predicting civil unrest
January 27, 2011 10:01 AM Subscribe
Professors' global model forecasts civil unrest against governments - With protests spreading in the Middle East (now Yemen - not on the list) I thought this article and blog on a forecast model predicting "which countries will likely experience an escalation in domestic political violence [within the next five years]" was rather interesting.
BONUS
the 'youth bulge' (hehe) - "that an excess in especially young adult male population [who cannot find jobs] predictably leads to social unrest"
The Young and the Restless: Population Age Structure and Civil War - more on "youth bulges"
Civil War: Political Violence and Robust Settlements - game theoretical approaches combined with on the ground field studies to analyze war and conflict
Can Statistics Predict the Outcome of a War? - "A University of Georgia scientist has developed a statistical system that can, she claims, predict the outcome of wars with an accuracy of 80 percent."
BONUS
According to the profs, Egypt is ranked just behind Belgium on the forecast. Hmm.
posted by docgonzo at 10:09 AM on January 27, 2011 [1 favorite]
posted by docgonzo at 10:09 AM on January 27, 2011 [1 favorite]
Funny how they missed out Britain.
I have a real dislike for this kind of thing. You can make models say whatever you want.
posted by knapah at 10:17 AM on January 27, 2011 [2 favorites]
I have a real dislike for this kind of thing. You can make models say whatever you want.
posted by knapah at 10:17 AM on January 27, 2011 [2 favorites]
I hope the beer exports from Belgium won't be interrupted.
posted by furiousxgeorge at 10:18 AM on January 27, 2011 [2 favorites]
posted by furiousxgeorge at 10:18 AM on January 27, 2011 [2 favorites]
the 'youth bulge'
Wasn't that supposed to have remade Iran by now?
posted by Artw at 10:19 AM on January 27, 2011
Wasn't that supposed to have remade Iran by now?
posted by Artw at 10:19 AM on January 27, 2011
What is "political violence," according to the first set of profs? What violence really happened in Ireland recently, other than burning some placards at the march and some tussles in a march of 50,000 people? I was not under the impression that Ireland was in any real state of violent unrest. Wouldn't it be trivially easy to find "political violence" in any country, if you cast your net wide enough?
posted by Sticherbeast at 10:24 AM on January 27, 2011 [5 favorites]
posted by Sticherbeast at 10:24 AM on January 27, 2011 [5 favorites]
Related: In Tunisia, women play an equal role in revolution
Women in Tunisia are unique in the Arab world for enjoying near equality with men. And they are anxious to maintain their status.posted by filthy light thief at 10:31 AM on January 27, 2011 [1 favorite]
...
Najet, a criminal lawyer, explains the difference between Tunisian women and their sisters in the rest of the Arab world.
"We feel more free and more civilized than other Arab women," she says. "And especially since our revolution, we pity the women in neighboring countries. Look at Libya where they have to wear headscarves and can't even talk with men. This is a catastrophe."
Take predictions seriously only when they are made scientifically, about phenomena with some underlying quantifiable mechanism. Testable hypotheses are necessary.
Predictions about historical particulars like wars can only be "tested" by seeing if the event occurs. At that point, the model is unnecessary anyway.
These people are not testably hypothesizing. This is not science; Nostradamuses should be ignored.
posted by jjray at 10:33 AM on January 27, 2011 [4 favorites]
Predictions about historical particulars like wars can only be "tested" by seeing if the event occurs. At that point, the model is unnecessary anyway.
These people are not testably hypothesizing. This is not science; Nostradamuses should be ignored.
posted by jjray at 10:33 AM on January 27, 2011 [4 favorites]
Nostradamii
posted by dougrayrankin at 10:34 AM on January 27, 2011 [5 favorites]
posted by dougrayrankin at 10:34 AM on January 27, 2011 [5 favorites]
Nostradamodes.
posted by Sticherbeast at 10:35 AM on January 27, 2011 [5 favorites]
posted by Sticherbeast at 10:35 AM on January 27, 2011 [5 favorites]
Nostradamodes
More like nostradameans amirite?
posted by thsmchnekllsfascists at 10:40 AM on January 27, 2011 [3 favorites]
More like nostradameans amirite?
posted by thsmchnekllsfascists at 10:40 AM on January 27, 2011 [3 favorites]
Predictions about historical particulars like wars can only be "tested" by seeing if the event occurs. At that point, the model is unnecessary anyway.
