« Hurricane futures | Main | Whither right-wing thought? »
No Prisoner's Dilemma on the Western Front
Robert Axelrod's story of how cooperation developed between British and German soldiers in the trench warfare of World War I is so elegant few people have questioned it. Yet in a single sentence, Andrew Gelman says the emperor has no clothes and looky, looky, he's right!
The crux of Axelrod's story is that the soldiers were trapped in a prisoner's dilemma: individual incentives were to shoot the enemy while the socially optimal outcome was cooperation. Axelrod then introduces his famous ideas of tit for tat etc. etc. to explain how cooperation could evolve even under these most hostile of conditions.
But Gelman asks why should we think that shooting the enemy was in a soldier's best interest? Indeed,
...it seems more reasonable to suppose that, as a soldier in the trenches, you would do better to avoid firing: shooting your weapon exposes yourself as a possible target, and the enemy soldiers might very well shoot back at where your shot came from.
I believe that on this point Gelman is totally correct [insert dope slap here]. But, as he continues, "If you have no short-term motivation to fire, then cooperation is completely natural and requires no special explanation."
Axelrod's story and the large literature following it sometimes suggest that cooperation is always the thing to be explained. Cooperation is what happens when the natural order is overcome. Gelman reminds us that sometimes cooperation is the norm, it's conflict that needs to be explained. In this case, we need to explain why the soldiers fought.
Comments are open.
Posted by Alex Tabarrok on August 24, 2005 at 07:10 AM in Economics, History, Political Science | Permalink
TrackBack
TrackBack URL for this entry:
http://www.typepad.com/services/trackback/6a00d8341c66b253ef00e550834c538833
Listed below are links to weblogs that reference No Prisoner's Dilemma on the Western Front:
» Two Thoughts on Cooperation from The Work Better Weblog
The first two posts I read in the news reader this morning were on cooperation. Both with the same “conflict is generally unnecessary” point. I thought you’d enjoy.
“…cooperation is the norm, it’s conflict that nee... [Read More]
Tracked on Aug 24, 2005 9:36:43 AM
» Gallipoli As Game Theory from A Stitch in Haste
Last week, as a result of gross negligence on my part regarding the maintenance of my Netflix queue, I wound up watching Gallipoli.
The movie can be summed up thusly: ... [Read More]
Tracked on Aug 24, 2005 10:58:16 AM
Comments
What this analysis misses, however, is that the soldiers on both sides DID fire -- they just didn't hit each other. From your Axelrod link:
"In trench war, a structure of ritualised aggression was a ceremony where antagonists participated in regular, reciprocal discharges of missiles, that is, bombs, bullets and so forth, which symbolized and strengthened, at one and the same time, both sentiments of fellow-feelings, and beliefs that the enemy was a fellow sufferer."
So, both sides regularly fired and made themselves targets. The Prisoner's Dilemma comes in specifically in the fact that the shooters did NOT become targets.
Gelman's analysis would make sense if the phenomenon observed was "no firing." The phenomenon, however, was "firing at place X at 7:01. No one goes to X until 7:02. Don't shoot at source, as he won't fire against until 7:01 tomorrow."
Posted by: Richard Bellamy at Aug 24, 2005 11:16:01 AM
Further, notice the phenomenon with the artillery firing at regular times, and at regular targets. Also, notice that supply lines weren't regularly hit. Notice that snipers routinley fired, just not at people, and that work parties would go out and not provoke firing.
In other words, I agree entirely iwth what Richard is saying.
Posted by: Michael Last at Aug 24, 2005 12:07:48 PM
There's another possibility: they fired such that they weren't targets (in other words, fired blindly). Check out David Friedman's the Economics of War. I read in that essay that on average 100,000 shots are fired for every kill in modern warfare.
Posted by: Jason at Aug 24, 2005 12:09:13 PM
The arguments above would make more sense if there was no killing occurring along with that firing, but that assumption is absurd. WWI was first and foremost a bloodbath. Obviously, any benefit that might have come from cooperation was overshadowed by costs--demotion, facing a firing squad--incurred by not shooting to kill.
Axelrod's story, however, reminded me of Stanley Weintraub's excellent book SILENT NIGHT: THE STORY OF THE WORLD WAR I CHRISTMAS TRUCE. Before those costs were imposed by the opposing sides' armies, there was a definite inclination for cooperation. The governments involved could have none of that.
