So, it’s not a fluke. Remember when we discussed the Santa Cruz experiment to predict crime? So that the police could beef up patrols in those areas that were likely to have crimes perpetrated. Well, now STEM (science, technology, engineering, math) may have a valid method to predict incidents of war (not the incidence of war, though).
Dr. Guido Sanguinetti (University of Edinburgh), along with Drs. Zammit-Mangion, (Edinburgh & British Heart Institute) Dewar (Columbia) , and Kadirkamanathan (University of Sheffield) are about to publish their findings in the Proceedings of the National Academy of Sciences (PNAS, approved for print).They employed data from 2004 to 2009 to develop a model that predicted which regions would have violent episodes and which would not, including the levels of that violence. Something like this (or the Santa Cruz prototype) would never have been possible before computing power reached a critical threshold; rigorous quantitative analysis is now supplanting subjective suppositions.
Somehow DARPA has failed in this arena. Their Integrated Crisis Early Warning System (ICEWS) simply fails to predict many crises. And, this is the most “accurate” system the Pentagon has- older systems provide worse results.
The trick this group used was to avoid trying to discern military operations or political events- it just based all of its analysis assuming any sort of violence was equivalent. (In other words, it did not seek to predict a certain event, but all events in general.) And, that let them predict the 2010 situation very clearly- time, place, and intensity. (Yes, I know it’s 2012. But, they needed to test their theory- and verify it with real data!) This was true in the restive south and the relatively placid northern provinces (where there are fewer data points, besides) of Afghanistan. Among the data analyzed included population density, distance from major cities, terrain, and distance from Pakistan.
(On a political note, the study provided a very discouraging result. The surge would not dispel the insurgency. The other political result- without the WikiLeaks Afghan War Diary data, this study would have been impossible to obtain, analyze, and determine the model’s usefulness. )
For example, the model predicted (ok, it would have predicted this if it were done in 2009 for 2010, instead of in 2011 for 2010) 128% increase in armed violence from 2009 to 2010 in one province. The actual data demonstrated a 120% increase (100 incidents in 2009 to 222 in 2010). More importantly, the model estimated the probability of correct prognosis- which means one can discern those portions of the model where predictions are difficult (and where more model refinement may be useful).
This type of study- replete with “noise”- means we can begin to truly develop plans of action based upon scientific principles to change the course of events. Kudos to this group of friends!
Wow; this is very interesting, Roy… I tool like this could really save lives, when they get the ability to bring the data up to a current state and keep it there.
On the other hand, I can easily see some of the problems that could be caused if the software and data falls into the wrong hands. We need to do a better job of protecting tools and information that becomes part of our military arsenal than we have done in the past.
Peggy
Peggy-
You have outlined the dilemna of this sort of technology. And, I wish I had the answer to that, too!
Roy
All I can say about this is Wow! Truly amazing stuff Roy – thanks for sharing this interesting research!
Tor Constantino, MBA (@torcon) recently posted..Theology is no Substitute for Waiting Under an Oak
What made this amazing to me is that we (the non-military) have only a portion of this data (thanks to WikiLeaks). What could one do with complete data, Tor?
Thanks for the visit AND the comment.
Roy
Sounds like a very worthwhile study! If only we could shoot ahead in time and have hindsight and know in our lifetimes how this worked out.
Carolina HeartStrings recently posted..CAROLINA ROOTS – TEA
Alessa:
Not sure that we need hindsight. The hindsight will determine how accurate our models are and where they need tweaking. But, assuming only one side has the “computer power” AND the model, then this should work.
Roy
Thanks for this description, Roy. Sounds pretty compelling, based on my naive intuition and a hefty dose of wishful thinking. However, I suspect that the progress of warfare, like the progress of economics, doesn’t follow a rational course. Much as we might wish for predictable model, my sense is that the frenzy of warfare is a boatload of loose cannons. And all violent acts are NOT equivalent, with similar consequences. Given all that, why are you so hopeful for this project?
Robbie Schlosser recently posted..Plan a Surprise for Wedding and Party Music
Robbie:
I’m with you and against you on this. No, all violent acts are NOT equivalent. But, if one is trying to determine where to put troops to protect the most people- then, it is. Because, the attempt to protect against the one isolated event that may occur is an important act- but one that will be difficult if nigh to predict. Unless one IS at war, and everyone in the citizenry has subrogated their rights. And, by eliminating the bulk of the violent acts,the citizenry will be better protected.
Given that, there is a difference between fighting against an enemy in war and enemy in thought. And, this sort of program will never work against an enemy in thought.
Roy
Yeah everyone here all they can say is WOW, because sir Roy, this page is really interesting to everyone…
Fatima Hipolito recently posted..parenting classes
Yes, Fatima, I agree with you- but, then, again, I wrote it 🙂
Yeah you are really a good writer sir Roy, keep it up more articles to come in your site…
Fatima Hipolito recently posted..sleep consultant
Thanks so much, Fatima. I aporeciate that approbation!