« Books I love on topics I don't care about | Main | Assorted »
Moral Wiggle Room
If Bob and Alice prefer vanilla to chocolate ice cream then when given the choice we wouldn't be surprised to see each of them choosing vanilla. Now suppose that Bob and Alice are given the following choice, if either chooses vanilla they both get vanilla only if both choose chocolate do they receive chocolate. If Bob and Alice prefer vanilla to chocolate it seems plausible that they will continue to choose vanilla. But suppose that we observe Bob and Alice choosing chocolate in the second experiment. How might we explain this?
Imagine that chocolate is considered sinful and vanilla is thought to be nice. Bob and Alice might want to be sinful but they choose nice in the first experiment to avoid social condemnation. In the second experiment, however, Bob and Alice are sinful only when both sin. True preferences are revealed only when no individual can be singled out for condemnation.
That's the setup of one of the clever experiments in an excellent new paper, Exploiting Moral Wiggle Room, except the experiment isn't about chocolate and vanilla ice cream it's about fairness in a division game. In the first experiment Alice and Bob must each decide whether to choose $6 for themselves and $1 for a third party (Cindy) or $5 for themselves and $5 for Cindy. In this experiment most Alice and Bobs choose to be nice, they divide "fairly" with Cindy. Many researchers have concluded that Alice and Bob must have a preference for niceness.
But put Alice and Bob together in the second experiment and Alice and Bob are each much more likely to choose the sinful division, $6 and $1. Alice and Bob may prefer chocolate after all.
Read the paper for several other experiments along these lines. The implications for societal organization are profound.
Hat tip to Robin Hanson.
Posted by Alex Tabarrok on November 13, 2006 at 07:19 AM in Economics, Philosophy | Permalink
Comments
Anyone who has worked on a personnel selection committee knows this effect only too well.
Posted by: Dylan at Nov 13, 2006 8:55:12 AM
The game also applies to adultery
Posted by: Yan Li at Nov 13, 2006 11:33:04 AM
I haven't read the paper, so I don't know whether they control for this, but another possible explanation is that Alice is more concerned about Bob's utility than about Cindy's and chooses accordingly. There is plenty of behavioral research showing that individuals expect others to be more greedy and value fairness less than they do themselves. So if Alice thinks Bob would strongly prefer the $6/$1 split, and Bob thinks Alice would strongly prefer it also, they may choose that option in spite of their own individual preferences. This could operate in addition to the authors' moral license story, or might simply serve as Bob and Alice's conscious justification for their choices.
Posted by: anon at Nov 13, 2006 11:52:51 AM
It may also be that these 'parlor game' results don't translate into real-world behaviors -- at least not in any straightforward way. That is, people may behave differently when the situation is artificial rather than natural and when the sums involved are trivial rather than substantial. I'm certain that people are more likely to behave selfishly when their acts are unknown rather than known -- but that was never really in doubt, was it? But can these artificial 'dictator game' results really reveal anything more detailed than that? I have my doubts.
Posted by: Slocum at Nov 13, 2006 12:03:31 PM
This reminds me of various jokes about going to restaurants on Yom Kippur (you are supposed to fast) confident in the knowledge that anyone who sees you there is also breaking the rules.
Posted by: Bernard Yomtov at Nov 13, 2006 12:27:39 PM
anon: they don't specifically control for that. They assume since the players are completely anonymous that there is no reason to value the welfare of player Y greater than the welfare of player Z.
I think that's a reasonable assumption, given the double blind nature of the experiment, but it might have been worth experimenting with different sets of outcomes to test it more certainly.
I identified another small problem with that same experiment. The social welfare is decreased by this experiment than in the baseline experiment, because there is only one recipient between the two. In the baseline experiment you have the option to sacrifice $3 in the social welfare ($10 instead of $7 paid out) in order to increase your personal payment by $1. In the modification Tyler describes, two players can collude to gain $1 each by eating $2 of the social welfare.