The degree of cooperation in society is inversely related to the size of government. World War I was a government affair, which explains the lack of cooperation between adversaries.
Posted by: Chris Westley at Aug 24, 2005 12:34:25 PM
In response to your post on why the soldiers may have fought:
In the short term, they may well have lacked the incentive to fire the first shot, but man is restless. They fought due to the various incentives presented by time. The risk of the war ending with your squadron facing off the enemy with neither side having made a move is one incentive to fight. This creates the motivation to act because if you do not act and your teammates lose, then you lose based on the sheer size of your numerical disadvantage. Another incentive to fight may rest in the desire of the soldiers to go home to their families. As game theory suggests, this future reality works its way back to the current dilemma. If you can reason your way back to the present in this way, you are presented with a competitive bargaining game rather than a cooperation game.
Dixit and Nalebuff, in Thinking Strategically, Chapter 2 “Anticipating your Rival’s Response”, lay out the several situations which provide a context for understanding “why they fought” in WWI. “Rule 1: Look ahead and reason back. Anticipate where your initial decisions will ultimately lead, and use this information to calculate your best choice. In the case of our battle, we have two sides that know that the first party to move has a disadvantage because they have shown their hand.
While the two sides weigh this out, they apply differing probabilities to the possible scenarios that could play out. The German forces most likely reason that they have the incentive to wait it out. After all, they are the defender and the British/U.S. force the invader. They also may be questioning the strength of their opponent. If the opponent is stronger, then the Germans reason they need every advantage they can get. They reason that this either leads to a surprise first-move or a wait it out strategy. On the other side of the battlefield, the American/British forces are far from home, miss their families and the weather is uncomfortable. They want to get things over as soon as possible. They reason that they are in foreign lands and the advantage is naturally weighted toward the Germans so they are pre-disposed to fight.
But we still lack a strong catalyst. What is the solution to the statement you posed, “we need to explain why the soldiers fought.” At some point in the future, the benefit of not risking your life to rise from the bunker, not fully understanding who had the competitive advantage, and various other uncertainties would be outweighed by one important catalyst.
The answer I would say is wet socks!
Posted by: tyson at Aug 24, 2005 2:00:01 PM
One word counter-argument: officers.
Th cost of cooperation in the Axelrod sense is the sum of the costs imposed by both the opposing soldiers and his own officers.
Men that failed to shoot/attack could be and were murdered by their own officers. Czarist officers used this tactic extensively. That is the historical reason why officers typically carry weapons such as swords and pistols that are highly effective at less than 20 paces but militarily worthless: so they could "motivate" their own troops. To this day the American officers retain the (collective) right to murder non-compliant soldiers. Eisenhower explicitly ordered Private Slovik executed, for example.
The puzzle as posed by Axelrod is alive and well. Of course, the puzzle of why officers are willing to both fight and "motivate" their own men remains.
Posted by: Tylerh at Aug 24, 2005 2:13:51 PM
Unsurprisingly, the military has developed a large body of knowledge on the topic of getting soliders to fight. For Americans the canonical reference is the WWII officer S.L.A. Marshall, a colorful figure whose methods have been questioned but whose results are widely believed. Basically, soldiers will act like soldiers, but generally are unwilling to kill others. To actually be soldiers(i.e. kill on command, and only on command) people need to be motivated. Cultural exhortation (eg "For King and Country," "Smote the Infidels") may have been the original method. As I noted above, another common tool historically has been the death penalty for desertion. During the middle part of the 20th century offiicers discovered two new effects: disintermediation and peer pressure. People are more willing to kill if there is a mechanical disintermediation between their own action (pulling on a lanyard) and the death of another human being (a shell exploding on the other side of the hill). I might pull a trigger on a pistol when I would hesitant to sink a knife into your chest. So fighting via mechanical means (artillery, UBoat, Cruise Missile) side steps the individual's innate unwillingness to risk the self to kill another. The "peer pressure" is that individuals don't want to be viewed as "slackers" by others they care about: no wants to be the fink who lets down this band of brothers. This is one of the reasons why the Wehrmacht was so fond of the half-track: all the infantry fought within sight of each other, and thus fought more effectively.
