Conscientiousness and online education | Hacker News Reader

Conscientiousness and online education

gwern.net

77 pointsgwern29 comments

This article raises some interesting science fiction-y ideas. Quite a long and in depth read, but worth it because gwern's mind is one of our real treasures. Two things I found interesting:

1. If everyone in the world had the exact same Khan Academy style learning experience it would control all of the random factor associated with education. This means that the only differentiating factor would be raw intelligence, which is scary for those of us who were born as geniuses.

2. Conscientiousness (aka discipline or work ethic) is more important in online classes. So future Einsteins might be too bored by mundane assignments to take their work seriously.

Improvements to IQ and personality testing might mitigate both of these problems in the future. If we can accurately measure a student's IQ we can give him or her harder or easier questions, essentially baking a curve into every assignment. Personality testing might yield the ability to identify students who are lacking Conscientiousness in order to target them with more interesting problems at a challenging pace. As usual, I think that machine learning and data science could lead to great benefits for everyone.

This was an incredibly well researched piece that provided some interesting hypotheticals regarding online education. As Gwern points, out drastic technological changes rarely benefit everyone equally. Education tech is going to create some big winners and losers over the coming decades.

"the only differentiating factor would be"

And language / cultural stuff. I can't imagine trying to do trig or calculus with Roman numerals or Mayan numerals. And that's sticking to obvious math examples.

There have been theories that (insert language) has always been superior for (insert hard science) because of inherent linguistic advantages.

You haven't lived until you've heard William Shakespeares plays in their original Klingon and all that. I have a decent sorta-classical education and I find it much easier to read modern translations of ancient greeks than to try to puzzle thru Olde English.

I do wonder about some subjects like somewhat advanced physics, is the "charm" factor of a subatomic particle "charm" when Chinese talk about it, or does it get localized?

WRT making every assignment challenging, I was near the top of my class in most subjects in K12 and that would have driven me completely bonkers with the unfairness of it all. Teens are all, and always have been, a bundle of angst but I would go nuts if little Bobby Tables only had to power on the computer without hurting himself or the machine, but I had to write a malloc() function for the same assignment or similar. There's more than a little prison time mentality in contemporary K12 (intentionally?) and arbitrarily giving people hard time or easy time is going to make social issues. I would probably have gone on strike until grade 12 by coasting.

"There have been theories that (insert language) has always been superior for (insert hard science) because of inherent linguistic advantages."

I've never heard any of those theories coming from anybody who wasn't some know-nothing engineer whose idea of high literature was old Larry Niven stories.

I've been told that a lot of Chinese and Japanese prefer to discuss technical or philosophical topics in English or using English vocab, supposedly because those languages tend to the concrete rather than the abstractions necessary.

It is probably because it holds more opportunities for jobs in the US.

Are they really going to preferentially talk to each other in English out of a long-term abstract desire to practice their English and get a job in the US? Having gone to university on a campus with a very high proportion of Chinese students, I can tell that even when they're actually in the US, they feel free to use their native languages all the time with each other.

And you don't think this has anything to do with, for example, the fact that a lot of the available resources in the world of technology and philosophy are in English?

That matters, certainly, but the terms can be assimilated and then any further conversation carried out in the more convenient and fluent native language.

One argument that Latin is superior to mathematics because of its moral (not linguistic) advantages is found here: http://www.classicalliberalarts.com/library/whylatin.htm "Why Latin and Greek Must Be Studied", under section V. "Because Classical Literature is Morally Superior to Mathematics".

That is (a) not really the same kind of thing and (b) one of the least sufferable defenses of classical education I've seen in many years.

One example is previous century German physicists, chemists, and engineers had a very high opinion of themselves. Go back two centuries and add philosophers.

I will agree its curious how those observations tend to appear from people describing their own accomplishments.

See Malcolm Gladwell's Outliers. Last I checked, he wasn't a know-nothing engineer.

Last I checked Malcolm Gladwell was a 'journalist' and barely a couple notches above Thomas Friedman on the intellectual rigor scale.

