Join 3,555 readers in helping fund MetaFilter (Hide)


you are the result of a social experiment; it's the simplest explanation
May 19, 2010 8:42 AM   Subscribe

The experimental method: Testing solutions with randomized trials -- In trying to help explicate the complexity of society Clark Medal-winner Esther Duflo is raising the productivity of social policies by increasing our knowledge of what works and doesn't work through repeated social experiments of randomised controlled trials. She has a large surplus labour pool, a veritable industrial reserve army, to worth with.

BONUS
-The orthodox loss of faith: "we are witnessing the biggest silent shift in macroeconomic thought since the Second World War"
-History's greatest theme: "the metastasis of exchange, specialization and the invention it has called forth, the 'creation' of time"
-The New World of the Anthropocene: "the current interval of time, one dominated by human activity"
posted by kliuless (18 comments total) 25 users marked this as a favorite

 
Good to see some of these political ideas put to an empirical test. I hate to say it, but a lot of debate on policy just ends up being "this is wrong because it disagrees with my philosophical outlook!"
posted by mccarty.tim at 8:50 AM on May 19, 2010


And wherever politicians must pander to the masses to stay in power, you can be sure that they will continue to disagree with the results of this tests.
posted by ZenMasterThis at 9:10 AM on May 19, 2010


These.
posted by ZenMasterThis at 9:11 AM on May 19, 2010


TED Talks can be hit or miss, but this is one of the hits for sure.
posted by etc. at 9:15 AM on May 19, 2010


I hate to say it, but a lot of debate on policy just ends up being "this is wrong because it disagrees with my philosophical outlook!"

To clarify: It is wrong if it disagrees with my philosophical outlook.
posted by philip-random at 9:22 AM on May 19, 2010


To clarify: It is wrong if it disagrees with my philosophical outlook.

To restate, maybe: "This will not work because it disagrees with my philosophical outlook!"
posted by monocyte at 9:29 AM on May 19, 2010


"to worth with"? In this context, that might not be a typo....
posted by anotherpanacea at 9:31 AM on May 19, 2010


I was hoping someone would make a post on Duflo. I was intrigued by the TED talk, but the New Yorker article really helped me better understand her own perception of her work. I think that one oughtn't overestimate the extent to which these RCT results can be applied to other areas (an extremely effective intervention in Rajasthan may not fare so well in Uttar Pradesh, and it may be a total waste in Malawi due to any number of unmeasured confounders), but her push for quantifiable results in public policy is refreshing, something that we should have been doing since waaaay before the green revolution started.
posted by The White Hat at 9:41 AM on May 19, 2010 [2 favorites]


Yea, empiricism!
posted by Mental Wimp at 10:08 AM on May 19, 2010


mccarty.tim: “Good to see some of these political ideas put to an empirical test. I hate to say it, but a lot of debate on policy just ends up being ‘this is wrong because it disagrees with my philosophical outlook!’”

I don't think I believe there's such a thing as an empirical test of policy. I haven't had time yet to wade through many of these very interesting links (great post - thanks, kliuless) but it seems to me that "social experiments" aren't like actual scientific experiments in any serious sense. What's most striking about the difference between "social science" and hard science is the significant difference in approach: a physicist would never try to understand the way matter works by first doing statistical models of how often a mass does this or that, or what its tendencies are. A physicist would (a) try to determine unchanging rules – not tendencies, not statistical probabilities, but sureties; and (b) try to determine what exactly is going on at a much lower level. Yes, the result of this approach is that we're now able to discuss the motions of clouds of atoms by statistical analysis; but that result comes after we've attained the best possible understanding of what's going on inside objects at an atomic level.

Society is a mass of people. The chief difficulty of political science is that psychology is prior to it; we can't really start thinking in any clear way about how a multitude of human beings interact until we understand basic, obvious, certain things about how individual humans act. And, just as statistical analysis of the chemical makeup of a gaseous cloud would be sort of ridiculous if we have no idea how molecules combine or behave, any science that tries to proceed statistically by considering a vast aggregate while ignoring the individual instances seems a little misguided.