This is like saying I shouldn't get a prognosis from a doctor because either I'll die or a won't, at which point I'll know.
posted by DU at 10:46 AM on January 27, 2011
This is like saying I shouldn't get a prognosis from a doctor because either I'll die or a won't, at which point I'll know.
posted by DU at 10:46 AM on January 27, 2011
Free Belgium! Down with the oppressive King Albert II! No longer should Belgians live under the tyranny of the Bicycle King! End the 30 year ban on Shimano breaks imposed by the ruling Flemish elite. In Belgium you can only find jobs as a research assistant to a major pharmaceutical, an accountant for a global tax advisory firm or a civil engineer. Whoa is it to the Belgium who after years of subsidized college and healthcare emerges into a marketplace where bonuses are only sort of fair! This is the face of corruption, viva revolution!
posted by geoff. at 10:49 AM on January 27, 2011 [1 favorite]
posted by geoff. at 10:49 AM on January 27, 2011 [1 favorite]
Nostradumbass
posted by absalom at 10:49 AM on January 27, 2011 [4 favorites]
posted by absalom at 10:49 AM on January 27, 2011 [4 favorites]
(In all seriousness, the reason it is ranked so high is because of the150+ days they spent without a government)
posted by geoff. at 10:50 AM on January 27, 2011
posted by geoff. at 10:50 AM on January 27, 2011
The New Yorker: Le Divorce: Why Belgium Has Never Been More Disunited.
(Behind the paywall, but an interesting read.)
posted by Floydd at 10:51 AM on January 27, 2011 [3 favorites]
(Behind the paywall, but an interesting read.)
posted by Floydd at 10:51 AM on January 27, 2011 [3 favorites]
These people are not testably hypothesizing.
I agree that this is not good science, but I disagree that their claims are not testable. Nothing is easier to test than a system that makes predictions. The tricky part is coming up with reasonable standards for what can be considered "unrest."
As for predicting success in military engagements, that sounds absolutely absurd. The reason that they arrived at 80% is probably because a lot of conflicts are pretty lopsided. What is the success rate of military analysts? I'm guessing it's quite a bit better. There are simply too many variables for me to imagine that this is achievable. The same is probably true of civil unrest, though perhaps to a slightly smaller degree. Both circumstances can hinge on emergent technologies.
posted by Edgewise at 10:52 AM on January 27, 2011
I agree that this is not good science, but I disagree that their claims are not testable. Nothing is easier to test than a system that makes predictions. The tricky part is coming up with reasonable standards for what can be considered "unrest."
As for predicting success in military engagements, that sounds absolutely absurd. The reason that they arrived at 80% is probably because a lot of conflicts are pretty lopsided. What is the success rate of military analysts? I'm guessing it's quite a bit better. There are simply too many variables for me to imagine that this is achievable. The same is probably true of civil unrest, though perhaps to a slightly smaller degree. Both circumstances can hinge on emergent technologies.
posted by Edgewise at 10:52 AM on January 27, 2011
I have no problem believing that Italy made number 12 on that list and that frightens me greatly. There is a breaking point, unfortunately, and Italy has a history of violent domestic terrorism.
Then again, I am surprised not to see the US on that list, given the recent events in Arizona. And by recent events, I mean the shooting of the Flores family in May 2009, allegedly by armed border patrol members, for one.
posted by lydhre at 10:58 AM on January 27, 2011
Then again, I am surprised not to see the US on that list, given the recent events in Arizona. And by recent events, I mean the shooting of the Flores family in May 2009, allegedly by armed border patrol members, for one.
posted by lydhre at 10:58 AM on January 27, 2011
I'm with Sticherbeast. From the article: "It contains the frequency and intensity of domestic political violence from 1990-2009. According to Bell, this violence includes anything from a sit-in that turns into a physical altercation to an embassy bombing."
I've gotta say, a sit-in where somebody throws a punch (or gets trampled? Or gets hit by a police baton?) is a damn sight off from an embassy bombing. Now that Scotland Yard has that "kettling" technique, where it seems someone always ends up hurt, is every demonstration in London "violent"? And if so, is there any use to this metric at all?
And that's assuming it even is a testable, disprovable hypothesis.
posted by rkent at 11:05 AM on January 27, 2011
I've gotta say, a sit-in where somebody throws a punch (or gets trampled? Or gets hit by a police baton?) is a damn sight off from an embassy bombing. Now that Scotland Yard has that "kettling" technique, where it seems someone always ends up hurt, is every demonstration in London "violent"? And if so, is there any use to this metric at all?
And that's assuming it even is a testable, disprovable hypothesis.
posted by rkent at 11:05 AM on January 27, 2011
As a grad student, I was asked to help write the codebook for an immense data set on militarized international disputes for a rather prominent scholar. It was a joke. It relied on ridiculous assumptions, such as you can find totally accurate information (and all the information you need) by looking at a handful of sources (New York Times, Facts on File, etc.). The other grad student and I would research the same two countries to count the number of MIDs over the centuries and I would often have 3 or 4 times as many as he did because I used a much wider set of sources (and in a couple instances I had much more detailed knowledge because I'd spent my time doing something besides coming up with increasingly arcane regression formulas). So the scholar just threw out what I came up with and used the other student's research to set up the code book because that was a lot easier. This data set is very well established in the political science literature now and god knows how many stupid articles have been written using it.