I'd argue that an optimized curriculum would start to resemble a video game that people would play for fun. To beat a video game takes grit but not conscientiousness in the same way that it's normally framed for learning (where it usually means slogging through boring material and lots of rote memorization or unpleasant repetition).

In v1 of such a curriculum, IQ would reign supreme, but as the sophistication advanced things like creativity, ability to innovate, and mental control/focus would become increasingly important. Things like letting your mind wander just the right amount, knowing which instincts to trust and which to temporarily ignore, etc.

There are very few natural impediments to the progress of thought, so training intended to prepare students for the real world would necessarily deal a lot with optimizing around constraints. The deeper one goes into this line of reasoning, the more education starts to mean what it intuitively means, life skills preparation.

Few jobs today require the level of pure "knowledge work" that an optimized curriculum would prepare a person for, which is why the current system is generally adequate and being skilled at slogging, memorization, and obedience are still so highly valued.

regarding "discipline/conscientiousness/work ethic": It will probably always be a big factor (it always has been). Probably bigger than usual for the next few years as we figure out this online self learning stuff. We just have to remember that these things are very very new.

Imagine taking a bunch of smart people and turning them into middle school teachers. None of them has ever taught before. They have access to nothing. No textbooks or curriculum. No experienced teachers to help. They probably won't do a great job, not at first. It'll take time to learn what works.

We have time to figure this stuff out. I like that we are in a rush as individual players. Khan Academy can't let coursera blow them away. That's great. But as a culture, we have time. Maybe generations to figure out how to teach people using these new tools.

I believe solutions will eventually be better for almost everyone. We'll probably need to tailor the tools for people with a naturally low self discipline, but I think that is possible.

"Imagine taking a bunch of smart people and turning them into middle school teachers."

An interesting analogy might be .mil drill instructors and other .mil instructors. They have a very high opinion of the value of self discipline, even for future REMFs like I was. I did a lot of OJT when I was in .mil many years ago, on both sides, and decades later all I can really analyze it to at this moment is "its complicated" not just easy or difficult. What little I can get out of it, is the average motivated dude can pretty successfully teach the average motivated dude pretty well.

There is also the business model issue where you profitably churn out three times as many kindergarten teachers as the market can adsorb. Very optimistically you'd end up with a perfect top third being hired, and the rest being my baby sitter or waitress (true story). This survivorship bias naturally leads to peculiar ideas like only the top fraction of people have what it takes to be able to teach; not so, it means the market and culture produce waaaay too many teachers...

> But as a culture, we have time. Maybe generations to figure out how to teach people using these new tools.

I'm concerned that we don't have as much time as one might think. The employment changes seem to be happening already. Maybe as a culture we have time, but in the long run, you and I are dead.

If you focus on measuring present achievement, you can skip faffing around trying to correlate question difficulty to intelligence (that is, just ask a question targeted at that level of achievement). And you don't even need better IQ tests to do it.

Of course, this requires abandoning the idea that students should be competing with each other.

A minor technical objection: you actually can improve slightly if you include a measure of IQ, because if you estimate from just the person's performance on a few questions, there's going to be a lot of inaccuracy in your estimate and a measure of IQ will reduce it (as would, say, a measure of previous grades, or data about previous questions, even in an unrelated area).

This topic was also discussed on the most recent EconTalk episode [1] with Tyler Cowen.

[1] http://www.econtalk.org/archives/2013/09/tyler_cowen_on.html

Indeed! I actually have a whole laundry list of Cowen quotes I need to incorporate at some point - not just from that podcast, but also from his book _Make Your Own Economy_.

I did a quick search of the linked article, did not find any historical analysis of conscientiousness as relates to all previous distance learning initiatives, from correspondence courses to radio classes to film-strips to VCR tapes to analog laserdiscs to multimedia cdroms.

I'm assuming the brain has no magic way to respond differently to a video lecture delivered over analog NTSC on a PBS station in 1960 than to a .h264 file delivered over the internet in 2013.

It would be interesting to know if this was all figured out WRT postal mail correspondence courses circa US civil war era or whatever.

