Insurgent Attacks Follow Mathematical Pattern 181
Hugh Pickens writes "Nature reports that data collected on the timing of attacks and number of casualties from more than 54,000 events across nine insurgent wars, including those fought in Iraq between 2003 and 2008 and in Sierra Leone between 1994 and 2003, suggest that insurgencies have a common underlying pattern that may allow the timing of attacks and the number of casualties to be predicted. By plotting the distribution of the frequency and size of events, the team found that insurgent wars follow an approximate power law, in which the frequency of attacks decreases with increasing attack size to the power of 2.5. This means that for any insurgent war, an attack with 10 casualties is 316 times more likely to occur than one with 100 casualties (316 is 10 to the power of 2.5). 'We found that the way in which humans do insurgent wars — that is, the number of casualties and the timing of events — is universal,' says team leader Neil Johnson, a physicist at the University of Miami in Florida. 'This changes the way we think insurgency works.' To explain what was driving this common pattern, the researchers created a mathematical model which assumes that insurgent groups form and fragment when they sense danger, and strike in well-timed bursts to maximize their media exposure. Johnson is now working to predict how the insurgency in Afghanistan might respond to the influx of foreign troops recently announced by US President Barack Obama. 'We do observe a complicated pattern that has to do with the way humans do violence in some collective way,' adds Johnson."
Uhuh (Score:5, Funny)
Re:Uhuh (Score:5, Insightful)
Re:Uhuh (Score:4, Informative)
Sadly,there's been an assload of data.
Re: (Score:2, Funny)
Indeed. You don't get insurgents without an occupying power*.
* For the semantic pedants: While technically insurgents could resist a domestic government, it's been the case in the 20th century and since that insurgent warfare is a response to invading forces.
Re:Uhuh (Score:5, Insightful)
Indeed. You don't get insurgents without an occupying power*.
* For the semantic pedants: While technically insurgents could resist a domestic government, it's been the case in the 20th century and since that insurgent warfare is a response to invading forces.
Uh, really, explain to me which were the invading forces that triggered a response from the following which are perhaps best known and most representatives of 20th insurgency: - UNITA insurgency during the Angolan Civil war) (Angola)
- Tamil Tigers (Sri Lanka)
- Lord's Resistance Army (Uganda)
- AFDL (Congo)
- FSLN/MILPAS/Contras (Nicaragua) - FMLF (El Salvador) - Shining Path (Peru) - Tupac Amaru (Peru) - EZLN (Mexico) - CPN-M (Nepal)
- India's Naxalite insurgents
- People's Mujahedin of Iran
Re:Uhuh (Score:5, Insightful)
So I'm not sure how useful this pattern is.
My dinner spending patterns might follow a mathematical pattern too. I spend 10 bucks on dinner a lot more often than I spend 100 bucks. Whoopee.
Re:Uhuh (Score:5, Interesting)
One could probably form a strong argument (perhaps even with a valid mathematical basis) that suggests that so-called "insurgent" actions have worn out their welcome, and news of them floats in a featureless sea of similar actions. It doesn't help the "insurgents'" cause that they have little record for being nice to their own people, so they can only garner support from the most polarised of those they choose to leave alive.
Media exposure is variable (Score:2)
Media channels have a fixed capacity. There's exactly this much first page in printed papers, only so many minutes of prime time news.
News reports don't depend linearly on atrocities committed, they are a mix where whatever is considered newsworthy is given a portion of the available space.
Re: (Score:3, Insightful)
Warning - a lot of things look like they follow a power law.
Exactly. And in case it doesn't fit into a power law, you can probably make it fit into a Gaussian distribution.
Re:Uhuh (Score:5, Funny)
Warning - a lot of things look like they follow a power law.
Exactly. And in case it doesn't fit into a power law, you can probably make it fit into a Gaussian distribution.
at which point it all becomes a blur
Re:Uhuh (Score:4, Funny)
Normally, you can.
Re: (Score:2)
Heck, even a gaussian distribution looks like power law if you look at a small enough piece. I.e. drop enough terms on the taylor series.
Reminds me of the "Chemist's Rule." There's always some number of points that fits the graph paper you happen to have.
