Tuesday, October 23, 2012

matching markets

Speaking of matching markets, it's pretty strange that the economics job market, which is so nicely streamlined and centralized, doesn't use a stable matching mechanism. In terms of the original Gale-Shapley algorithm, it's missing the feature in which women can call of whatever engagement they've previously accepted if a better man asks in the next round of offers. In other words, once you accept a job, you can't really get out of it to accept a better one. That's a problem because offers aren't made at the same time by all universities.

This causes a lot of stress for job market candidates, and I'm sure for universities too, who play a convoluted timing game to try to get their most preferred candidate. Would it really be so hard for universities to agree to a set time period in which offers can be made and accepted and then reneged if something better comes along? I'm not sure who stands to lose, and I'm pretty sure the gains must outweigh whatever losses there are.

Thursday, October 18, 2012


I grew up listening to a lot of classical music, especially for violin. As I got older and started listening to music of other genres, a particular aspect of classical music started to really bug me: there is such emphasis on virtuosity, above and beyond the emotional/aesthetic impact of the music.

Virtuosity should be only a tool to enhance that impact, not an end in itself. Cadenzas, the solo interludes in concertos in which the performer gets to show off his own skills and strengths and style, should fit with the rest of the music, should be beautiful, and shouldn't sound forced.

Compare: Itzhak Perlman's cadenza in the Beethoven violin concerto, probably my all time favorite. Beautiful, effortless. On the other hand, the Paganini caprices. Aesthetically borderline offensive, blatantly designed just to show off one skill after another, painfully forced. Impressive, most definitely. But who cares? It's not a competition. Or if it is, technical skill isn't the relevant aspect.

I was reminded of this last night by an amazing concert by Kelly Joe Phelps, an acoustic/slide guitarist who is the pinnacle of effortless understated virtuosity. His musical genius is undeniable but it only serves to make the music more beautiful. Listen (main song starts at 1:30. I like the vocals on the album version a little better but I'm too lazy to upload it. I've never heard him play the same song the same way twice; more proof of the genius...)

Monday, October 15, 2012

Al Roth matched correctly

One of my favorite economists won the Nobel Prize today. Al Roth (along with Lloyd Shapley) won for his incredibly cool work on matching and mechanism design. If you've read about school choice in New York City, matching medical school graduates with residencies, or chains of kidney donations up to 60 donors long that overcome the problem in which willing donors (friends and relatives of the recipient, usually) aren't a match, you've read about his work. Heroically awesome applications of game theory, is how you might summarize it. (Take that, Ariel.)

Plus, he's a really nice guy. And he writes an excellent blog.

Shapley is also obviously a fantastic choice; he's already an icon of the profession. I blogged about the Gale-Shapley algorithm previously (which forms the foundation of many mechanisms designed by Roth.)

Friday, October 5, 2012

prerequisites and performance statistics

To answer my own question, I think we need to see more clearly differentiated tracks for classes and more clearly defined prerequisite structures.

The online class model is a significant improvement over 'learning things on the internet with disjointed articles, videos, and wikipedia' because the duration is long enough to build up from simple first principles to more complicated ideas. Now they need to improve a step further with clearer prerequisites. Rather than teaching a quantum mechanics class where "you don't need to know calculus! We will present the material in the most accessible way possible", they can teach a quantum mechanics with a calculus prerequisite (...and here's a link to the calculus class you should take first; you can sign up without it of course, but we will explicitly assume that knowledge.)

And/or, different tracks of classes should be more clearly defined. Already many classes have 'optional' assignments and supplementary material for more advanced students, and their certificate of completion sometimes says something about that. Why not make it more explicit? Every certificate of completion should say which track out of which options it's for, what the prerequisites were, and what your score was compared to the average among people who completed all coursework.

I don't think these things are directly beneficial to the "maximize audience in the short run" objective of course offerers, in play at this early point in time. But I also don't think it's contrary to that objective, and could be done in a clearly beneficial way. It could maintain current interest but also attract additional interest and conglomerate statistics of "number of people who took any track of this class or initiated this course sequence" are just as impressive as "60,000 people signed up for this course".

But more importantly, I think these changes are vital for long-run success of online education. People won't chase meaningless certificates if they want credentials, and they won't chase empty dumbed-down curricula if they want real education and employers won't give a crap about certificates that don't have a clear meaning. It's great for those who want a cursory introduction, for fun or curiosity or a jumping off point for more serious independent learning, but that's nothing that's going to ever be able to compete or seriously supplement traditional education. It's something worthwhile in its own right, but it's not "online education".

Thursday, October 4, 2012

online education

Online classes are better subject to forces of competition (no one is required to take them, and there are more options both within and between subjects.) Therefore the following forces operate:

For a given quality/rigor of curriculum, improving teaching methods will increase the number of people who want to take an online class.

For a given curriculum, lowering the rigor will increase the number of people who want to take an online class.

The former effect is limited: no matter how good the teaching is, work is required to understand and master complex subjects.

The latter may also be limited: reducing rigor loses a few people at the top end who want a very thorough understanding of a subject and picks up a lot more (because the distribution is skewed) on the lower end who want a simple introduction. But, at some point even the lower end isn't tempted by a class that can be completed by 7 year olds.

I think that on the current margin, the latter is more effective at increasing audience. At least, it seems to be the tactic overtly chosen by many classes.

Loosely speaking, the former effect is welfare-improving, while the latter is welfare-reducing, but both are potential avenues for an online course offerer whose objective is to maximize its audience. (At this point in time, at the dawn of online education, I think that is indeed the main objective.)

How do we set up the system so that the former is the more tempting option?