I have no idea how the data set was put together for this material, but when a list of countries that might have increased political violence doesn't include, say, Nepal or India, I'm very wary of giving it any validity. I looked at a couple of their so-called successful predictions and they seem so vague as to be useless. The reason for any violence in Ireland was different than what they predicted, so they're not actually predicting anything. I'm sure their web site will be updated to claim Egypt as another successful prediction.
Behavioralism has meant the death of critical thinking in political science. It's the stupidest fucking discipline on the planet.
posted by williampratt at 11:05 AM on January 27, 2011 [12 favorites]
I have no idea how the data set was put together for this material, but when a list of countries that might have increased political violence doesn't include, say, Nepal or India, I'm very wary of giving it any validity. I looked at a couple of their so-called successful predictions and they seem so vague as to be useless. The reason for any violence in Ireland was different than what they predicted, so they're not actually predicting anything. I'm sure their web site will be updated to claim Egypt as another successful prediction.
Behavioralism has meant the death of critical thinking in political science. It's the stupidest fucking discipline on the planet.
posted by williampratt at 11:05 AM on January 27, 2011 [12 favorites]
A friend in Turkey write today to tell me that all is well there, and yet, a substantial Muslim population, pretty much running things but allowing a very flourishing economy to grow with secular rule
The question is not so much will there be revolutions but who or what will replace those rulers who might be toppled:
the army? Sharia law? Jihadists?
Hezbollah, for example, now pretty much in control of Lebanon. What will they do?
posted by Postroad at 11:20 AM on January 27, 2011
The question is not so much will there be revolutions but who or what will replace those rulers who might be toppled:
the army? Sharia law? Jihadists?
Hezbollah, for example, now pretty much in control of Lebanon. What will they do?
posted by Postroad at 11:20 AM on January 27, 2011
You say Nostradamato, I say Nostradamahto.
Let's call the whole thing off.
posted by MrVisible at 11:31 AM on January 27, 2011
Let's call the whole thing off.
posted by MrVisible at 11:31 AM on January 27, 2011
This is the holy grail of social science: to be able to predict wars, especially revolutions, before they break out. One of my sociology professors long ago mentioned that revolutions tended to occur after the populace of a given country had seen its standard of living first increase appreciably, hold steady for a while, then decrease. People get really pissed off when the economic gains they've made are taken away from them, it seems.
I only wish I had asked for some citation, clarification or something on that little factoid (or that I'd paid more attention - it's possible a citation was given at the time) because it has stuck with me through the years.
Off to finish reading all the links in the OP now.
posted by Marla Singer at 11:44 AM on January 27, 2011
I only wish I had asked for some citation, clarification or something on that little factoid (or that I'd paid more attention - it's possible a citation was given at the time) because it has stuck with me through the years.
Off to finish reading all the links in the OP now.
posted by Marla Singer at 11:44 AM on January 27, 2011
It relied on ridiculous assumptions, such as you can find totally accurate information (and all the information you need) by looking at a handful of sources (New York Times, Facts on File, etc.).
Nope. It didn't rely on the NYT having all the information and totally accurate information. It relied on the NYT having a similar set of biases across all or most dyads, so that you weren't (much) more likely to pick up a MID in a dyad just because one of the countries involved happens to have better internationally-available press.
The other grad student and I would research the same two countries to count the number of MIDs over the centuries and I would often have 3 or 4 times as many as he did because I used a much wider set of sources (and in a couple instances I had much more detailed knowledge because I'd spent my time doing something besides coming up with increasingly arcane regression formulas).
That doesn't make your data better. If anything, it makes it worse, because now whether or not a MID appears in the data is largely dependent on how familiar you are with the countries of the relevant dyad. It means that whether a MID gets counted depends in large part on how good your information-searching skills were that day, or on how much time that day you could spend finding your extra sources. And it means that whether a MID appears or not depends in large part on whether or not information about that MID is available in a language you happen to understand. If you understand English, Spanish, and French, then you're very likely to relatively overcount MIDs in Latin America and Africa but relatively undercount MIDs in central Europe or Asia simply because you'd be unable to process the information from there.