Re: (Score:3, Funny)
Re: (Score:3, Interesting)
There is a whole cottage industry of trying to fit power laws to data and being amazed whenever it fits. I guess I don't understand this one though; it sounds like they're just saying small attacks are more numerous than large attacks, which would seem obvious. What am I missing?
That's not it, there is more (Score:5, Informative)
What they are saying is that regardless of culture, location, enviroment decade, reasons behind the conflict, etc., the relation between large and small attacks appears to be a constant.
That wasn't obvious at all.
Re: (Score:3, Insightful)
What they are saying is that regardless of culture, location, enviroment decade, reasons behind the conflict, etc., the relation between large and small attacks appears to be a constant.
I wonder if said constant relationship is the implicit criteria that we use to label the conflict an "insurgency" in the first place.
Re: (Score:3, Informative)
More interestingly, many things which are at their heart completely random follow power laws.
For example, the arrival time distributions of cosmic rays, or the energy distribution of those particles one might observe. (ok, so I'm a cosmic ray physicist, so that's the topic I think a lot about). Thus, you can't use this information to predict anything about any one cosmic ray (or insurgent attack). What yo
Re: (Score:2)
Re: (Score:2)
The math in 'Numb3rs' is the IT version of "A GUI interface in VB to track an IP address" in CSI:NY...
There was a TED talk on this (Score:5, Informative)
Re:There was a TED talk on this (Score:5, Interesting)
Re: (Score:3, Insightful)
Re: (Score:2)
No, a war is a reciprocated armed conflict between two or more factions; these are just massacres.
Re: (Score:2, Interesting)
So World War Two didn't start when Germany took over Poland with almost no resistance? Good to know.
Re:There was a TED talk on this (Score:4, Informative)
Re: (Score:2)
Well, there is no fixed start date of World War II only fixed dates of the smaller wars inside the total conflict. It's true that France and the UK declared war on Germany when they invaded Poland but by the media of the time it was referred to as a ghost war or a phony war(the french called it a joke of a war). Ofcourse that all changed when Germany initiated "Fall Gelb" and attacked the Benelux in order to properly invade France.
The war activity went full force from that moment on. When the US bombed pear
Re:There was a TED talk on this (Score:4, Funny)
Unless there's a REALLY GOOD conspiracy theory out there that I don't know about, I think you mean "...the Japanese bombed Pearl Harbor..."
Re: (Score:2)
Mind boggles (Score:2)
Mind boggles reading garbage like this offered as at the end of the sentence as if it was a fact.
Do you consider over 80,000 killed within a span of a few weeks (15,000 Germans, over 65,000 Poles) as "almost no resistance"? How about the fact Germany lost nearly 700 tanks and over 500 aircraft in the same short amount of time? That is "almost no resistance"? Where did you learn history?
Re: (Score:2)
It is not a WAR. War can only be declared by congress so it must be avoided officially because it is not legal. Politically, it is a war and not a war whenever the need arises to be either - and the public falls for it.
Which leads to the next point: insurgents have good reason to be unhappy with the people who stand bye and do nothing which empowers their enemy. The military/police can also be unkind to those who help insurgents or do nothing - not that one can equate them, after all the people pay to suppo
Re:There was a TED talk on this (Score:5, Interesting)
Excellent point. But it make me question his definition of an insurgency.
Apparently, an insurgency that's crushed quickly doesn't count as an insurgency. And an insurgency that grows into a civil war doesn't count as an insurgency.
Only if the counter-insurgency is somewhat effective in reducing but not eliminating the number of attacks does he include it in his data set. In conclusion (and most remarkably) the data in his data set show a strong correlation across "insurgencies".
Re:There was a TED talk on this (Score:5, Insightful)
Which may lead to the conclusion that the 'law' that he found describes his inclusion concept (friendly version).
or
He fine-tuned his inclusion algorithm to the point that he could publish a valid 'law' and thus be eligible for DHS funding (reality insurges).
CC.
Re: (Score:2)
Thanks for that; I *KNEW* this sounded familiar and had seen it, or something like it, once before.
Re: (Score:2)
Nothing new under the sun. What he's saying it's that you shouldn't waste your efforts either in doing too many attacks that kill a few people each, or in doing a few attacks to kill a huge bunch of people each.