So someone who is a partial utilitarian (wants to maximize social welfare but values their personal gain somewhat more so) might have relative social vs. personal utility preferences that make them select the "fair" option in the baseline, but select the "get me more" option in the modified experiment. To some extent, the other experiment modifications control for this since they represent the same personal vs. social tradeoff as the baseline and don't experience significantly different results.
I have two other problems.
What if some of the actors *were* utilitarians, but did some second-level economic thinking and realized that the total social welfare in reality (as opposed to that of the experiment created sub-world) was exactly the same no matter which option they chose. Their decisions affected only a balance of payments between 3 or 4 actors, one which is the experimenters! No wealth was created in this experiment, except for the possible value of the results.
We are supposed to assume that all of the participants think of the experimenter's purse as a magic money box rather than as a person who is spending money in accordance with your dictator's wishes, and whose welfare is presumably affected by those decisions. Could it be that the mavericks who always choose the "selfish" option are not in fact selfish, but simply represent those who see through this facade and realize that their actions do not affect the general welfare, merely the relative outcome?
Also, the sample sizes seem small to me, though. The chi^2 analysis indicates a statistically significant difference, but that isn't a very strong claim. The chi^2 calc gives a 5% chance that the difference in results between the two experiments is entirely due to chance (sampling error), and leaving a very large chance that the size of the population who changes their behavior is much smaller (or greater) than the roughly one-third they speak of.
It's very hard for me to believe that doing this experiment with 2-3 times the number of samples would have been so much more expensive. Most of the expense had to be in the setup. Doing so would have reduced the potential for sampling error to insignificance and made the results much stronger. Doesn't this general tendency to halt and publish as soon as statistical significance is achieved irk everyone interested in the pursuit of knowledge?
Posted by: Michael Sullivan at Nov 13, 2006 1:24:33 PM
I wrote:
In the modification Tyler describes, two players can collude to gain $1 each by eating $2 of the social welfare.
That should be "Alex", not "Tyler". Sorry for the misattribution.
Is there a way to edit comments on Typepad? That's what I would usually do on blogger or LJ.
Posted by: Michael Sullivan at Nov 13, 2006 1:27:57 PM
Great comments, Michael. I hadn't picked up on the total social welfare issue in my quick skim, and you're absolutely right. Having run some of these types of experiments with students in classroom settings, they are very aware of how much money they are taking out of the experimenter's pocket, and will often seek to maximize that amount (kind of a soak-the-professor approach). If the reduction in social welfare is smaller under the group experiment, it could certainly be having an effect.
For my comment about Alice and Bob valuing each other's outcome more than Cindy's, even with random anonymous assignment, people very quickly take on an "us vs. them" mentality in these games. Taking on a specific "partner" role changes outcome valuation substantially. Work along these lines was done by folks like George Loewenstein and Linda Babcock, and since the first author of this paper is a student of Loewenstein's, I would have expected this issue to be more specifically addressed.
Posted by: anon at Nov 13, 2006 2:01:05 PM
This experiment just says what I already knew. That lots of people chose their own self-interest over the greater good.
Posted by: Half Sigma at Nov 13, 2006 2:27:55 PM
No offense, but duh?
Posted by: BillWallace at Nov 13, 2006 3:56:51 PM
I will take Half Sigma and Bill Wallace's comments as a compliment to how well I framed the argument! If you think about it, however, the result certainly isn't obvious because a large literature has developed on "preferences for fairness". The experiment shows that it's only a preference to be seen as fair which is quite a different thing.
I second Michael's comments which are spot on.
Alex
Posted by: Alex Tabarrok at Nov 13, 2006 4:06:02 PM
http://vark.blogspot.com/2006/11/at-marginal-revolutions-alex-has.html
Posted by: arbitrary aardvark at Nov 13, 2006 4:52:41 PM
Slocum: "I'm certain that people are more likely to behave selfishly when their acts are unknown rather than known -- but that was never really in doubt, was it?"