Following SLA Marshall the post-WWII US military developed another, even more effective paradigm: operant conditioning. Soldiers were drilled, drilled, and drilled again in environments as cognitievly similar to battle as possible. circular "bullseye" targets were replaced with silouhettes on the firing range, for example. Huge sums were spent on places like Ft. Irwin so soldiers could operate their weapons with live ammunition on almost-real targets. The result is a soldier thoroughly conditioned to operated his or her weapon effectivley, almost reflexively, without engaging in the kind of utility caclulations so valued in economics.
Which is all a long, roundabout why of saying that the analysis in the original post is specious becuase it completely ignores both the training and the social environment of the soliders. By careful design those men were not operating in a regime you or I would relate to as "normal."
Posted by: tylerh at Aug 24, 2005 2:39:49 PM
Also, some men like to fight, either because they like to kill, or they want to remove the Germans from their village in France now occupied by the Germans, or because they fear that the French will invade Germany once before as they did regularly before 1871, or similar reasons. If the Germans (or French) do not respond they will be conquered, so tit for tat responses follow even when both sets of soldiers would just like to go home.
Posted by: wkwillis at Aug 24, 2005 2:53:09 PM
There were several good comments regarding the details of the actual situation in the trenches in 1914. Some of the claims, however, were incorrect in that the German Army was the aggressor/invader, while the French and the British were trying to hold them back. Training certainly is a factor in a soldier's inclination to fight (or react in the desired way in a combat situation). Still, even with the training US soldiers get nowadays, the first time under fire induces a sort of paralysis that is broken only by the actions of more experienced soldiers, plus NCOs and officers. Also, the physical disconnect between action and result certainly lessens the sensitivity to killing, although you still know what your actions will accomplish.
Trench warfare devolved into a kind of defensive situation, excluding the various attempts at breaking through the other side's lines -- and which invariably resulted in massive casualties on both sides. In a defensive situation you are generally waiting to repel an attack, rather than initiating one. In addition, the sheer massiveness of the armies facing each other made for a rather slow-motion form of combat. Still, that doesn't explain the behavior of line soldiers on both sides.
My best guess is that not shooting, aided and abetted by little training and almost no combat experience, was a survival method employed by men who were caught up in a massive conflict. Once the grim reality of kill or be killed set in, that live and let live behavior more or less stopped.
Here's another puzzle for everyone: The American Civil War is regarded as the first war in which massive firepower directed against massed troops was a reality. It is regarded as the first war in which 20th Century armies had to operate. There are, as far as I know, no such behaviors recorded in the American Civil War.
Posted by: George at Aug 24, 2005 3:53:32 PM
A quibble here, the Wehrmacht may have been fond of half tracks, but didn't use that many. For all its Blitzkrieg reputation, the German army in WW 2 was a largely horse drawn and foot slogging operation. The proportion of their troops that fought from half tracks was miniscule.
Posted by: Fizz at Aug 24, 2005 5:11:40 PM
re George: the civil war puzzle.
I'll take a stab: before smokeless gunpowder became widespread (after 1900) most units spent most of the battle waiting for the smoke to clear -- literally. The actual time in active combat was typically a few hours at most, and often much less, because individual units could not see through the smoke long enough to make sustained attack possible. Thus one of the key premises of Axelrod's cooperation theory, that one knows the identity of the other player in the "game," was generally not met on most civil war battlefields.
The exceptions would be extended sieges, like Vicksburg. Perhaps your example is lurking there. But it's going to be hard to find becase I doubt the reporters of the day would have reported this kind of cooperation even if it had emerged.
Posted by: tylerh at Aug 24, 2005 5:16:33 PM
Further to Tylerh's 2:13:51: Most of the casualties were caused by artillery. The typical line infantry-man rarely saw, let alone had any chance of firing his rifle at, enemy artillery-men. This talk of regular firing is misleading; even the allies eventually changed and developed their tactics. http://www.amazon.com/exec/obidos/tg/detail/-/0060084332/qid=1124920666/sr=1-1/ref=sr_1_1/002-0818342-7827251?v=glance&s=books
As for German Army aggressor/invader status, it's a little more complicated. Who mobilized first? Why is that important?