The important factor is much less "DID WE COUNT EVERY MID THAT WE COULD POSSIBLY HAVE COUNTED?" and much more "DID WE USE A CONSISTENT DECISION RULE TO DECIDE WHETHER OR NOT TO COUNT A MID? AND IDEALLY WAS THIS DECISION RULE NOT MADE BY US BUT BY PEOPLE WHO HAVE NO INTEREST IN OUR PROJECT?"
posted by ROU_Xenophobe at 12:08 PM on January 27, 2011 [1 favorite]
Nope. It didn't rely on the NYT having all the information and totally accurate information. It relied on the NYT having a similar set of biases across all or most dyads, so that you weren't (much) more likely to pick up a MID in a dyad just because one of the countries involved happens to have better internationally-available press.
The other grad student and I would research the same two countries to count the number of MIDs over the centuries and I would often have 3 or 4 times as many as he did because I used a much wider set of sources (and in a couple instances I had much more detailed knowledge because I'd spent my time doing something besides coming up with increasingly arcane regression formulas).
That doesn't make your data better. If anything, it makes it worse, because now whether or not a MID appears in the data is largely dependent on how familiar you are with the countries of the relevant dyad. It means that whether a MID gets counted depends in large part on how good your information-searching skills were that day, or on how much time that day you could spend finding your extra sources. And it means that whether a MID appears or not depends in large part on whether or not information about that MID is available in a language you happen to understand. If you understand English, Spanish, and French, then you're very likely to relatively overcount MIDs in Latin America and Africa but relatively undercount MIDs in central Europe or Asia simply because you'd be unable to process the information from there.
The important factor is much less "DID WE COUNT EVERY MID THAT WE COULD POSSIBLY HAVE COUNTED?" and much more "DID WE USE A CONSISTENT DECISION RULE TO DECIDE WHETHER OR NOT TO COUNT A MID? AND IDEALLY WAS THIS DECISION RULE NOT MADE BY US BUT BY PEOPLE WHO HAVE NO INTEREST IN OUR PROJECT?"
posted by ROU_Xenophobe at 12:08 PM on January 27, 2011 [1 favorite]
I only wish I had asked for some citation, clarification or something on that little factoid (or that I'd paid more attention - it's possible a citation was given at the time) because it has stuck with me through the years.
Search for "relative deprivation."
posted by ROU_Xenophobe at 12:10 PM on January 27, 2011 [1 favorite]
Search for "relative deprivation."
posted by ROU_Xenophobe at 12:10 PM on January 27, 2011 [1 favorite]
Power threatened fears history. It can control the science and the art to some extent.
Push start.
The 'revolution' to explore does not involve guns or plots or secrets. One central premise is that everyone involved tends to not want to bomb another
anymore.
posted by clavdivs at 12:26 PM on January 27, 2011
Push start.
The 'revolution' to explore does not involve guns or plots or secrets. One central premise is that everyone involved tends to not want to bomb another
anymore.
posted by clavdivs at 12:26 PM on January 27, 2011
This is funny to me because a long time ago I worked on (and abandoned) a poem that had this image in it: A sort of weather man on the evening news, gesturing to a world map and forecasting outbreaks of pogroms as if they were patches of bad weather. It seemed like an absurd idea at the time. Now it doesn't seem so implausible.
(Presumably this new meteorology of human misery will take some time before it can approach the pin-point accuracy of modern weather forecasting.)
Also: Nostradammit!
posted by saulgoodman at 12:59 PM on January 27, 2011
(Presumably this new meteorology of human misery will take some time before it can approach the pin-point accuracy of modern weather forecasting.)
Also: Nostradammit!
posted by saulgoodman at 12:59 PM on January 27, 2011
This is bad 'science' journalism.
The fact that countries that experience unrest all appear on the list, is not the same thing as all countries that experience violence will appear on the list. But the reporting makes it seem that the model predicts the latter.
posted by carter at 1:05 PM on January 27, 2011
The fact that countries that experience unrest all appear on the list, is not the same thing as all countries that experience violence will appear on the list. But the reporting makes it seem that the model predicts the latter.
posted by carter at 1:05 PM on January 27, 2011
There's been civil unrest in Ireland? No-one told me.
posted by a womble is an active kind of sloth at 1:09 PM on January 27, 2011 [1 favorite]
posted by a womble is an active kind of sloth at 1:09 PM on January 27, 2011 [1 favorite]
gonna have to be more than a few random hits to be taken seriously by anyone thinking it through. Just cause your watch says 3:30 doesn't mean it will say 3:31 in 60 seconds time.