To my mind, this is also the problem with the popular notion that psychology can proceed by doing "studies" of large groups of people; these can be interesting indicators, but theory has to be explained and demonstrated on the individual level to have any concreteness. Most psychology actually does this, but the papers still delight in printing the latest studies on whether people tend to do this or that.
posted by koeselitz at 10:21 AM on May 19, 2010 [1 favorite]


What's most striking about the difference between "social science" and hard science is the significant difference in approach: a physicist would never try to understand the way matter works by first doing statistical models of how often a mass does this or that, or what its tendencies are.

Yes, well, I don't think the intent of these experiments is to form a unified theory of human behavior as much as to determine the pragmatic results one gets from a particular intervention. I think it is possible to generalize to some extent by such experiments, or at least determine whether an intervention can work in reality. Imagine that vouchers were offered on a lottery basis and we compared the results among winners with those among losers. Obviously the results can only be generalized to those who volunteer for vouchers, but, then again, that's the problem with observational studies of vouchers: those that seek them are different from those who don't a priori. If vouchers don't work for this crowd, they don't work. Even though we can't claim to have discovered some grand truth about human behavior, we have learned something very valuable about vouchers and human beings.
posted by Mental Wimp at 10:34 AM on May 19, 2010 [2 favorites]


Oh, and another thing: experiments in quantum mechanics are very much about the statistical modeling of particle behavior.
posted by Mental Wimp at 10:37 AM on May 19, 2010


"to worth with"

how did that happen!?

also btw here's elizabeth kolbert on the Anthropocene Working Group :P
posted by kliuless at 10:40 AM on May 19, 2010


Some criticisms of the this style of methodology in development (Acemoglu) and macro economics (Sims).
posted by diftb at 11:08 AM on May 19, 2010


here's another on the causality revolution: "there must therefore remain uncertainty and ambiguity about the breadth of application of any findings from randomized experiments..."

otoh maybe we can make computers do all the work: "At the heart of each of these new languages is a so-called inference algorithm, which instructs a machine-learning system how to draw conclusions from the data it’s presented." :P

cf. Identity Economics: "In standard economic models, individuals maximize their self-interest. In the identity economics model described below by George Akerlof ond Rachel Kranton, identities and norms are basic motivations."

viz. Joblessness and Perceptions about the Effectiveness of Democracy: "We find that personal joblessness experience translates into negative opinions about the effectiveness of democracy and it increases the desire for a rogue leader."

cheers!
posted by kliuless at 12:12 PM on May 19, 2010 [1 favorite]


Her work on microfinance is pretty interesting. There are a couple of examples in the New Yorker piece. I continue to be uneasy about microlending and she is one of the few people bringing a complete contextual analysis to the area - not just saying "this feels good and the money gets paid back, so yay!" but asking "does the availability of this loan make possible a long-term improvement in the conditions of the borrower's life?"
posted by Miko at 1:20 PM on May 19, 2010


I am reminded of Asimov's Foundation and psychohistory. (of course the speaker's methods seem more practical in the real world)
posted by RobotVoodooPower at 2:21 PM on May 19, 2010


koeselitz: A physicist would (a) try to determine unchanging rules – not tendencies, not statistical probabilities, but sureties; and (b) try to determine what exactly is going on at a much lower level. Yes, the result of this approach is that we're now able to discuss the motions of clouds of atoms by statistical analysis; but that result comes after we've attained the best possible understanding of what's going on inside objects at an atomic level.

This is how math works, but not science. You might in principle be able to derive all the rules of chemistry from physics, biology from chemistry, and so on. You might in principle be able to derive the rules of any kind of science from some more fundamental science, but science doesn't work that way in practice, and it's a good thing it doesn't. If it did, we couldn't study anything more complicated the fundamental particles until physicists work out a Grand Unified Theory. And even if we could discover such absolute rules for everything, they would be useless, because to apply them, we would need complete and perfect knowledge of the systems we wanted to analyze, right down to the subatomic level. The Uncertainty Principle means we can't have that level of knowledge even for a single particle, much less a molecule, a person, or a society.

tl;dr: Don't discount statistical rules, because they're the only rules we can discover and apply outside of pure mathematics.
posted by shponglespore at 11:57 AM on May 20, 2010


« Older The Guardian has an article on Pimm's, a tradition...  |  Anti-Identity-Theft Firm Lifel... Newer »


This thread has been archived and is closed to new comments