As for the civil war question: The Central Powers had the men, machines, and guns to cover the whole line, so it could dig in; the Rebels didn't and couldn't. Thus, whereas in WWI you could have all quiet on the Western front, in the Civil War there weren't fronts as such, let alone significantly static lines.
Posted by: Chris at Aug 24, 2005 6:28:18 PM
I don't buy Gelman's argument insofar as we're trying to explain the absence of effective artillery shelling rather than just rifle fire. I think you do have to explain why artillery (which presumably would be able to target the other side without directly exposing itself) would fail to try to maximize casualties on the other side.
FYI Ken Binmore, who accepts the conflict paradigm as a given, has also questioned Axelrod's approach, arguing that the tit-for-tat model does not really explain the emergence of cooperation:
"Axelrod (1984) gives one striking example of the emergence of such co-operation. In the First World War, there were several fleeting outbreaks of implicit collusion between units of the British and German armies, in which each side ceased to shell the other. Axelrod (1984) attributes this behaviour to tit-for-tat reasoning, but such an explanation overlooks the obvious fact that the players didn't begin by being nice to each other. I agree that it is vital to understand how such co-operation between groups who treat each other as outsiders can get off the ground, but there seems no point at all in seeking to analyse the emergence of co-operation using a model that takes the conclusion for granted." (JASS 1997 - see http://jasss.soc.surrey.ac.uk/1/1/review1.html)
Posted by: Brendan Nyhan at Aug 24, 2005 6:56:47 PM
Here's another puzzle for everyone: The American Civil War is regarded as the first war in which massive firepower directed against massed troops was a reality
Paddy Griffiths, in his book on tactics of the American Civil War, really undermined this common assumption - he may have overstated his case at times, but the essentials of his argument - that whatever the theoretical advantages of percussion cap rifles over flint-lock muskets may have been, in the chaos and panic of battle, they turned out to be fairly immaterial in affecting how battles were fought - seem reasonably sensible to me.
Now the two wars the Prussians fought in 1866 and 1870, massed firepower was an issue then, but not really in the American Civil War...
Posted by: Richard at Aug 25, 2005 8:57:12 AM
"[A]rtillery [was] able to target the other side without directly exposing itself."
True and false. True if by direct exposure one means viewable by eyesight or binoculars by enemy infantry or artillerists. False if one includes airplanes, dirigibles, spotters and deductive reasoning, i.e. presuming they must be within X meters of the falling of the shells, where could they be firing from? Both sides, given the high stakes, developed some skill at finding out where the shells were coming from. Indeed, in response to French developed expertise, the Germans in the late stages of the war took to short barrages, without taking the time to "register" their guns, before infantry follow-up.
Thus there exists a possible non-cooperative reason for cease fire: Jesus Christ man you'll give our position away!!!
Posted by: Chris at Aug 25, 2005 12:52:34 PM
Because of the risk of getting out of the trench
rifles with a periscope device were invented and used at Gallipolli.
Posted by: MarcinGomulka at Aug 27, 2005 9:44:44 AM
Information about global offshore outsourcing market.
Posted by: global offshore outsourcing market. at Mar 31, 2006 12:24:32 AM
Crimsonlink is leading Bug Tracking Software Issue Tracking Software Defect Tracking Software
Posted by: Issue Tracking Software at Mar 31, 2006 12:25:18 AM
we are IT Offshore Consulting Company India
Posted by: IT Consulting Company at Mar 31, 2006 12:25:57 AM
we are IT Offshore Consulting Company India
Posted by: Offshore Company India at Mar 31, 2006 12:27:21 AM
We are expertise in Offshore IT Outsourcing India
Posted by: it-outsourcing at Mar 31, 2006 12:27:53 AM
We are leading IT Outsourcing India Company
Posted by: Outsourcing India at Mar 31, 2006 12:40:26 AM
we are leading IT Outsourcing Solutions Service Provider India
Posted by: IT Outsourcing Solution at Mar 31, 2006 12:40:57 AM
We are leading Software Development Outsourcing Company India
Posted by: Software Development Outsourcing at Mar 31, 2006 12:41:35 AM
We are leading Offshore Software Development Company India
Posted by: Offshore Development India at Mar 31, 2006 12:42:10 AM
