This is a perfect illustration of why non-profit education is broken. There is no principal, so the agents are running the show. (See https://en.wikipedia.org/wiki/Principal-agent_problem if you are unfamiliar.)

In a competitive market dominated by for-profit principals, this wouldn't be an issue. The effective principals would order their agents to cut costs and they would, allowing these principals to dominate the market by cutting costs.

And isn't the higher education market in the u.s. open to competition from for-profit universities ? isn't the problem that for profit universities have a bad reputation vs non-profits so employees prefer people who went to non-profit universities ?

If the product educates students just as well, and at a better price point, why would the cuts be "draconian"?

I would deem it intelligent business and a better use of government resources.

If you can replace one crappy American programmer with an excellent programmer from Wakanda for less money, that's just good business and I've never seen a good argument against that.

However, every programmer has heard the horror tales of companies firing their best (and highest paid) programmers. These programmers were not replaced with equally skilled Wakandans for less pay, but rather the cheapest code monkeys that Latveria has to offer. It saves a bunch of money this quarter, but causes problems in the long run because you didn't want to pay for quality.

In the same way, I'm suspicious of some of the motives here. I don't doubt for a second that software instructors can improve the learning process. I'm working on my own platform to help my students with some of their math anxiety. This is a good idea and is the path forward.

However, imagine a university with three stats lecturers. The first two have twenty years of teaching experience and are well liked by their students. The third one was just hired in as a temp to cover the excess course load and mumbles through her lectures. With the new software, the administration should get rid of the mumbling temp and now let all the students benefit from the good professors' lectures. On the other hand, the university can save $40k by firing one good lecturer and keeping the mumbler. As a final option, the university could fire both of the experienced lecturers, have an overloaded mumbler give a course that's even worse than the one she was doing before, and save $160k. I'd argue that the first option was progress, the second poor planning, and that the third set was draconian. I also believe that the third choice is exactly the one that the administration will make. That's why I'm nervous about this software.

How well do you think American society and institutions were doing during the hay day of not-for-profit education versus how well do you think it's doing now? Does your criteria include providing for all citizens and having a robust public debate?

It seems more likely to me that the marketizing measures of today have done a lot to break the community building aspect of higher education but that for-profit education hardly promises anything better. Suuure, blame the government for the scams that constitute for-profit institutions now but without the student loan scheme, they wouldn't exist at all.

Could you be more specific about what periods you are referring to? Personally, I would characterize now as the hay day of not-for-profit education - we have far more of it than ever before.

Suuure, blame the government...

Huh? I was blaming principal-agent problems in nonprofits, and suggesting they are less of an issue in for-profits (though it is still an issue even there). What does this have to do with the government?

Broken it may be, but I certainly prefer free and inefficient education to the alternative.

Why not? How does your normative point (that education should be free) relate to the positive point (that education isn't currently competitive) at all?

Also, I'm curious which service you believe should be free? Education, evaluation or credentialing?

Specifically, suppose we create a marketplace of for-profit education sellers as you suggest, and that marketplace creates good-enough-cheap-enough* slots for 80% of eligible students. What happens to the remaining 20% of students? In an ordinary marketplace, that demand would go unfilled, creating an incentive for more good-enough-cheap-enough slots to be created (up to some equilibrium, not necessarily enough for everyone). But we can't let "demand" go unfilled in this case (i.e., throw smart poor kids under the bus). So does the government fulfill its guarantee by favoring unacceptable sellers, either paying extra or lowering its standards -- giving sellers who charged less an incentive to raise their prices or provide a worse service? Or does it fulfill the guarantee by re-creating public universities, abandoning the goal of a for-profit marketplace?

These aren't rhetorical questions -- we're currently using both strategies, and more. It's not an unsolvable problem, but it's not a trivial one.

(Check out the Commonwealth Health Connector to see how Massachusetts is solving the same problem in health care, combining for-profit providers with a universal guarantee. They're not doing a bad job, but it sure ain't a conventional free market.)

* And then there's the question of what's good enough and cheap enough, when it's a democratic/lobbying process setting standards -- that's pretty different from an ordinary marketplace between lots of buyers and sellers.

Ultimately I believe the problem with what you propose is the idea that the government can guarantee that every buyer will find a seller at a (marginal to the seller) cost below X. This is just an attempt to pass a law against scarcity, which can't really work. All you can do with law is create different rationing mechanisms - perhaps replacing price with queues or favoritism.

