go on to finish the course. That's actually a remarkably high completion rate IMO given the scale of the classrooms.

Not sure if it's the same researcher, but I heard an interview with one which talked about another problem with MOOCs - they're generally only useful to those people who already know how to learn. The autodidacts or people close to being autodidactic.

My own criticism of MOOCs is that they're not ideal for any subject with physical skills. I was a class tutor in the late 90s doing biology and electrical engineering classes. At this time the 'can learn everything from a computer!' crowd was getting really underway, and I remember wondering 'how is a computer going to teach you how tight to tie a liguture?'. Similarly, my favourite lecturer mentioned that for a lot of skills, a technical college is better. At university we might be expected to solder together a circuit, but zero time was spent teaching this skill. At a tech college, they do physical skills, starting with things like "here's a bunch of nails in a block of wood, solder wires across their heads" - exercises solely for honing physical skills.

And if you are having trouble doing a physical skill, having an expert pair of eyes see what you're doing is a hundred times better than trying to post it on a forum. For example, I'm an indifferent guitar player and have skimmed forums here and there for years, asking questions about this or that... but it was during my first lesson with a professional guitar teacher that I had considerable issues cleaned up with my technique.

MOOCs are a great tool to have in the arsenal, but they're not a replacement technology.

The data on that page isn't loading for me but either you're interpreting the data too naively or their study is missing some very important factors. With zero cost (money or otherwise) to enroll in a course the number of students who will sign up should be ridiculously inflated. There are people who just want to get their feet wet and try the first few assignments, some who just want to download all the videos, etc. Even if you only measure the people who did all the assignments or exams it's still not a direct comparison to a traditional classroom because there are: no curves (yet– though a few courses drop the lowest assignment grade), a hugely diverse student body without any verified qualifications/prerequisites, varying English (or other) abilities, etc.

Characterizing the completion rate as 'absymal' suggests that the number is intrinsically significant which really isn't the case outside of a traditional school– where the enrollment pool is already carefully filtered and there is external pressure to keep rates at some arbitrary level.

Edit: Forgot to refresh before submitting my reply so this is repeating some things which have already been said.

It's my belief that the MOOC, and online education in general, are something of a fad. It's new, and therefore interesting. But we've always had correspondence courses, and libraries where you can go and self-study almost any topic under the sun.

Online makes that easier (you don't have to leave the house, or type up papers and mail them in) but fundamentally, online education still requires a lot of self-discipline.

If most people were really good at self-study, we wouldn't have ever built up institutions of education. The real problem with undergraduate university environments right now is that the perceived value has fallen dramatically as tuition rates have skyrocketed. The universities and providers of easy financial aid are to blame for that.

> It's my belief that the MOOC, and online education in general, are something of a fad. It's new, and therefore interesting. But we've always had correspondence courses, and libraries where you can go and self-study almost any topic under the sun.

MOOCs differ from just reading a book on important aspects:

1. You have a teacher break down the content for you

2. You have graded assignments, which helps giving confidence you're absorbing the content

3. Interactive media (video, software) can be a more effective media to explain certain concepts

4. You have interaction with other students learning the same subject

> If most people were really good at self-study, we wouldn't have ever built up institutions of education.

Back in the day, your only chance of having access to knowledge was at institutions, so I'm afraid this is more of a tradition than anything else. Remember that the church was once the source of knowledge too. Nowadays though, people aren't going to college for knowledge as much as for filling a pre-requisite to enter the job market.

The problem with just looking at the completion rate is that it doesn’t really distinguish between the “just browsing” crowd and the “I really want to finish this course” crowd.

And the reason for looking at the completion rate I'd that neither does a headline saying "1 MILLION STUDENTS ENROLLED"

People learn in many different ways, and therefore I dont think you can make the sweeping statement: "Learning, IMO, doesnt happen like this."

I have always been someone who prefers self directed learning to the classroom (this coming from someone who spent 7 years in college / grad school), I think MOOCs are great and I get as much out of them as I did many college classes. I found the freedom to study any topic whenever I wanted to very liberating; I ended up studying more. Some MOOCs are certainly better than others, but you could say the same of regular classes.

There is definitely no replacement for office hours with professors, but facebook groups, community TAs, and Skype actually close the gap more than you'd think.

Completion rates are a red herring for a variety of reasons. The biggest is the total lack of risk inherent in joining a free class with no transcript. In college you should be more conservative about enrolling for classes because you can ruin your academic record or waste money, and you also often have degree requirements and schedule conflicts to worry about. With MOOCs you might as well join whatever you're even half-interested, even joining more classes than you can possibly realistically complete and finish whichever ones seem the most promising after a "trial period".

Generally speaking, for any system, reducing the cost of failure will make the failure rate go up.

Online courses are cute. The stanford machine learning course is a watered down travesty that doesn't compare to Andrew's original material out there. There is a reason for that too. The harder the courses get, the more it is nice to have an actual person to talk to about this shit. My real analysis class made sense only because I was able to talk to the professor after class and understand why I should care about open balls or closed balls.

Similarly from a systems perspective, my OS class was the hardest C.S. class that I took. There was very little that I remember from the lectures but the most I took away was during designing shit and getting stuck and then going back and talking to the professor and brainstorming. This is stuff that inherently doesn't scale well to millions of people.