> I'm assuming the brain has no magic way to respond differently to a video lecture delivered over analog NTSC on a PBS station in 1960 than to a .h264 file delivered over the internet in 2013.

Besides the differences in image/sound quality, you have differences in how much interaction there is, and I don't know if any of those video lecture courses had equivalents to discussions forums or peer grading of assignments. And obviously any programming course (or course with automatable grading) wasn't going to be nearly as good.

More importantly, the Big Five/OCEAN model wasn't nailed down until the 1980s, so you wouldn't expect to see any useful research examining Conscientiousness until then. (There might be MBTI studies since that came into use in the 1940s, but I think it's more or less junk...)

Just musing: Many of the learning methods you cited lack user control. If you miss something on PBS in 1960, you don't rewind it, you just lose. Same for most classroom experiences (except for the 10% who actually raise their hands). That's a difference big enough to matter. Postal mail correspondence sound like a interesting comparator, though. (I started with "valid"; it probably isn't. Online education also has the possibilities to incorporate a ton of extremely swift feedback. But at least interesting in terms of measuring the effects of conscientiousness.)

Good points. Although even a brief comparison of similar yet somewhat inapplicable examples might be handy.

I'm old enough to have spent considerable time doing interactive arithmetic drills on a Commodore PET... Other than not being online, and having much lower res graphics (as if that matters) I'm not sure if there is a difference in the last 30+ years in that small sub-genre, so that tiny sub-genre would be a good area to explore.

Its still a good article, within its self imposed bounds. There's so much interesting / exciting stuff just barely over that boundary...

If you would like to read studies on self-paced learning with immediate feedback going back to the 1960s, search for "programmed instruction": http://scholar.google.com/scholar?q=programmed+instruction

While data is not the plural of anecdote, I've found the same things with friends who have taken MooCs or previously have followed along MIT OCW -- for some it has worked out very well, for others they fell prey to procrastination and taking easier routes.

I've also noticed similar things in community colleges: I attended a community college and then transferred to a university; this saved me a great deal of money. My community college (De Anza) has been known as a "federal school" for UC Berkeley and UCLA, yet very few of my classmates took that route. I wondered why, but came to the same conclusion: unless you're strongly motivated and self-disciplined, it's too easy to fall through the cracks in a community college (no requirement to graduate by a certain date, no financial pressure, no residential learning, no exam study sessions, etc...). To me the desire to study upper division math/physics/EE/CS is what drove me; was already used to working independently and learning from textbooks. On the other hand, those who have always studied in groups, who didn't have quite a clear educational or career goals in mind tended, or who were not used to having clearly spelled out requirements ("you complete these classes in four years with this GPA, otherwise you can't get in" vs. "these sets of universities demand these classes and usually accept students these with these GPAs") would languish, either dropping out without transferring, transferring only after 4+ (3 years was very common), or doing poorly once they've transferred.

As is think MOOCs and any other high-autonomy self-structured educational programs are of great benefit[1] to autodidacts with great motivation and self-discipline. However, nothing says they can't be combined with traditional teaching environment. As an example, consider the case of Calculus -- it's the foundation of all modern mathematics; you have to get it to study all even slightly quantitative disciplines from Computer Science to Political Science. Yet, we entrust its teaching to high-school AP (or IB for rest of the English speaking courses) teaches or graduate TAs. With MooCs we could have these classes taught by professors distinguished for their teaching ability, but conducted with the structure of a regular high-school/lower-division college environment.

... or take the case of a machine learning, a specialized and quickly evolving discipline. A small university might only have a single professor who specialized in it, but she may not have the resources to teach the class (or teach enough sections of the class). She would, however, likely have the resources to organize an experience around a MOOCs and assist the MOOC students.

> if every child is in an environment that lets them develop and flourish to their fullest extent, then any remaining differences in their development will be due to hereditary factors!

Well, almost! There's also variation in diet and other aspects that fall in the "nurture" category.

All of those are part of the environment, and, consequently, addressed by the premise "if every child is in an environment that lets them develop and flourish to their fullest extent".

Reply on news.ycombinator.com