« Fun debates | Main | Markets in everything, virtual reality edition »
Chimps more "rational" than humans
Chimps play the ultimatum game more "rationally," that is to say more in line with standard economic models, than do humans. I'm not sure whether this says more about chimps or economists.
Thanks to Jeff Smith for the pointer.
Posted by Alex Tabarrok on October 9, 2007 at 09:28 AM in Economics | Permalink
Comments
researchers concluded chimpanzees do not show a willingness to make fair offers and reject unfair ones. In this way, they protect their self interest and are unwilling to pay a cost to punish someone they perceive as unfair
reputation must not matter to chimps
Posted by: jgray at Oct 9, 2007 9:47:30 AM
My first and immediate suspicion is on communication. Can we be sure we were really telling the chimps the proper rules of the game? If we weren't, the obvious lesson here is that if you're going to screw someone, don't tell them anything.
Posted by: perianwyr at Oct 9, 2007 9:48:10 AM
Chimps love Ayn Rand!
Posted by: Mike Giberson at Oct 9, 2007 10:23:31 AM
I guess it's all about framing of the article.
The economist took a different (more appropriate view - in my less rational viewpoint)
http://www.economist.com/science/displaystory.cfm?story_id=9898270&CFID=26890186&CFTOKEN=42349972
Posted by: Saul Weiner at Oct 9, 2007 10:38:56 AM
Game theory is for chimps?
Posted by: indiana jim at Oct 9, 2007 10:49:47 AM
Game theory is for chimps?
Posted by: indiana jim at Oct 9, 2007 10:50:12 AM
We are inclined to punish unfairness because we recognize that very few things are one-time deals, and punishing unfairness is perfectly rational if this is something that will be repeated again and again. While punishing someone for an unfair share may be irrational in this case (I assume that it's framed such that this is a one-time thing), the fact they still do is perfectly understandable because this one-time-thing is rare and unrealistic in real life and thus we instinctively pursue fairness. Irrational? Yes. Understandable? Yes, because in most real-life cases, it is perfectly rational.
As for the person doing the dividing, as long as there is a reasonable expectation that there will be punishment for unfairness (even if it is irrational), the person doing the dividing will have an incentive to divide fairly.
Posted by: Anonymous at Oct 9, 2007 11:43:13 AM
Has anyone ever conducted this experiment with real money? I mean it's one thing to reject an "unfair" proposal when someone offers you $.50 out of $5. Not too costly to send a message. What would happen if the proposal was $500 out of $5000? would the fairness bug still bite people? I kind of doubt it. I think the fact that grapes are important to chimps, $.50 not so important to most people can help explain these results.
Posted by: angus at Oct 9, 2007 11:58:49 AM
"Has anyone ever conducted this experiment with real money? I mean it's one thing to reject an "unfair" proposal when someone offers you $.50 out of $5. Not too costly to send a message. What would happen if the proposal was $500 out of $5000? would the fairness bug still bite people? I kind of doubt it. I think the fact that grapes are important to chimps, $.50 not so important to most people can help explain these results."
The traditional way to test this hypothesis is to run your experiment in a poor country, so that your experimental budget can allow for stakes that are large relative to local wage rates.
Here is one such study of the ultimatum game, played in the Slovak Republic, with maximum stakes equal to 62.5 hours of work.
http://kuznets.fas.harvard.edu/~aroth/papers/lihsug.pdf
They found that initially, (1) offers were similar between low and high stakes versions of the game and (2) rejection rates of low offers were slightly lower in the high stakes version. But as the game was repeated (with different partners, to avoid repeated game effects), the average offers went down.
Posted by: Steven Joyce at Oct 9, 2007 12:18:41 PM
"We are inclined to punish unfairness because we recognize that very few things are one-time deals" - Anonymous
I gather that the interesting question is whether there are more one-time deals now, in modern society, than there were in the past (for most of man's evolution). We might thus have a bias out of sync with our times.
(Not that these biases are bad! A bias toward cooperation with strangers might be quite food, even if uneconomic.)