The internet is such a big place I'm sure I can find predictions that a giant ant infested asteroid will decimate Togo in 2013
posted by edgeways at 1:31 PM on January 27, 2011 [1 favorite]
The internet is such a big place I'm sure I can find predictions that a giant ant infested asteroid will decimate Togo in 2013
posted by edgeways at 1:31 PM on January 27, 2011 [1 favorite]
I, for one, welcome our new giant ant overlords.
posted by Marla Singer at 1:57 PM on January 27, 2011
posted by Marla Singer at 1:57 PM on January 27, 2011
between the ants and the Tea Party I'd gladly support the local Formicidae chapter.
posted by edgeways at 2:14 PM on January 27, 2011
posted by edgeways at 2:14 PM on January 27, 2011
Want to know the next revolutions? Just find the dictators the US is supporting. I can name three; Tunisia, Egypt and Yemen. There are many more on the list.
posted by shnarg at 2:19 PM on January 27, 2011
posted by shnarg at 2:19 PM on January 27, 2011
ROU_Xenophobe : FWI, outside of youtube and 4chan, it's considered quite rude to use all caps.
It didn't rely on the NYT having all the information and totally accurate information.
Actually, the scholar did specifically say that if a MID wasn't in the NYT or the other "accepted mainstream sources," it simply wasn't important enough to be included. He maintained this even when I pointed out that one MID the NYT and other sources didn't mention at all resulted in >100,000 battle deaths.
It relied on the NYT having a similar set of biases across all or most dyads,
No, he specifically said that if you use the NYT and the other "accepted" sources, there would be no significant bias. It was easy for me to show that, for example, those sources consistently downplayed/ignored MIDs initiated by the US and its friends (or countries perceived as democracies) while significantly highlighting those initiated by official enemies of the US (or those perceived as authoritarian or communist). Even if the biases are consistent, as they were, if you do not correct for them and are oblivious to their existence, then you have a flawed data set and you cannot actually draw any scientific conclusions from it.
so that you weren't (much) more likely to pick up a MID in a dyad just because one of the countries involved happens to have better internationally-available press.
I'm not following your point. With the resulting codebook, MIDs were being identified based upon their appearance in US major media. What that means, is that MIDs in a dyad were being picked up or not in strong correlation with US interests, either in general terms or very specifically. This consistent bias affected everything about the MIDS.
That doesn't make your data better.
Your points about not consistently using a codebook are correct, but beside the point. We were putting together the code book, not using one. What we had to go on was his definition of MID. On the basis of our research, the code book was then going to be put together. We both used the definition consistently, but I got all of the other student's MIDs plus many others that he missed (to which there was a consistent pattern). My results were thrown out because 1) it was asserted that there was no problem getting rid of them and only using US major media sources, and 2) it would be a big pain to do research beyond those main sources and if he didn't get the data set done on time he wouldn't get another big grant. Based on this initial research foray, a codebook was put together which was concerned about facilitating easy, cheap, income-producing research, not accurate research. Not actual data.
In this case, yes my data was better. The results were much more consistent with reality. Had they provided the basis for a code book it would have resulted in a much better data set. One can make all kinds of pragmatic arguments about limiting the number of sources, but that doesn't change the problems with the resulting data set. At any rate, with one exception I was not an expert in any of the dyads and did not have research proficiency in any of the local languages. We looked at a lot of dyads. I didn't do much more than look at one or two of the major english-language papers from the region to come up with my MIDs.
DID WE USE A CONSISTENT DECISION RULE TO DECIDE WHETHER OR NOT TO COUNT A MID
Sorry, I don't feel like re-typing to get rid of your caps. I'm not actually yelling back at you. It is in fact a big problem to use a consistently biased methodology to investigate a phenomenon on the assumption that this will somehow result in objective, empirical data which can then be scientifically applied to every single example of that phenomenon in some form of predictive model. It's about putting a scientific gloss on received information and assumptions, and perpetuating certain interests. It ends up being a mockery of actual science. That's a consistent problem with behavioralism in political science, along with its opposition to critical thinking. It's not without value but it has done far more damage than good.
WAS THIS DECISION RULE NOT MADE BY US BUT BY PEOPLE WHO HAVE NO INTEREST IN OUR PROJECT?
I actually didn't have any interest at all in the project. I was only involved because I got paid to do research for 4 or 5 months.
This is bad 'science' journalism.
Perhaps, but it also seems like bad science. I can't get to their site right now but I when I could I noticed they didn't have anything about Thailand. When whatever happened in Ireland last year gets counted as "political violence" and "affirmation" of their model, but what happened in Thailand does not, it suggests there's some confirmation bias.
posted by williampratt at 2:57 PM on January 27, 2011 [3 favorites]
It didn't rely on the NYT having all the information and totally accurate information.
Actually, the scholar did specifically say that if a MID wasn't in the NYT or the other "accepted mainstream sources," it simply wasn't important enough to be included. He maintained this even when I pointed out that one MID the NYT and other sources didn't mention at all resulted in >100,000 battle deaths.