Re: (Score:2)
Or you can nuke everyone and that will end fairly quickly as well...
Just pointing out an alternative option :)
I must be missing something (Score:5, Insightful)
I don't see what it is they think they've discovered. If you take a loose collection of 5000 people with a weak desire to cooperate you're going to get way more groupings of 10 than 100 than 1000. The desire for safety in numbers is offset by the risk of exposure by size. In fact I'd have drawn almost exactly their curve if somebody had asked what the distribution would look like.
If the likelihood of an event is a coupled with critical mass of groupings then the event distribution will follow pretty much the same curve.
If somebody understands what it is these folks found could you explain it.
Re:I must be missing something (Score:5, Informative)
Also being able to draw a straight line on a log log plot is all well and good but if you get the slope off by even a small amount you will soon be orders of magnitude off in your predictions. Thus while you might expect a power-law distribution from simple arguments getting the specific value is much more difficult.
Re:I must be missing something (Score:5, Interesting)
Hmm, well shame on me, I saw the talk existed but expected just a verbal representation of the article.
I had missed the point about stability around alpha. I have to admit the graphs of alpha vs events like the surge or elections are pretty interesting.
Equally interesting though is the rapid return to alpha=2.5. I guess the real question at this point would be: Can repeated examinations of alpha be used to measure the positive effect of a strategy or is it merely a measure of the temporary perturbation and inevitable return to 2.5 because humans are after all humans and 2.5 merely represents the steady state of humans desire for coalescence vs fragmentation.
In short it's a question of cause and effect. Would a different species have a different alpha that's just as stable because it's a reflection of their physiology and psychology.
The research is certainly more interesting than I originally credited, thanks.
Correlation (Score:3, Insightful)
Solid post. This comment
Can repeated examinations of alpha be used to measure the positive effect of a strategy or is it merely a measure of the temporary perturbation and inevitable return to 2.5 because humans are after all humans and 2.5 merely represents the steady state of humans desire for coalescence vs fragmentation.
is about the best "correlation vs. causation" post I've seen lately.
Correlation vs. Causation has turned into an overused meme IMO-- not around here, just digg and reddit.
Re: (Score:2)
I'm sure there's something important in the details, but power laws always turn up in statistically independent events.
Insurgent mathematics . . . (Score:5, Funny)
Insurgent: "Hey, chief, there's a big column of Americans coming! Let's skank 'em!"
Chief: "Hold on, let me get out my calculator . . . damn it! I should have paid more attention to the Linear Programming and Game Theory courses at the Madrasah! Go ahead and attack . . . then turn on CNN to see if we got any media exposure. And please bring me some more pencils and paper . . . this mathematically based insurgency strategy *really* sucks!"
Re:Insurgent mathematics . . . (Score:5, Funny)
Re: (Score:3, Funny)
Re: (Score:2)
Predicting humans (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Unless you compelety mapped my brain and had complete understanding of how information is procssed and the relationship between inputs and outpus, the model you create would be incomplete.
To get around the lack of understanding of very complex systems, we use statistics to determine the probability o
Re: (Score:2)
The ones organizing those attacks are somewhat a small group of people, and if they get aware of this they could change behaviour.
Re: (Score:2)
Things will (and already begin to, IMHO) change once you have enough (micro-)data regarding an individual. Imagine you have a complete history of a person sampled at a small enough scale (including physiological variables). You might then use a 'single-case' approach on a rich real life dataset lending itself to statistical analysis of sorts.
CC.
Re: (Score:2)
Re: (Score:2)
No, having been a psychologist in a former life
Which then will further complicate the notion of 'free will' [stanford.edu].
CC.
Location? (Score:2)
Re: (Score:2)
Troop surges concern several ten thousand people, so it's almost impossible to keep them under wraps until they actually happen. The word will get out and around early enough for the insurgents to do what they do: adapt.