Well, it was kind of never considered in this literature before, as in typical experiments, actions directly translate into payoffs (for yourself and/or your opponent). And if you think about it, in typical experiments the subjects' acts are unknown in the sense that the other players will never know who it was that took a certain action, or even who the other players were. So in that sense it's surprising that additional 'noise' leads to people being more selfish, as this entails that otherwise they would have cared about what an unknown stranger thinks about them...
Posted by: fff at Nov 13, 2006 9:55:18 PM
From the paper,"in the final manipulation, the dictator would be randomly
cut off". would it bring bias to the selfishness assumption?
Posted by: sally at Nov 14, 2006 5:57:32 AM
Alex says
"however, the result certainly isn't obvious because a large literature has developed on "preferences for fairness".
I contend that the latter doesn't prove the former. Just because a bunch of people spent a bunch of time looking for something that they really hope is there doesn't mean that it's not obvious that it isn't.
Posted by: BillWallace at Nov 14, 2006 5:55:50 PM
Is the person who chooses the greater amount of money for themselves definitely being selfish?
I read an article that pointed out an extra fillip. Imagine you are an altruist who wants to help the worse-off people, and imagine also that for some reason you are participating in one of these sorts of games.
Now it is highly unlikely that your opponent/partner at the university where these are normally set is particularly badly off, by world standards. So an altrusit would surely want to make as much money as possible themselves, so they can donate it to their favourite charity. Then you can be much surer of helping someone who really needs it, that you would be if you helped your anonymous fellow player.
These experiments are picking up on people valuing cooperation, or some other ethic, not altruism. One can't rule out the possibility that someone who is aiming to get as much money as possible themselves is actually being nice and not selfish.
Posted by: Tracy W at Nov 14, 2006 6:22:46 PM
I am not sure how I feel about these concepts. I do agree that individuals act differently when they are in groups. I think that it was more socially acceptable for Alice to choose to split the money evenly when she was by herself but when Alice making a decision along with Bob, they could choose to be selfish together. As little as one dollar really is, sometimes people can be greedy enough. The thing is, is that Bob can't judge Alice and Alice can't judge Bob because they both made the same selfish decision.
Posted by: Christine at Nov 15, 2006 6:06:30 PM
数控机床数控机床。。数控改造。。
数控机床。。
数控机床改造。。
数控机床仿真软件数控机床改造。。
数控机床。。
数控机床。。
数控机床。。
数控车床数控车床。。
数控车床。。
华中数控车床。。
大连数控车床。。
无锡数控车床山东数控车床。。
南京数控车床。。
数控车床。。
数控车床。。
广州数控车床数控车床。。
广州数控车床。。
华中数控车床。。
数控车床。。
数控车床北京数控车床。。
数控车床简介。。
教学车床。。
教学数控车床。。
数控铣床南通数控铣床。。
数控铣床。。
数控铣床。。
数控铣床。。
教学数控铣床教学铣床。。
数控机床实训设备。。
加工中心。。
加工中心。。
加工中心加工中心。。
微型加工中心。。
数控加工中心。。
加工中心。。
加工中心加工中心。。
加工中心。。
上海加工中心。。
加工中心
Posted by: 南京北春 at Aug 20, 2007 10:38:00 PM
hey, what's up?
Posted by: free psp games at Jul 24, 2008 3:36:28 PM
Posted by: at Oct 14, 2008 2:27:57 AM
Weight loss is easier than you think!
Posted by: adipex at Mar 4, 2009 4:29:05 PM
best-tutor.com
http://www.best-tutor.com
good site
Posted by: tutor at May 27, 2009 12:05:52 AM
thank you
Posted by: sohbet at Aug 10, 2009 5:44:05 PM
thank you
Posted by: sohbet at Aug 10, 2009 5:44:32 PM