It relied on the NYT having a similar set of biases across all or most dyads,
No, he specifically said that if you use the NYT and the other "accepted" sources, there would be no significant bias. It was easy for me to show that, for example, those sources consistently downplayed/ignored MIDs initiated by the US and its friends (or countries perceived as democracies) while significantly highlighting those initiated by official enemies of the US (or those perceived as authoritarian or communist). Even if the biases are consistent, as they were, if you do not correct for them and are oblivious to their existence, then you have a flawed data set and you cannot actually draw any scientific conclusions from it.
so that you weren't (much) more likely to pick up a MID in a dyad just because one of the countries involved happens to have better internationally-available press.
I'm not following your point. With the resulting codebook, MIDs were being identified based upon their appearance in US major media. What that means, is that MIDs in a dyad were being picked up or not in strong correlation with US interests, either in general terms or very specifically. This consistent bias affected everything about the MIDS.
That doesn't make your data better.
Your points about not consistently using a codebook are correct, but beside the point. We were putting together the code book, not using one. What we had to go on was his definition of MID. On the basis of our research, the code book was then going to be put together. We both used the definition consistently, but I got all of the other student's MIDs plus many others that he missed (to which there was a consistent pattern). My results were thrown out because 1) it was asserted that there was no problem getting rid of them and only using US major media sources, and 2) it would be a big pain to do research beyond those main sources and if he didn't get the data set done on time he wouldn't get another big grant. Based on this initial research foray, a codebook was put together which was concerned about facilitating easy, cheap, income-producing research, not accurate research. Not actual data.
In this case, yes my data was better. The results were much more consistent with reality. Had they provided the basis for a code book it would have resulted in a much better data set. One can make all kinds of pragmatic arguments about limiting the number of sources, but that doesn't change the problems with the resulting data set. At any rate, with one exception I was not an expert in any of the dyads and did not have research proficiency in any of the local languages. We looked at a lot of dyads. I didn't do much more than look at one or two of the major english-language papers from the region to come up with my MIDs.
DID WE USE A CONSISTENT DECISION RULE TO DECIDE WHETHER OR NOT TO COUNT A MID
Sorry, I don't feel like re-typing to get rid of your caps. I'm not actually yelling back at you. It is in fact a big problem to use a consistently biased methodology to investigate a phenomenon on the assumption that this will somehow result in objective, empirical data which can then be scientifically applied to every single example of that phenomenon in some form of predictive model. It's about putting a scientific gloss on received information and assumptions, and perpetuating certain interests. It ends up being a mockery of actual science. That's a consistent problem with behavioralism in political science, along with its opposition to critical thinking. It's not without value but it has done far more damage than good.
WAS THIS DECISION RULE NOT MADE BY US BUT BY PEOPLE WHO HAVE NO INTEREST IN OUR PROJECT?
I actually didn't have any interest at all in the project. I was only involved because I got paid to do research for 4 or 5 months.
This is bad 'science' journalism.
Perhaps, but it also seems like bad science. I can't get to their site right now but I when I could I noticed they didn't have anything about Thailand. When whatever happened in Ireland last year gets counted as "political violence" and "affirmation" of their model, but what happened in Thailand does not, it suggests there's some confirmation bias.
posted by williampratt at 2:57 PM on January 27, 2011 [3 favorites]
Nostradamodes
More like nostradameans amirite?
I think it's more of an issue with nostradamedians.
posted by eviemath at 4:37 PM on January 27, 2011
More like nostradameans amirite?
I think it's more of an issue with nostradamedians.
posted by eviemath at 4:37 PM on January 27, 2011
I'm quite skeptical about the seeming equating of radicalism, social unrest, and domestic political violence. This does not add positively to my impression of the researchers' interpretive abilities either (from the "article" link):
For me, the most surprising thing was that human rights matter: If a country uses political imprisonment or political disappearances, domestic violence against that country is likely to increase.
If you count domestic terrorism, civil war, and repressive police in a totalitarian state cracking down on any sort of protest (which generally gets reported as "protest turns violent" or "riot"), all under the same umbrella of "domestic political violence", seems to me the above observation should be a no-brainer.
posted by eviemath at 5:02 PM on January 27, 2011
For me, the most surprising thing was that human rights matter: If a country uses political imprisonment or political disappearances, domestic violence against that country is likely to increase.
If you count domestic terrorism, civil war, and repressive police in a totalitarian state cracking down on any sort of protest (which generally gets reported as "protest turns violent" or "riot"), all under the same umbrella of "domestic political violence", seems to me the above observation should be a no-brainer.
posted by eviemath at 5:02 PM on January 27, 2011
Perhaps, but it also seems like bad science.
Yes, agreed. But the journalist can't tell the difference.
posted by carter at 5:08 PM on January 27, 2011
Yes, agreed. But the journalist can't tell the difference.
posted by carter at 5:08 PM on January 27, 2011
I assume this is the COW data or its successors, which are widely acknowledged as being problematic. But that doesn't mean that the things that you disagreed with were terrible decisions.