If they intend to hit everyday targets, possibly focusing on civilians or military installations, moving the attack to an earlier date would help their chances. Their insights on the current security forces around will stay the same, so they can get by with minimal changes in planning. If th
MMPI Comparison (Score:2)
This is a really interesting article. Although the point out the weaknesses of the theories behind the attacks, it is interesting that there is a pattern at all. Perhaps one way to look at this might be the same way we determined the validity of the Minnesota Multiphasic Personality Inventory (MMPI). Basically, the researchers looked for patterns within certain target groups (e.g., depressed, schizophrenic, bipolar) and detected answer patterns. In the same way, detecting the patterns of attack without at
Hmm (Score:2)
It will be interesting to see if there is any real "predictive" value behind this hypothesis. There's only one way to find out, and that's waiting to see if FUTURE (not past) data correlates with the model. Then there's the whole argument about the model itself changing the way insurgents are dealt with (since I assume the security forces aren't going to sit around and wait for people to die if the model predicts a "high" probability) and thus changing the expected results. And what about people being ident
Re: (Score:2)
It will be interesting to see if there is any real "predictive" value behind this hypothesis. There's only one way to find out, and that's waiting to see if FUTURE (not past) data correlates with the model.
Yes, that will be interesting.
And what about people being identified as "terrorists" on circumstantial evidence strictly because of the "higher probability" of an imminent attack?
The researchers didn't suggest that their model can or should be used to identify terrorists. You seem to be sliding down a slippery slope.
Past Data (Score:2)
I know that if I continuously flip a coin that it will come up "heads" about one half of the time.
But, that does not mean I know whether the next flip will be "heads" or "tails".
Re: (Score:2)
Your case assumes true randomness. We're dealing with a large group of humans, and humans have tendencies. This is more like predicting what the caller will say based on data like previous flips...
Of course, my example is still overly simplified, but closer to the intent and idea of the article... I think.
Re: (Score:3, Insightful)
Of course, you are correct. Thanks for the additional input.
My position was though was that the randomness instead of being 50-50 like the coin flip might instead be 60-40 or even 70-30 but still a large unknown exists to predict with any sense of integrity of accuracy.
Perhaps if it collapsed nearer to 100% such as 95-5 or 98-2 then I could see a usefulness for predicating a potential.
It does remind me of the old adage "Figures lie and liers figure"
Just Biology (Score:5, Insightful)
The result is cool, and important in the details, but is not that interesting in terms of breaking new ground. As a biologist, having measured countless number of behavioral parameters that all follow power laws, it is not surprising that yet another biological behavior, waging a particlar kind of war in this case, follows a power law. That part is ho-hum.
Similarly it would only surprise me if things like, oh, the size of undergraduate populations at different universities, the number of cars in each country, the number of stray dogs in each city, the average brain mass for each species, or the number of bullets used in any given firefight, do NOT follow a power law. It's just biology. That's the way things work.
And, to keep things in perspective, I'm just a biologist. It could be that all natural phenomena follow that sort of pattern, like the mass of celestial objects, the surface areas of land masses, the percent cloud cover at each point on Earth, etc. The basic idea of power laws -- lots of small versions of a thing, only a few big ones, and a smooth distribution between -- seems inherently universal to my small brain.
Re: (Score:2)
And, to keep things in perspective, I'm just a biologist. It could be that all natural phenomena follow that sort of pattern, like the mass of celestial objects, the surface areas of land masses, the percent cloud cover at each point on Earth, etc. The basic idea of power laws -- lots of small versions of a thing, only a few big ones, and a smooth distribution between -- seems inherently universal to my small brain.
Yeah, it is observed in non-biological systems [wikipedia.org], too. Interesting to note that power laws help explain why Benford's law exists.
Re: (Score:3, Interesting)
The reason the basic idea sounds familiar to not just you but everybody here is that it is the characterizing property of fractals [wikipedia.org]. I wouldn't go so far as to relate this idea to biology per se, however. It commonly occurs in physics as well.
Intuitively, fractals (and therefore power laws) ought to arise whenever a finite resource i
Caveat in re: power laws in empirical data (Score:5, Interesting)
Cosma Shalizi [cmu.edu] rants a lot about scientists' (often physicists') claims about having found a power law description of some empirical phenomenon (upshot: finding a straight line on a log-log plot isn't enough). See the following:
http://cscs.umich.edu/~crshalizi/weblog/491.html [umich.edu]
http://cscs.umich.edu/~crshalizi/notebooks/power-laws.html [umich.edu]
The WERE following a pattern (Score:2)
Publishing this just upset the pattern. Knowledge should not be released to the public as they can use it for bad purposes!! They are just supporting the terrorists, warning them of their mistakes.