Incidentally, even ignoring all government intervention, college education is not a conventional free market. Inefficiencies abound, most notably the prevalence of huge numbers of non-profit seeking agents and reputation effects.

For all those who like to discuss market failures, this is the place to find them.

Anyway: when pantaloons said "you can't have a competitive market for what should (ostensibly) be a free and universal service", they were making a normative claim that universal education is desirable, but a positive claim about how to achieve it: that "universal" and "competitive" are mutually exclusive qualities of any marketplace. That's the argument I was expanding on. I claim that it is very difficult to structure a competitive marketplace while also ensuring that everyone who needs some can get some. Thus if the goal is universal access to a good, affordable college education, your free market proposal is an incomplete strategy.

You seem to be saying that government cannot ensure that everyone has access to a good, affordable college education ("the government can['t] guarantee that every buyer will find a seller"). That's a position I think might be hard to support. The government absolutely can pass a law that prevents scarcity -- which is to say, ensure that citizens have access to something the free market doesn't make enough of. It can, for example, provide universal access to high school. It can provide universal access to assault rifles for healthy young men during a war. It can provide universal access to school lunch. It can provide universal access to health care for people over 65. It could, if it chose to make that a priority, provide universal access to college education for those who need it.

If you propose that "non-profit education is broken," and that the solution is for-profit education, then you need to either explain how for-profit education can provide universal access, or explain why universal access shouldn't be a goal.

The government absolutely can pass a law that prevents scarcity...

To some extent, yes. The government can redirect resources from other uses towards some fixed use. But the extent to which this is done is still fundamentally limited. For example, you can't significantly increase the amount of medical services provided for at least 7-8 years, which is the amount of time it takes to train a doctor. (You might, however, redistribute medical services from some people to others.)

Although I generally believe the burden of proof is on the people arguing in favor of taking action (e.g., ensuring "universal access"), I'll link to a post where I argue we should tax education rather than subsidize it: http://news.ycombinator.com/item?id=2454363

I didn't indicate that I think any should be free. I think education should be universally accessible.

This is certainly a concern for non-tenured, non-research faculty. The plain vanilla instructor may really have something to worry about.

Looking at the original study:

These were intro stats classes with 100-1000. These classes are already basically a teacher lecturing, minimum engagement, simple material, no ability to have complex homework, and from the looks of things not in the student's major specialization (so super 'in-depth' learning is unnecessary)

Universities have realized for years that they are not very good at these types of classes. They've tried to move these courses online or encourage students to take these at community colleges. But, you don't (or shouldn't) go to university for these classes, you go to the university for the more specialized upper division training in your department. I would be more interested to see if a high level, small, for majors course could be replaced by a similar program. It's worth noting the even the Khan Academy classes are still the 'factory' classes, he has very few 'specialized' classes in more complicated subjects.

Secondly, this is an interesting result, but not too surprising. Working through exercises on your own with software that presumably identifies and targets area of weakness ought to put the onus of learning on the student.

On the other hand, many college students sitting in a lecture don't engage the material themselves to the same degree and instead rely on the professor to "teach" them instead attempting to learn it on their own.

In fact, I'm sort-of surprised the machine-learning students didn't do better than the lecture-listening ones.

On the other hand, material in a medium to large lecture is rarely custom tailored to individual students. If half of the class doesn't understand what slope represents in a supply curve, the experience is ruined for the more gifted students desiring to learn the finer points of economics. Once a lecture regresses into basic concepts you already know, at the loss of interesting finer points, it's quite easy to zone out and lose focus. When the "instructor" to student ratio is 1:1, custom tailoring can really benefit those outside the norm.

Even more, the student is driving the educational vehicle. It's quite empowering to be able to skip over videos about concepts I already know, especially when I can confirm my thoughts with a basic exercise (e.g., a multiple choice question, or short code prompt).

Where does this idea come from that the title should be the same? Granted, the current title isn't quite perfect either -- "Study Reports Software Speeds Up Teaching" would be more accurate -- but it's at least more informative, and certainly no more editorial, than "Score One for the Robo-Tutors".

It is suggested although not mandated by the Hacker News guidelines:

http://ycombinator.com/newsguidelines.html

"Please don't do things to make titles stand out, like using uppercase or exclamation points, or adding a parenthetical remark saying how great an article is. It's implicit in submitting something that you think it's important."