Posted by: odograph at Oct 9, 2007 12:46:58 PM
It means economists are nothing but monkeys
Posted by: Luis at Oct 9, 2007 2:08:46 PM
This definitely put a crimp in the "Bush is a chimp" line.
Posted by: Yancey Ward at Oct 9, 2007 3:00:30 PM
I gather that the interesting question is whether there are more one-time deals now, in modern society, than there were in the past.
It doesn't really matter if the deal is one-time in the sense that you'll never make a deal with that person again. If the deal is merely observed or is public knowledge (or may become public knowledge), then reputation matters. In modern society, it is much harder to escape a bad reputation -- you can't just move where nobody knows you or switch into a different line of business, because it's too easy for people in the new place or industry to find out about you now.
Posted by: Slocum at Oct 9, 2007 3:01:13 PM
I'm pretty sure it's the case that if (1) assignment of roles is based on some method of meritocracy and (2) the experient is double blind and (3) you frame the game as a seller-buyer rather than a proposer-responder that human play starts to look like chimp play. I believe any one of these will begin to shift the offer distribution towards the game theoreitc prediction. The only reference that I can think of off the top of my head is Hoffman, McCabe, and Smith (AER 1996?) and the only one of those I'm certain that is dealt with is the double-blind issue.
The economist link provided by Saul Weiner does a little better job of explaining exactly how the study was conducted, although some issues are still unclear. I'm assuming the chimps could see how many grapes were in each pot, otherwise this would be pointless (if you played an ultimatum game with an unknown pie and someone "offered" you something would you take it?). If the amount of grapes in each pot is observable to both chimps, there is still a difference in this game and the typical human ultimatum game in that the human subject makes the division, whereas the chimp subject merely is selecting one of two pots that has already been divided. To make this consistent with the human game wouldn't one have to have a 3rd party make the initial division of goods in the human version and then have the proposer choose his pot and the responder choose to reject or accept his? That is, if a 3rd party splits the amount 70/30 and the proposer selects the 70 share would the responder reject his 30 share?
Posted by: AZ at Oct 9, 2007 3:37:57 PM
Not only are they rational, but they're good dancers too!
Posted by: Brian Hollar at Oct 9, 2007 4:28:09 PM
I wouldn't be surprised if language is what causes this "irrationality". In a world with no language, being a chump costs you nothing, and conversely rejecting free stuff simply to not be seen as a chump gains you nothing. But with language, the fact that you're a chump can be communicated to others, meaning you may end up with less forever. I wouldn't be surprised if this "irrationality" in humans is actually a quite-rational adaptation that was selected for after the development of language.
Posted by: Josh at Oct 9, 2007 4:49:00 PM
Brian Hollar,
Uh oh! I can see TV network executives coming up with the next variation of "Dancing with the Stars".
Posted by: Yancey Ward at Oct 9, 2007 5:31:24 PM
Social animals like chimps, wolves, and lions tend to have strict dominance hierarchies.
Within such groups, beta males fully cooperate in hunting even though they will eat less (and later) than alpha males. It is evidently in their "economic" self-interest to do so, since if it wasn't, such animal societies would evolve towards solitary lifestyles and "lone wolves" would come to be more common than wolf packs. Presumably, some combination of factors for each species determines whether social or solitary living is more efficient: chimps are social, orangutans are solitary; lions are social, leopards are solitary.
In this sense, acceptance of unfairness in sharing resources is built into the very structure of such animal societies, so perhaps it's not surprising that is reproduced in a lab experiment. Also, the hierarchies in animal societies tend to be "meritocratic", being based on relative physical prowess rather than caste or class. Last year's beta males can and often do turn the tables on the aging incumbent alpha, and thus may more readily accept unfairness because they may someday benefit from it.
And by the way, as Jane Goodall first discovered, chimps do hunt (small monkeys, for meat, in troops of beta males led by an alpha male).
Posted by: at Oct 9, 2007 10:16:58 PM
I think when they do these sorts of tests in "primitive" societies the results are very much in line with the chimp tests.