( just being sarcastic here.. people are stupid, they will ALWAYS follow patterns, and information should always be free )
Human Solidarity (Score:3, Interesting)
the way in which humans do insurgent wars — that is, the number of casualties and the timing of events — is universal
Did anyone else find it ironic that human solidarity was found in acts against human solidarity?
A more interesting pattern (Score:3, Interesting)
I wonder what mathematical laws are in play that results in the reported number of insurgents killed during any attack by coalition forces weirdly hovering around 30. Google "30 Taliban killed", or "30 insurgents killed", or "30 militants killed" and you see a lot results going all the way back when the wars were started. See this blog entry http://securitycrank.wordpress.com/2009/12/07/winning-the-war-30-taliban-at-a-time/ [wordpress.com] for more discussion.
Re:A more interesting pattern (Score:5, Interesting)
Ehhhh... I don't think so.
A series of searches of "x insurgents killed" yields:
2= 14,700
3= 30,700
4= 164,000
5= 20,000 results
10= 160,000
15= 64,000
20= 306,000
25= 41,000
30= 58,400
31= 10
32= 75,400
33= 4,460
34= 26,400
35= 36,000
40= 57,000
41= 484
42= 28,400
43= 9
44= 1
45= 9,180
I think it would be difficult to draw any conclusions about how many insurgents are killed at once. How do you decide when an incident starts and ends? Operations can last days. How close do they have to be to each other when they die? I can almost guarantee that we are taking out insurgents one by one or two by two for the most part. They don't run around in packs of 30, they sneak at night in pairs.
That's just my experience, though. Keep your fun little "23" theory.
-b
Re: (Score:2)
If true, it might have something to do with the max size of close social group that humans form; it falls somewhere in "up to 30" range.
Re: (Score:2)
That blog entry links to this article: Timeline Pakistan 2009 [wordpress.com]. Looking through those, I see no pattern at all. In fact, I'd argue that there is a stro
Hari Seldon. (Score:5, Insightful)
'nuff said.
Re: (Score:2)
Really? (Score:2)
So, the bloody results of combat can be quantified using a few formulas and tables?
I think a few wargame designers at Avalon Hill are shaking their heads and rolling their eyes right now.
Asimov was right! (Score:2)
Re: (Score:2)
Your comment is the most pertinent of all here, it's sad it didn't get noticed.
A Seldonist.
_Everything_ follows an approximate power law. (Score:2)
Consider, for example, the quality and snarkiness of comments on Slashdot.
Not surprising at all. (Score:4, Insightful)
Let's see. It takes more energy, time, and complexity, to move into place the resources needed for a bigger attack. So, its not really surprising at all that bigger attacks occur less frequently or even obey a power law.
Re: (Score:2)
Re: (Score:2)
It's not surprising that the occur less frequently, but that the exponent is the same across different wars and cultures did come as a surprise.
Not really. It's very likely the researchers threw out things and emphasized others to get the story that there was an exponential commonality to wars.
The 2.5 Exponent (Score:4, Interesting)
The value of the exponent is interesting. If one assumes that the smallest attacks happen roughly once a day then the attacks that are an order of magnitude larger happen about once a year. This implies that there may be some sort of calendar event that triggers these larger events. If these events can be identified then it may help avoid some of the large attacks. It would be interesting to check this by looking at the timing of the largest attacks in the data set that was used for this study.
Re: (Score:2)
While it may be true that some events are linked to events in a calendar, i believe you are interpereting what is actually said incorrectly.
Re: (Score:3, Interesting)
That is certainly true, but it would be interesting to see if there is some sort of periodicity, particularly considering that there are many different annual events and cycles that could affect insurgencies and the way that they plan and carry out attacks. The 2.5 exponent may be completely unrelated to the year, but it is interesting that it does roughly correspond to an order of magnitude larger attack on roughly annual timescales.