. . . .

"You can make up a new title if you want, but if you put gratuitous editorial spin on it, the editors may rewrite it."

The retitling in this case seems to have been effective for eliciting more discussion, and I appreciate the various interesting comments here. I regret cases where I submit an article with an original title, only to find out that the same article was submitted hours before with a retitled submission title, one that is missed by Hacker News participants and does not elicit discussion. Most professionally published articles have well written headlines, and it is often hard for HN participants to improve upon those.

Headlines may be professionally written, but they are written with incentives which don't favor HN readers. Most of the time, I'd prefer to hear why the submitter thinks I should read something rather than why the author does.

Honestly, I would be more interested in seeing the comparison between variation in time/scores for hybrid + traditional formats. My hunch is that the smartest people are significantly dragging up the hybrid scores. My personal experience in class is that professors tend to teach at a rate acceptable to the top ~75% of the class, leaving the top ~5% feeling like class is moving like a glacier. And then there's the fact that each student typically gets stuck at slightly different points, so if a lecturer is to clarify each of these points, there's a LOT of redundancy being thrown around.

Most introductory statistics courses are crap, going wrong right at the start by hiring the wrong instructor and specifying the wrong textbook.

http://statland.org/MyPapers/MAAFIXED.PDF

Statistics, the "science of data," is actually an endlessly fascinating subject, from the introductory level on up, but few statistics courses are taught in a manner that requires anything but a robot to present the material, and those statistics courses are so dull that they don't encourage student engagement. In fact, statistics gets into some deep issues about the nature of correct inference right from the beginning,

http://escholarship.org/uc/item/6hb3k0nz#page-1

and it occurs to me that a really, really good statistics course might be a lot harder to automate (although this would be a worthy challenge for artificial intelligence researchers) than the bleak general run of introductory statistics courses.

As a good human statistics teacher, and perhaps even a good automated statistics teacher, would say, further research on this point is needed. One of the best examples I have seen of application of artificial intelligence (software) to teaching of a foundational subject is the ALEKS

http://www.aleks.com/

online K-12 mathematics course, used by a variety of higher education institutions to supplement remedial courses, and used by an increasing number of secondary and elementary schools and self-taught learners of all ages to learn mathematics and some closely related subjects. (ALEKS has a statistics course

http://www.aleks.com/about_aleks/course_products?cmscache=de...

that mostly covers the dull issues rather than the cool issues in statistics.) I like ALEKS a lot for my children learning mathematics and for myself and my wife learning basic principles of sole proprietorship accounting. The ALEKS learning spaces model

http://www.aleks.com/about_aleks/research_behind

is based on some very interesting research, and I have previously recommended that the Khan Academy developers assimilate that research base as they continue developing the Khan Academy courses.

Based on real-world examples, it just might be that automated presentation of many forms of course content will be more efficient (resulting in more learning for fewer dollars in less time) than current in-person models of instruction. And everyone should be happy about that. There will always be a role for in-person teachers, and they do their best work if they use their comparative advantage of being able to do some things better than machines, even if everyone acknowledges that some machines do some kinds of teaching better than most human beings. Let the research continue, until we have an abundance of good data.

It was during the black death when there was a shortage of lecturers and many students wanted to go away from the city to learn remotely in their own homes

So they came up with these portable reading tablets, a sort of p-book, which contained the lecturers words and could be read anywhere, even in bright sunlight. Eventually large stores of these were collected together in data repositories- a bit like tape libraries but with human robots.

- they also invented exams when they realised it was cheaper to examine a 100 students at once and outsource the marking than have the professor viva each student individually.

Remote-instructor learning through software differs markedly from the same through books in a variety of ways. ie the software can provide more compelling extrinsic motivations, but can also provide increased distractions (both if used on a computer with internet access, and through the affordance of screen-based tools).

So its good not to treat this as an either or problem (does software work as well as face-to-face) because there's a number of other considerations (is the face-to-face personal or lecture hall, is the software highly interactive with immediate feedback, etc

I had enough teachers who were barely able to grasp the material themselves, vindictive, or emotionally unfulfilled in their personal lives, that I'd rather have learned from a book, computer, and videos. But hey, maybe dealing with crazy or incompetent educators is an important socialization step for young people...

As opposed to _generating_ successful anthropological-demographic predictions of linguistic behavior? Shouldn't students be "liking" professors' "asides" anyway?