Posted by: Owen at Oct 10, 2007 4:26:32 AM
interesting article which shows that altruistic 'punishment' allows to form bigger groups.
So it is people who are better adapted to form groups and organizations, not chimps.
http://www.iew.unizh.ch/home/fehr/papers/NatureOfHumanAltruism.pdf
Posted by: Sergey Kurdakov at Oct 10, 2007 6:17:00 AM
Chimpanzees have a rule. This rule is it doesn't matter who you are or what your position is in the group, you can't take away food from another chimpanzee. You can ask to be given some food (beg) but it's entirely up to the other chimp how much food to give, if any, and the begging chimp has to be satified with what they get no matter how small. The results of the study are entirely consistant with normal chimpanzee behaviour. This behaviour appears common to all chimp societies. Human however make more complex rules about what can be taken and what should be given. These are formalized by tradition or, since the advent of civilization, by law. These rules exist in all human societies including those of hunters and gatherers.
Posted by: Ronald Brak at Oct 10, 2007 11:48:06 AM
Forget the typical ultimatum game results - the game that the chimps play and the game the typical human ultimatum game ARE NOT THE SAME (see my comment above - I have had 3 fairly well known young experimentalists respond separately when I asked them about this and all tended to agree). Take a look at Sally Blount (1995). "When social outcomes aren't fair: The effect of causal attributions on preference" Organizational Behavior and Human Decision Processes, 63, 131-44. Look at pg. 136, Table 1, picture c, for the "randomly generated treatment" (meaning the proposed split was randomly determined, which is likely how the chimps view it, although I can't really speak for how the chimps view anything). About 80-85% of the HUMAN subjects would accept a minimum offer of 0 or 50 cents in this game. The "fair offers only" accepted are clearly in the minority in this game.
Posted by: AZ at Oct 10, 2007 6:37:50 PM
Sorry, it's figure 1 in the Blount paper (above) not table 1.
To be fair, there is a "3rd party treatment" in the Blount paper and the acceptances do shift from the typical results (also replicated by Blount, although less extreme than the results that are usually cited). The "third party" treatment does not generate the same acceptances as the "randomly generated" proposals treatment because in the third party treatment the subjects expect that the third party (human) will have some notion of "fairness" (I hate that word - it has 31 meanings on dictionary-reference.com) or "equal splits" (this is table 1). Again, I would think that chimps would view the appearance of food in a pot as a random division, not a 3rd party division, but I can't speak intelligently on the thoughts of chimps.
Posted by: AZ at Oct 10, 2007 6:49:11 PM
It would seem to me that the social signal even in a one-off experiment with a random stranger is potentially much more important than the few dollars which are likely at stake in the game.
I'm pretty sure some kind of fairness enforcement is closer to optimal in repeated games among a population of people with various strategies in this game. I don't think we have anywhere near enough information to argue that the game-theoretic solution is "more rational" in any real world experiment unless the game is played with a very large amount of money. By large, I mean money where a typical "unfair" offer could still represent a noticeable fraction of the responder's social capital. In the rich world, that probably requires pots in excess of $10,000, given per person social capital in the 1/2 million range.
Trust is an extremely valuable (but difficult to "value") economic resource. It reduces transaction costs and enables incredible wealth building from trade. The web of trust in human society is the main engine of economic and social wealth. Our sense of fairness and justice and willingness to send costly signals about our commitment to those ideals is what makes the trust web work. It's just not worth messing with that for a few dollars unless they are the difference between starving and not.
Posted by: Michael Sullivan at Oct 11, 2007 10:56:33 AM
Another resource for pondering from Scientific American Mind
Is Greed Good?
Economists are finding that social concerns often trump selfishness in financial decision making, a view that helps to explain why tens of millions of people send money to strangers they find on the Internet
By Christoph Uhlhaas
http://www.sciammind.com/article.cfm?articleID=1D9DFCAC-E7F2-99DF-3C2DE7F657B943E5
Posted by: BrianDRPM at Oct 12, 2007 3:48:55 PM
