Conversation and Cooperation in Social Dilemmas: a Meta-Analysis of Experiments

Conversation and Cooperation in Social Dilemmas: a Meta-Analysis of Experiments

David Sally

Rationality and Society 1995


This is a very admirable piece in my opinion. The author compiled a bunch of statistics from over 100 prior game theory studies. More data, in general, seems to be a good thing. Results from one study with 50 participants can be telling, but the results would be very preliminary. Combine 35 years of smaller studies, and you have a lot of results to look at, and those results can be very telling.

Sally begins with a typical two criminals being arrested narrative to introduce the PD. I’ll bring up one objection I had to a statement he made on pg 59. In this narrative, he claims that if the prisoners were able to communicate, their conversations would have been “without any truthful content whatsoever” once they were separated and had to make their choice. Even with common knowledge of the payoffs and “rational” way of playing means playing defect, and the fact that their earlier talk was “cheap”, I don’t think their previous conversation was necessarily without truth.

Bob can make a promise to Joe that he will not defect. Joe can either believe or not believe Bob. When they are later separated to make their plays, Bob can either defect or cooperate. If Bob plays D he lied. If he plays C, he told the truth. Maybe Bob is way more likely to play one or the other, but that does not prevent him from playing cooperate. In fact, the idea that he might play cooperate and that he signaled that in his cheap talk could allow for the possibility of a CC play in a one shot PD (which is rare and unlikely, but not impossible!).

Anyways, he then moves into talking about the difference between game theory and the “reality” of playing games and how and if they differ. He makes a point on pg 60 that game theorists, when applying game theoretic principles, will estimate the cooperation rate to be very low (~0%), and even when they are asked to give their opinion outside of strict theory, they still underestimate cooperation (~20%) as compared to what experimental studies show (40% with a range of 5% to 96.9% and a standard deviation of 23.7%).

There’s really no point in my going through every independent variable he looked at, because if you wanted that you could just read the paper! So,  I’ll just list some of the really cool and interesting findings thoughts:

  • Econ grad students cooperated less than other types of people (pg 63)
  • Game theory PD studies/experiments can suffer from a number of experimental design issues such as the way you frame the terms:
    • Normative framing of the terms “cooperate” and “defect” versus saying play L or R.
    • Milgram related experimenter authority bias
    • The instructions of the goals of the experiment
  • One shot has less cooperation than repeated (but I think we knew that)
  • Up to a certain point, payoffs do not change cooperation rate
  • Seeing people as you are playing them could pull in social tendencies

Now, one independent variable I want to give special attention to is group size. A two person PD can obviously be played with only 2 players. But cooperation rates can be measured when you look at other social dilemmas. Sally doesn’t go into great details as to what those dilemmas were, but I assume they were common pool resource games. I think multiplayer games are incredibly interesting.

In fact I’ll take some time to write some thoughts about the number of people in a game.

Negative numbers= that doesn’t really make sense except maybe in an alternate reality

0= still doesn’t really make sense to have a 0 person game

1= a 1 person game can really only be played by a person with dissociative identity disorder (I wonder if there are any studies of that?)

2= well, this is a mainstay of game theory it seems. 2 player PD along with all the other seminal games

3= Truels! These are also interesting and add another layer of coordination and cooperation issues. You can also divide and conquer with three people.

4=Partners and coalitions! Even numbers of people versus odd numbers of people are also interesting from a social psychology phenomenon angle for games of 2 or more people. Just from personal experience I know I act differently, and my friends act differently whether we are in groups of 2 or 3 or 4. Three is a particularly interesting number because two people usually end up leaving the other person as a third wheel to some degree.

5~25= this seems like a typical group setting: a meeting, a sports team, a classroom etc. Has a certain dynamic.

25~50,000= This is about a small city level of people in a game.

50,000~10,000,000= big region/small country

10,000,000~500,000,000 big country/small continent

Billions= global scale

N=unspecified number of people

I stratify these populations/number of players in a game levels only point out that I think that when you are considering cooperation you need to consider what population size you’re talking about. Especially if you are talking about how to foster or force cooperation/coordination. A cooperation “solution” for a village of 700 may or may not hold true for New York City, or the European Union.

Lastly, a parting thought about all of this and political science and law. The principles of federalism allow for different governing bodies to make different types of decisions. (Supranational government) Federal government, state government, county government, city government, neighborhood government, household government all represent different numbers of people “playing” a game. Usually, they are all partaking in some sort of commons dilemma. What gets interesting is when counties and states interact, and of course, states and federal government interact. A lot of stuff to think about in intergovernmental cooperation/coordination and how laws manage cooperation at the micro and macro level.



Posted on August 30, 2011, in Uncategorized. Bookmark the permalink. Leave a comment.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: