Bloggingheads Community

Bloggingheads Community (http://bloggingheads.tv/forum/index.php)
-   Diavlog comments (http://bloggingheads.tv/forum/forumdisplay.php?f=9)
-   -   Science Saturday: Summer Doldrums Edition (http://bloggingheads.tv/forum/showthread.php?t=1933)

Bloggingheads 07-05-2008 09:36 AM

Science Saturday: Summer Doldrums Edition
 
AFTERTHOUGHT

When George freezes in place during John's explanation of a study of the effects of psychedelic mushrooms, we expect it will totally blow some viewers' freakin' minds. The effect is due to a technical glitch, for which we apologize.

--BhTV staff

Simon Willard 07-05-2008 10:13 AM

More is Different all over again
 
These guys had a more extensive discussion of "More is Different" some months ago: http://bloggingheads.tv/diavlogs/759...2&out=00:44:42

Here is today's reference:
http://bloggingheads.tv/diavlogs/125...3&out=00:51:33

And here is Anderson's paper:
http://www.cmp.caltech.edu/~motrunch...sDifferent.pdf

Bloggin' Noggin 07-05-2008 12:12 PM

"Let Google Run the Country"
 
Actually, _The Wisdom of Crowds_ suggests that that's, in effect, what democracy does. Crowds are foxes.
Tetlock's book certainly seems like a good riposte to Bryan Caplan(who appeared on BHtv a year or more ago), whose book compared economic views of ordinary voters to the "economic experts," gave ordinary voters a failing grade on that basis and argued that democracy was a pretty lousy system.
Well, libertarian economists like Caplan -- are big flaming hedgehogs with the one big idea of ideally free markets and the "invisible hand". The complexities of the actual world are mere details to them. The data don't clearly show that raising the minimum wage a bit raises unemployment, but so what? Our beautiful theory shows us that this MUST be happening. The Great Depression didn't really happen, because Say's law says it couldn't happen... and so on.
While I'm free-associating, Robert H. Frank argues in _Luxury Fever_ that Economists need to supplement Adam Smith's view of competition with Darwin's more ambivalent and subtle view. (As Frank points out, even Smith wasn't as extreme a booster for competition as many modern economists.) Perhaps it's the libertarian economists who ought to get the "F" while Democracy gets a B-.

Ocean 07-05-2008 01:08 PM

Re: Science Saturday: Summer Doldrums Edition
 
Following George’s example, let’s say an insurance company has found a strong correlation between houses with red shingles and claims for fire damage. Should they be charging higher premiums for red shingles? Well, perhaps there are geographical areas, with very dry and hot weather where it is fashionable to use red shingles and there the correlation. Of course this example is obvious, but the reality is that databases only have a limited number of variables and the mediators or confounders may not be in there. These huge databases are useful for some purposes which don’t require theory, or for which theory is already known. But the most we can expect is to get some patterns and correlations which should then be tested experimentally after articulating hypothesis. These databases only point at some aspect of reality, but we need to be mindful about its limitations. I don’t think any serious scientist could possibly agree with Anderson’s position except for very narrow practical induction.
In medicine, it certainly has some application, but ultimately we would want to know the physiopathology behind the syndrome so that different treatments can be developed. How easily we get carried away with, once again, the promise of a shortcut!

Now if we are going to put things in perspective, we may agree that there may be some waste of money in medicine. But what about dollars per capita that go to the military, or for oil/gasoline, or clothes, or food, or any other commodity in our materialistic, capitalistic culture? If we are going to socialize medicine, that ultimately I would support, can we socialize everything else? Shouldn’t we pick our battles? Is “excess” in health care such a central issue?
Give me a break!

Bobby G 07-05-2008 01:33 PM

Re: "Let Google Run the Country"
 
Quote:

Originally Posted by Bloggin' Noggin (Post 82285)
Actually, _The Wisdom of Crowds_ suggests that that's, in effect, what democracy does. Crowds are foxes.
Tetlock's book certainly seems like a good riposte to Bryan Caplan(who appeared on BHtv a year or more ago), whose book compared economic views of ordinary voters to the "economic experts," gave ordinary voters a failing grade on that basis and argued that democracy was a pretty lousy system.

I read the Caplan book and came away with something like the same impression you did--namely, that he's making the big assumption that the economists are right about their stances on free trade, the minimum wage, immigration, and so on. I certainly think they have strong cases to make, but you're right, the real world is often more complex than what's dreamed about by economists.

However, the thesis stated more formally is: people in the large are subject to predictable, systematic irrationalities regarding certain kinds of policies (due to things like framing effects); consequently, the assumption that democracy usually gives us the right (= most efficient?) policy should be called into question. And that seems correct to me. Now if we could only specify with greater precision what these systematic irrationalities are, and what we should do to rectify them...

themightypuck 07-05-2008 03:43 PM

Re: Science Saturday: Summer Doldrums Edition
 
I suppose it is an empirical question whether the black boxes start making better predictions than the scientists. This won't change some natural human drive to synthesize or "understand" things (whatever understand means).

tomarsaigo 07-05-2008 04:22 PM

Re: Science Saturday: Summer Doldrums Edition
 
"Should we turn over government to cloud computers and get rid of all the politicians?"
-John

Did we learn nothing from that whole W.O.P.R episode? Sure, the W.O.P.R. got it right eventually, but that was a bit too close for my liking. I'm sticking with flawed, but human leadership. :)

Kevin 07-05-2008 04:25 PM

Re: Science Saturday: Summer Doldrums Edition
 
I wonder what would happen if..... you took one of the black boxes, and wrote additional programs that said "expose the underlying routes from A to B" as sentences, or flow charts or something. Would they be incomprehensible beyond hope, or could we work backwards with understanding as a goal in its own right? Could you write a translator that we do deeply follow, by setting up a google-style translator and then digging in?

AemJeff 07-05-2008 04:45 PM

Re: Science Saturday: Summer Doldrums Edition
 
Quote:

Originally Posted by Ocean (Post 82290)
Following George’s example, let’s say an insurance company has found a strong correlation between houses with red shingles and claims for fire damage. Should they be charging higher premiums for red shingles? Well, perhaps there are geographical areas, with very dry and hot weather where it is fashionable to use red shingles and there the correlation. Of course this example is obvious, but the reality is that databases only have a limited number of variables and the mediators or confounders may not be in there. These huge databases are useful for some purposes which don’t require theory, or for which theory is already known. But the most we can expect is to get some patterns and correlations which should then be tested experimentally after articulating hypothesis. These databases only point at some aspect of reality, but we need to be mindful about its limitations. I don’t think any serious scientist could possibly agree with Anderson’s position except for very narrow practical induction.
In medicine, it certainly has some application, but ultimately we would want to know the physiopathology behind the syndrome so that different treatments can be developed. How easily we get carried away with, once again, the promise of a shortcut!

Now if we are going to put things in perspective, we may agree that there may be some waste of money in medicine. But what about dollars per capita that go to the military, or for oil/gasoline, or clothes, or food, or any other commodity in our materialistic, capitalistic culture? If we are going to socialize medicine, that ultimately I would support, can we socialize everything else? Shouldn’t we pick our battles? Is “excess” in health care such a central issue?
Give me a break!

In the first paragraph, you ask if insurance companies "should" take a particular action based on some measurement. It reads to me like you're asking for a provable causal connection between whatever is being measured and an effect that has actuarial consequences. I'd guess that an insurer would respond that just determining a correlation is enough to justify a decision to condition coverage somehow. In other words, if they see predictive value in a measurement, why should they care whether there's a deeper relationship?

Wonderment 07-05-2008 05:01 PM

George Johnson: Hero to non-flag-pin-wearing Americans everywhere
 
The Fourth of what?

jasonN. 07-05-2008 06:03 PM

Re: Science Saturday: Summer Doldrums Edition
 
john's point about the us having the most expensive health care and ranking 45th in life expectancy is misleading. the life expectancy statistic is not useful in comparing health care systems as the leader, Japan, has low rates of heart disease and cancer for largely unknown and possibly cultural reasons and its system isn't proven against a population with high rates of both such as the us. as far as the us having the most expensive health care being some sort of obvious negative, it ignores the obvious fact of our relative wealth compared to other countries. secondly, it ignores the extent to which us profits fund innovation and allow selling to other markets at lower prices, especially ones with nationalized care.

Ocean 07-05-2008 06:27 PM

Re: Science Saturday: Summer Doldrums Edition
 
Quote:

Originally Posted by AemJeff (Post 82311)
In the first paragraph, you ask if insurance companies "should" take a particular action based on some measurement. It reads to me like you're asking for a provable causal connection between whatever is being measured and an effect that has actuarial consequences. I'd guess that an insurer would respond that just determining a correlation is enough to justify a decision to condition coverage somehow. In other words, if they see predictive value in a measurement, why should they care whether there's a deeper relationship?

They should care because doing that is nonsensical and, frankly idiotic. A competitor insurer, being smarter and having figured out the reason, will be able to condition coverage to those that are at a real increased risk. It seems to me that it would be a more rational business. The problem is that the "cause" is not always evident. In those situations, shooting the bulk may be a temporary pseudo solution, but certainly not the intelligent one. This approach, applied to other areas, would have also very serious consequences. It seems to be that it would be like going back in time, to prescientific thinking, just with larger databases. Again, I acknowledge that there is a significant utility to this kind of approach, but there is no way it can substitute scientific method, including theory, of course. Now there is another deeper layer for this and has to do with the issue of the bias introduced by the observer. I think this would most likely be a problem for both traditional science and large database analysis. One could argue that we are really good at finding evidence to support our little stories.
And what about progress in science? If there is no theory, how do you move on to the next step? By gathering more data? How do you go about choosing what data to gather? Wouldn't you have to use some form of hypothesis or theory to make that decision? Going back to the insurance company, are they collecting data on weather, or proximity to flammable materials, etc? Why? Wouldn't that be driven by some kind of theory about what factors contribute to a fire?

uncle ebeneezer 07-05-2008 07:52 PM

Re: George Johnson: Hero to non-flag-pin-wearing Americans everywhere
 
Those Godless science writers are ruining America!! What next, George admits he doesn't like apple pie ;-)

AemJeff 07-05-2008 08:17 PM

Re: Science Saturday: Summer Doldrums Edition
 
Quote:

Originally Posted by Ocean (Post 82317)
They should care because doing that is nonsensical and, frankly idiotic. A competitor insurer, being smarter and having figured out the reason, will be able to condition coverage to those that are at a real increased risk. It seems to me that it would be a more rational business. The problem is that the "cause" is not always evident. In those situations, shooting the bulk may be a temporary pseudo solution, but certainly not the intelligent one. This approach, applied to other areas, would have also very serious consequences. It seems to be that it would be like going back in time, to prescientific thinking, just with larger databases. Again, I acknowledge that there is a significant utility to this kind of approach, but there is no way it can substitute scientific method, including theory, of course. Now there is another deeper layer for this and has to do with the issue of the bias introduced by the observer. I think this would most likely be a problem for both traditional science and large database analysis. One could argue that we are really good at finding evidence to support our little stories.
And what about progress in science? If there is no theory, how do you move on to the next step? By gathering more data? How do you go about choosing what data to gather? Wouldn't you have to use some form of hypothesis or theory to make that decision? Going back to the insurance company, are they collecting data on weather, or proximity to flammable materials, etc? Why? Wouldn't that be driven by some kind of theory about what factors contribute to a fire?

I'd say to you that an executive at our fictitious insurance company would be surprised to hear that he was being unscientific. He'd probably say that he was building an empirical model of the forces that affect his business, based on measurement, rather than a set of a priori assumptions. His business policies are probably not fixed, and as his database grows, if he's a smart businessman, he'll keep adjusting them based on what the data indicates. The competitor, on the other hand, who is, in a sense, trying to out-think the world, probably won't be as successful.

There's an interesting class of statistical methods devoted to this kind of analysis called Empirical Bayes Methods. One place where you're likely to see an application of this sort of thing is in your email SPAM filter, if you or your ISP uses one. This kind of modeling from data, rather than from pre-existing theories about what's going to be important, actually works really well if your aim is to respond to complex inputs. On the other hand, they don't answer any epistemic questions; and if your aim is to gain an organic understanding of how the world works, as opposed to, say, increasing the profitability of an insurance company, they really don't have much to offer.

Ocean 07-05-2008 09:31 PM

Re: Science Saturday: Summer Doldrums Edition
 
Quote:

Originally Posted by AemJeff (Post 82321)
I'd say to you that an executive at our fictitious insurance company would be surprised to hear that he was being unscientific. He'd probably say that he was building an empirical model of the forces that affect his business, based on measurement, rather than a set of a priori assumptions. His business policies are probably not fixed, and as his database grows, if he's a smart businessman, he'll keep adjusting them based on what the data indicates. The competitor, on the other hand, who is, in a sense, trying to out-think the world, probably won't be as successful.

There's an interesting class of statistical methods devoted to this kind of analysis called Empirical Bayes Methods. One place where you're likely to see an application of this sort of thing is in your email SPAM filter, if you or your ISP uses one. This kind of modeling from data, rather than from pre-existing theories about what's going to be important, actually works really well if your aim is to respond to complex inputs. On the other hand, they don't answer any epistemic questions; and if your aim is to gain an organic understanding of how the world works, as opposed to, say, increasing the profitability of an insurance company, they really don't have much to offer.

WOW!!!!
I was impressed with your reply!
I think we are talking about different issues. Not to mention we are looking at this from very different perspectives. First, I agree with you that in order to "filter out" or find patterns in very large, complex data, this approach is more practical. You get the "bulk". There are many practical applications, and yes, you can operate "blindly" drawing from the correlations to make some inferences. However, once you have these identified patterns (in science they are equivalent to observations), if you want to generalize to other, say, databases, or find other applications for them, you need to somehow "understand" those relationships to make informed predictions (hypothesis). My point is the difference between some guy called Newton saying "an apple fell on my head" (pure factual information) and developing a theory of gravity. Not to say that if Einstein hadn't engaged in a lot of pure a priori assumptions, we would still be happy with the falling apples. Talking about fruits, we are indeed talking about apples and oranges. Each has an application. A different taste for different customers...
By the way, I'm glad I'm not a business person. But I still think that my little rational insurance company would win at the end. That is if the huge brainless data crunching one doesn't buy it out!

AemJeff 07-06-2008 09:56 AM

Re: Science Saturday: Summer Doldrums Edition
 
Quote:

Originally Posted by Ocean (Post 82328)
WOW!!!!
I was impressed with your reply!
I think we are talking about different issues. Not to mention we are looking at this from very different perspectives. First, I agree with you that in order to "filter out" or find patterns in very large, complex data, this approach is more practical. You get the "bulk". There are many practical applications, and yes, you can operate "blindly" drawing from the correlations to make some inferences. However, once you have these identified patterns (in science they are equivalent to observations), if you want to generalize to other, say, databases, or find other applications for them, you need to somehow "understand" those relationships to make informed predictions (hypothesis). My point is the difference between some guy called Newton saying "an apple fell on my head" (pure factual information) and developing a theory of gravity. Not to say that if Einstein hadn't engaged in a lot of pure a priori assumptions, we would still be happy with the falling apples. Talking about fruits, we are indeed talking about apples and oranges. Each has an application. A different taste for different customers...
By the way, I'm glad I'm not a business person. But I still think that my little rational insurance company would win at the end. That is if the huge brainless data crunching one doesn't buy it out!

Thanks. I think you've hit the nail on the head - the difference is goals. A commercial entity like an insurance company has a different aim than a working theoretical scientist has. If your purpose is to do science that's a different thing from using science.

Paula Jones 07-06-2008 11:15 AM

Re: Science Saturday: Summer Doldrums Edition
 
Quoting Boggin' Noggin:Well, libertarian economists like Caplan -- are big flaming hedgehogs with the one big idea of ideally free markets and the "invisible hand". The complexities of the actual world are mere details to them

Actually, the idea of the "invisible hand" is more like the wisdom of the individuals within the crowd. They are the foxes, not the crowd as a whole. The "invisible hand" is the cumulative effect (predictions?) of the interactions of all of the foxes or ideas. The hedgehog would be the central authority dictating a policy (prediction).

I think that most libertarian economists are just pointing out that maybe we should be wary of the unintended (or intended for that matter) consequences
of dart throwing policy makers. On second thought, given what John said, maybe it would be a little better if they just threw darts.

Ocean 07-06-2008 12:56 PM

Re: Science Saturday: Summer Doldrums Edition
 
Quote:

Originally Posted by AemJeff (Post 82379)
Thanks. I think you've hit the nail on the head - the difference is goals. A commercial entity like an insurance company has a different aim than a working theoretical scientist has. If your purpose is to do science that's a different thing from using science.

True!

T.G.G.P 07-06-2008 09:40 PM

Re: "Let Google Run the Country"
 
I know I've abandoned the forums here for some time, but as I've unethically provided relevant materials you would otherwise have to pay for, I couldn't resist chiming in.

Bryan Caplan really thinks highly of Tetlock's book, but thinks it may give the wrong impression that laymen are as competent as experts, or even worse, moreso. His book review is in Jeffrey Friedman's journal, Critical Review. Friedman calls himself a "post-libertarian" who says those who tout the merits of the market or non-aggression axioms are on shaky ground. What actually is well established in his view is the deficient nature of democracies (his discussion of Converse in a previous issue gets to the meat of the matter). I provide the issue with Caplan's review here. I've got another post on Jeffrey Friedman here.

aranazo 07-07-2008 07:54 AM

Re: Science Saturday: Summer Doldrums Edition
 
To correct what I am almost certain is a mistake - one of the participants suggests Noam Chomsky is monolingual, I believe he is fluent in Hebrew as well as English.

nkirby 07-10-2008 09:45 PM

Re: Science Saturday: Summer Doldrums Edition
 
If, say, political experts are assumed to be masters of understanding the current and past political climate, is it any surprise that they would not foresee dramatic changes in the political climate? Things like the fall of the Soviet Union seem to be (not to sound too post-modern here) paradigm changing events.


All times are GMT -4. The time now is 12:16 AM.

Powered by vBulletin® Version 3.8.7 Beta 1
Copyright ©2000 - 2020, vBulletin Solutions, Inc.