Actually, the scholar did specifically say that if a MID wasn't in the NYT or the other
"accepted mainstream sources," it simply wasn't important enough to be included.
Even if that's wrong for some MIDs, that's a consistent decision rule. A consistent decision rule is a good thing. Myself, I'd rather they used a set of "newspapers of record" scattered across the world than relying so heavily on the NYT, but the point still remains that in a large dataset you really, really want to make as sure as you can that there's a simple, clear decision rule about what's going to be included.
I'm not following your point. With the resulting codebook, MIDs were being identified based upon their appearance in US major media.
...applied to every dyad. As opposed to having some MIDs that don't appear in the NYT or other sources, but you can easily spot them because you happen to have read a book about one of the countries once, or they happen to have a newspaper indexed electronically in a language you know, while others weren't just because all the writing about them was in a language you didn't know.
Your points about not consistently using a codebook are correct, but beside the point. We were putting together the code book, not using one.
No, they're not beside the point at all. Especially in a multicoder environment, it's very important that all coders work in the same way. Ideally, drawing from a single rubric for coding. On the one hand, it's surprising that your prof didn't give you one... but it almost sounds like he did, and you just didn't like it.
In this case, yes my data was better.
Nope. You took data with a single clear, understandable slant to it and added multiple new layers of selection effects to it.
I didn't do much more than look at one or two of the major english-language papers from the region to come up with my MIDs.
This is my point exactly. In addition to sampling from the universe of NYT pieces, you're now oversampling areas that happen to have newspapers you could get at that were in English. This isn't curing any bias, it's just adding a new one.
I actually didn't have any interest at all in the project. I was only involved because I got paid to do research for 4 or 5 months.
You were still working for someone who did, as part of a project with definite research goals besides producing the dataset. There can be a real value in having the decision about whether or not to count a particular case made by someone from a distinct organization, for a completely different purpose, who neither knows nor cares what the research goals of the project are.
posted by ROU_Xenophobe at 9:02 PM on January 27, 2011
Actually, the scholar did specifically say that if a MID wasn't in the NYT or the other
"accepted mainstream sources," it simply wasn't important enough to be included.
Even if that's wrong for some MIDs, that's a consistent decision rule. A consistent decision rule is a good thing. Myself, I'd rather they used a set of "newspapers of record" scattered across the world than relying so heavily on the NYT, but the point still remains that in a large dataset you really, really want to make as sure as you can that there's a simple, clear decision rule about what's going to be included.
I'm not following your point. With the resulting codebook, MIDs were being identified based upon their appearance in US major media.
...applied to every dyad. As opposed to having some MIDs that don't appear in the NYT or other sources, but you can easily spot them because you happen to have read a book about one of the countries once, or they happen to have a newspaper indexed electronically in a language you know, while others weren't just because all the writing about them was in a language you didn't know.
Your points about not consistently using a codebook are correct, but beside the point. We were putting together the code book, not using one.
No, they're not beside the point at all. Especially in a multicoder environment, it's very important that all coders work in the same way. Ideally, drawing from a single rubric for coding. On the one hand, it's surprising that your prof didn't give you one... but it almost sounds like he did, and you just didn't like it.
In this case, yes my data was better.
Nope. You took data with a single clear, understandable slant to it and added multiple new layers of selection effects to it.
I didn't do much more than look at one or two of the major english-language papers from the region to come up with my MIDs.
This is my point exactly. In addition to sampling from the universe of NYT pieces, you're now oversampling areas that happen to have newspapers you could get at that were in English. This isn't curing any bias, it's just adding a new one.
I actually didn't have any interest at all in the project. I was only involved because I got paid to do research for 4 or 5 months.
You were still working for someone who did, as part of a project with definite research goals besides producing the dataset. There can be a real value in having the decision about whether or not to count a particular case made by someone from a distinct organization, for a completely different purpose, who neither knows nor cares what the research goals of the project are.
posted by ROU_Xenophobe at 9:02 PM on January 27, 2011
i guess i should append the disclaimer that i think social science deserves an asterisk, altho i do approve the use of (discriminate) quantitative methods :P
like i think if you took relative cohort sizes, gender imbalances and unemployment rates and were able to come up with valid measures (for comparison) of "coercion, coordination and capacity," as discussed in the article, then you could fit a curve that identifies 'at risk' regions for civil unrest (however defined) with a certain probability and given time frame, like with the weather or in epidemiology... i mean it's not psychohistory[1,2] or precognition[1,2]!
maybe prediction is too strong a word -- and you want to be esp careful on grail quests -- because it presents the veneer of accuracy, but the challenge of any model is to understand its weaknesses + i do think empirical field work and continual testing are required to place data in the 'proper' context.