Re: (Score:2)
I am in no way trying to write off the suggestion that annual celebrations might have an effect on event severity/frequency. This is a very valid suggestion, but perhaps a little outside of what this research is tacklin
Re:The 2.5 Exponent (Score:4, Interesting)
Yes, we would see the same timescale regardless of the base that was used. The only difference would be the value of the exponent. The value of the exponent itself is not the key, it is the timescale that the exponent (in combination with the base) implies. The timescale may very well be a coincidence, but if it does merit some consideration to see if there is any evidence to suggest that the timescale is real. Fortunately, there are many tests that can be made to see if there is any evidence for some sort of periodicity or pseudo-periodicity. OF course, this whole idea falls apart if the timescale for the smallest attacks is significantly different from one day, which is another test of the hypothesis.
Re:The 2.5 Exponent (Score:5, Interesting)
These may be useful to you:
http://www.pbs.org/wgbh/pages/frontline/insurgency/etc/graph.html [pbs.org]
http://www.longwarjournal.org/archives/2007/12/iraq_by_the_numbers.php [longwarjournal.org]
I can't speak of afghanistan, but in iraq the insurgent attacks were higher and more effective:
-when the ground was dry (moving around in iraq during the rainy season is a nightmare)
-lots of blowing dust in the air, drastically reducing visibility
-around dusk
-toward the end of ramadan
That's just a taste of all the factors that you'd have to account for to get an accurate map of insurgent behavior. Even then, I think it'd be pretty useless, since they are not a regular army and do not usually coordinate among cells. Maybe they want to attack, but the shipment from libya isn't here yet, so they wait for that but now the americans are getting suspicious so they launch all 20 of their libyan mortars at once and high-tail it out of there. Seems like a major, coordinated attack when in reality things are very different.
Guaranteed to make your brain hurt.
-b
The Art of War (Score:3, Interesting)
Yea, who would have thought that war follows a predictable (even mathematical) pattern.
http://en.wikipedia.org/wiki/The_Art_of_War [wikipedia.org]
Re: (Score:2)
Re: (Score:3, Interesting)
yea, the guys with this study likly failed at both. I am not sure I would want to be the guy in the field getting shot at when it turns out they got one of the variables wrong (which from the article seems like they got more than a few wrong like this B.S. about the media).
My point was more aimed at the people that thought this was somehow a special discovery. The Art of War contains many specific (if not basic) formulas, mostly in regards to economics, about the nature of troop strengths, cost fielding tro
I wonder what the exponent would be... (Score:5, Insightful)
...if we brought them all home?
Re: (Score:2)
...if we brought them all home?
If we brought all of the insurgents home? I don't know if that's feasible. I mean, are they even housebroken?
Oh... you meant if we bring all our soldiers home. Well, since the insurgents generally attack the local government and civilians, I don't think their behavior would change very much.
Every collective human endeavor does this (Score:5, Interesting)
Power laws are ubiquitous in human affairs - almost everything we do as a group involves power laws. This works for the size of cities and the sale of books and traffic to web sites, so I am not surprised it also happens in insurgent attacks.
Whether that will actually result in the effectiveness of Army tactics is another question, and, frankly, I am dubious. The sale of hit records follows a power law, but knowing that doesn't make me into a better musician.
One more study to (Score:2, Insightful)
How does this help? (Score:2, Interesting)
Interesting... (Score:2, Insightful)
Patterns... (Score:2, Interesting)
there's an opportunity here (Score:2)
Re:Hello Captain Obvious! (Score:5, Insightful)
> The total amount of attack will/power stays the same, no matter what size
> the individual attacks are? No shit? I could have told you that too.
But you did not. I am constantly amazed that every time some sort of insightful discovery is
made there is a chorus of voices saying " I could have told you that". Wake me when someone
actually does "tell me that" before someone else publishes it.
Re: (Score:2, Funny)
Meh. I could have told you that.
Re:Psychohistory begins. (Score:5, Funny)
You're doing it wrong. There's no use in doing thousands of submissions if you don't follow the correct power law. An attack with ten submissions should be 316 times more likely to occur than an attack with 100 submissions.
Re: (Score:2)
At the end of any action ther