That course isn't representative. I don't know the thinking behind it, (it was one of their earlier MOOC courses) but there are many courses with much better and harder content.

Of course it is better to have a real person, but not everyone has that luxury, or their real professors are no good.

Sure. Learning from data (caltech) is a good course. But all I got from the course was to go pick up the book and learn this stuff by working through the problem set. And I have a pretty decent linear algebra and probability background. Imagine if you are getting into this stuff new (which is not an unreasonable assumption considering the diverse background of ML folks), this would be brutal. Math is a contact, no holds barred sport where some hand holding initially is pretty damn nice.

Also, I did some schooling in the third world and I am never going to come out and say that someone putting material for a course online is a bad thing. I am just annoyed by this idea that a bunch of online videos is a superior or equally effective replacement to an education at MIT or Stanford.

The online problem sets for the Caltech course aren't what you get in real life when the course is being actively offered, and also it was a pretty basic/easy course. That's not to say there aren't other hard classes online or that Yaser isn't great at teaching, but yeah...

I've taken a couple Coursera and other self-led courses and it's not a good fit for me. I recognize my deficiency at self-directed learning, and I imagine many many people share it. I just can't learn that way.

A good chunk of people will require social, interactive, and hands-on learning led by people to whom they can listen and converse. Until that experience is sufficiently reproduced online (and surely even after that), "professors" will not be out of a job.

Also the title is hyperbole.

I agree with you 100%.

I teach online and hybrid (part face to face, part online) courses. My students tell me they actively choose one or the other format. Some students cannot learn online and appreciate the ability to interact. I try to use the class time we have effectively, to demonstrate, to coach, to give feedback on their work. I save the "lectures" for their own time.

I've taken online courses from some pretty bright people and never felt that they were a good fit because they were simply lectures. I don't learn by lecture. I learn by doing and getting feedback. I do horribly in "traditional" instructor-focused courses, AKA "lecture/test" courses. I do very well in courses where I can apply things.

One thing I think is great about online education is the ability to reach a lot of people. But one thing that's terrible about it, in my humble opinion, is that it's harder to connect with students. My online classes have a cap of 27 students. I can manage a couple of courses like that. But in order to manage courses of 200-2000 people, I just would be putting up content and videos, and I'm sorry, but that's not teaching. That's nothing different than screencasts or books. If I get assessed, I want detailed feedback. I want to ask questions.

And if those questions are being answered by a support staff, then there goes the argument of "learning from the best and brightest" that MOOCs advocate.

I love teaching students hundreds or thousands of miles away. I don't think my role is going away any time soon, simply because I've evolved how I teach. The ones who should be worried are the ones who teach from the textbook and spend 64 hours per course lecturing from the front of the room.

[edit to clarify reply]

"Science cannot produce ideas by which we could live. Even the greatest ideas of science are nothing more than working hypotheses, useful for purposes of special research but completely inapplicable to the conduct of our lives or the interpretation of the world. If, therefore, a man seeks education because he feels estranged and bewildered, because his life seems to him empty and meaningless, he cannot get what he is seeking by studying any of the natural sciences, i.e. by acquiring "know-how." That study has its own value which I am not inclined to belittle; it tells him a great deal about how things work in nature or in engineering: but it tells him nothing about the meaning of life and can in no way cure his estrangement and secret despair." - E.F. Schumacher (from his essay The Greatest Resource - Education)

This quote is part a defense of the humanities and liberal arts as essential for presenting ideas of who, what, and why we are.

Schumacher makes a strong distinction between education and training and I lean towards categorizing MOOCs as the latter. While they may achieve wider credentialed status, they will not and should not replace traditional educational institutions. Pricing, degree structure, and allocation of resources (professor pay vs. recreational amenities) might all need to be looked at but the mentor-pupil dynamic is essential to the liberal arts.

I've taken several coursera courses and maybe the good ones are hiding in there somewhere but I'm really unimpressed by the ones I've seen.

I've had fun with the intro to python courses, but on searching around, there's not a lot for intermediate-level classes in the several topics I've looked for.

Which courses have you taken? I've completed the Machine Learning course and I'm currently taking Cryptography 1 from Coursera. They both have been great.

Machine Learning, Drug Development, Nuclear Energy, VLSI: Logic to Layout, Computer Architecture, and Analytical Chemistry were all somewhere between "good" and "great" in my book.

ML and NE were aimed at non-specialists and weren't rigorous. Maybe that's the objection?

That's part of it, yeah, lots of courses out there but they tend to be things like survey courses, etc. Some of this is that the video content turns out to be the equivalent of only a 1 or 2 credit course.

It's unfair to criticize Coursera for having survey courses if they also have deep courses, which they do. It would be nice if they had more, but that's not really a criticism.

There's a Stanford professor that has the videos from his class about the Fourier Transform on youtube, go compare that to (just about?) anything on Coursera.

Schools aren't just going to disappear. It's natural for technology to take over a sizable chunk of the teaching process, handling inefficiencies as technology has always done best. It's probable that most of the world's population won't need to interact with professors, but a select few will always have the privilege of face to face learning. I believe there are too many niche studies for professors to be entirely replaceable. Not to mention the indirect benefits of attending a school campus.

Will high quality textbooks and the printing press make the medieval university go extinct?

No more than ebooks have wiped out printing.

Reply on news.ycombinator.com