posted by kliuless at 4:56 AM on January 28, 2011
like i think if you took relative cohort sizes, gender imbalances and unemployment rates and were able to come up with valid measures (for comparison) of "coercion, coordination and capacity," as discussed in the article, then you could fit a curve that identifies 'at risk' regions for civil unrest (however defined) with a certain probability and given time frame, like with the weather or in epidemiology... i mean it's not psychohistory[1,2] or precognition[1,2]!
maybe prediction is too strong a word -- and you want to be esp careful on grail quests -- because it presents the veneer of accuracy, but the challenge of any model is to understand its weaknesses + i do think empirical field work and continual testing are required to place data in the 'proper' context.
posted by kliuless at 4:56 AM on January 28, 2011
Re: 'what violence happened in Ireland recently?'
An increase in intercepted and detonated bombs and ambushes of police officers. Discovery of weopons stashes. If you'd lived in London when frequent bombings killed people, or in the UK during the active years of 'the troubles', you'd be afraid of what might happen. In those days, it was thought there were three equally bloody and insoluble conflicts in the world that people thought wouldn't be solved in their lifetime: South Africa, Palestine/Israel, and northern Ireland. Whatever else happened under the Labour government, two big monuments only stand: introduction of the minimum wage, and the 'Good Friday' agreement for peace in Northern Ireland. It is agreed policy on Northern Ireland to not fan the flames of the kind of attention-seeking idiot who wishes they could have been killing people during the troubles and likes extreme rhetoric by not giving them any attention: the increase in northern irish violence and intelligence service warnings has been carefully buried in those parts of the papers read only by the most serious, but the story is clearly there. For instance, there was a brief story about the northern irish police requesting british intelligence service help (brits=the enemy for republicans), then one about republicans refusing it, another about a negotiated agreement, a mention in one of the reports of attacks of the advances made since the help received from british intelligence services, etc. So it's there if you want to follow it, but you wouldn't notice if you weren't looking.
posted by maiamaia at 5:41 AM on January 28, 2011
An increase in intercepted and detonated bombs and ambushes of police officers. Discovery of weopons stashes. If you'd lived in London when frequent bombings killed people, or in the UK during the active years of 'the troubles', you'd be afraid of what might happen. In those days, it was thought there were three equally bloody and insoluble conflicts in the world that people thought wouldn't be solved in their lifetime: South Africa, Palestine/Israel, and northern Ireland. Whatever else happened under the Labour government, two big monuments only stand: introduction of the minimum wage, and the 'Good Friday' agreement for peace in Northern Ireland. It is agreed policy on Northern Ireland to not fan the flames of the kind of attention-seeking idiot who wishes they could have been killing people during the troubles and likes extreme rhetoric by not giving them any attention: the increase in northern irish violence and intelligence service warnings has been carefully buried in those parts of the papers read only by the most serious, but the story is clearly there. For instance, there was a brief story about the northern irish police requesting british intelligence service help (brits=the enemy for republicans), then one about republicans refusing it, another about a negotiated agreement, a mention in one of the reports of attacks of the advances made since the help received from british intelligence services, etc. So it's there if you want to follow it, but you wouldn't notice if you weren't looking.
posted by maiamaia at 5:41 AM on January 28, 2011
maiamai, I've lived in Ireland and England for most of my life. I'm not unaware of the sad history of violence in Northern Ireland and the terrorist campaigns of violence. I do not live there now, but I read newspapers from both countries every day. My point is that there is not currently civil unrest in Ireland, (which colloquially means The Republic of Ireland). As the data is based on information from 1990-2009 it was not really referring to 'The Troubles.
Hence I am arguing that his point 'that there is civil unrest' in Ireland' is complete nonsense and I am highly suspicious of his overall model.
posted by a womble is an active kind of sloth at 3:29 PM on January 31, 2011
Hence I am arguing that his point 'that there is civil unrest' in Ireland' is complete nonsense and I am highly suspicious of his overall model.
posted by a womble is an active kind of sloth at 3:29 PM on January 31, 2011
Pentagon's Prediction Software Didn't Spot Egypt Unrest - "America's military and intelligence agencies have spent more than $125 million in the last three years on computer models that are supposed to forecast political unrest. But if any of these algorithms saw the upheaval in Egypt coming, the spooks and the generals are keeping the predictions very quiet."
posted by kliuless at 12:11 PM on February 11, 2011
posted by kliuless at 12:11 PM on February 11, 2011
« Older "Our Shooter's A Military Man" | Debtris Newer »
This thread has been archived and is closed to new comments
posted by dougrayrankin at 10:07 AM on January 27, 2011 [2 favorites]