gintis's Game Theory Evolving

[Nick Thompson]

Hi,

Has anybody any thoughts to share about Gintis's new book?  I have like some of Gintis's work as presented at conferences.  But I am struggling with this book because, despite an aura of userfriendliness, the book seems to leave huge steps out.  It seems to be a compilation of dozens and dozens of games with groovy names and silly stories. Is this what game theory IS when one gets close to it?   Is it true that game theory consists of story upon story as counterintuitive as the prisonner's dilemma game story. ( To "cooperate" means to me to be a "cooperative" witness; to defect, would be to renege on an agreement with the DA to cooperate;  teaching students what these words mean to game theorists is like making them drink Jamestown Kookaid;).  I have learned that there are more categories of games I have to worry about, and I suppose that is good.  I have learned that there are simultaneous games in which the players move at the same moment and serial games in which one player moves and then the other.  Also there are symetrical games in which, for instance your payoff playing strategy A with me is the same as my strategy playing Strategy A with you.  So, I have learned that the game I have spent most time thinking about ... Tragedy of the Commons type games lke PD games.are actually a narrow category of games, Simultaneous, symetrical, two player games.  (Please dont hesitate to correct me on any of this)

So, I wondering, within the scope of simultaneous symmetrical  two player games,  are there a zillion games that differ only in subtle changes in their payoff tables AND in their groovy names and silly stories?  Could all of this be collapsed into a 4d space (one dimension for each value in a 2x2 table and the space analysed?  The goal would be to identify interesting regions in this space.  

I understand about the importance of metaphors in science and about the value of "surplus meaning" in models, even  including the stuff which is just plain facetious.  I KNOW that one cannot disprove Darwinism by demonstrating that there is no great FarmerInTheSky called NATURE who is doing the "selecting".  But if this game theory literature is as it appears in Gintis's book, is it not surplus meaning gone wild????  

Feel free to jerk on my chain here:  I just dont  get it.

Nick


Nicholas Thompson
nickthompson at earthlink.net
EarthLink Revolves Around You.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: /pipermail/friam_redfish.com/attachments/20060810/40cf1261/attachment.html

[Owen Densmore]

Which one?
   Game Theory Evolving
   Moral Sentiments and Material Interests
   Unequal Chances: Family Background and Economic Success
   The Emancipatory Promise of Charter Schools
     ....

     -- Owen

Owen Densmore
http://backspaces.net - http://redfish.com - http://friam.org


On Aug 9, 2006, at 10:31 PM, Nicholas Thompson wrote:

> Hi,
>
> Has anybody any thoughts to share about Gintis's new book?  I have  
> like some of Gintis's work as presented at conferences.  But I am  
> struggling with this book because, despite an aura of  
> userfriendliness, the book seems to leave huge steps out.  It seems  
> to be a compilation of dozens and dozens of games with groovy names  
> and silly stories. Is this what game theory IS when one gets close  
> to it?   Is it true that game theory consists of story upon story  
> as counterintuitive as the prisonner's dilemma game story. ( To  
> "cooperate" means to me to be a "cooperative" witness; to defect,  
> would be to renege on an agreement with the DA to cooperate;  
> teaching students what these words mean to game theorists is like  
> making them drink Jamestown Kookaid;).  I have learned that there  
> are more categories of games I have to worry about, and I suppose  
> that is good.  I have learned that there are simultaneous games in  
> which the players move at the same moment and serial games in which  
> one player moves and then the other.  Also there are symetrical  
> games in which, for instance your payoff playing strategy A with me  
> is the same as my strategy playing Strategy A with you.  So, I have  
> learned that the game I have spent most time thinking about ...  
> Tragedy of the Commons type games lke PD games.are actually a  
> narrow category of games, Simultaneous, symetrical, two player  
> games.  (Please dont hesitate to correct me on any of this)
>
> So, I wondering, within the scope of simultaneous symmetrical  two  
> player games,  are there a zillion games that differ only in subtle  
> changes in their payoff tables AND in their groovy names and silly  
> stories?  Could all of this be collapsed into a 4d space (one  
> dimension for each value in a 2x2 table and the space analysed?  
> The goal would be to identify interesting regions in this space.
>
> I understand about the importance of metaphors in science and about  
> the value of "surplus meaning" in models, even  including the stuff  
> which is just plain facetious.  I KNOW that one cannot disprove  
> Darwinism by demonstrating that there is no great FarmerInTheSky  
> called NATURE who is doing the "selecting".  But if this game  
> theory literature is as it appears in Gintis's book, is it not  
> surplus meaning gone wild????
>
> Feel free to jerk on my chain here:  I just dont  get it.
>
> Nick
>
>
> Nicholas Thompson
> nickthompson at earthlink.net
> EarthLink Revolves Around  
> You.============================================================
> FRIAM Applied Complexity Group listserv
> Meets Fridays 9a-11:30 at cafe at St. John's College
> lectures, archives, unsubscribe, maps at http://www.friam.org

[David Eric Smith]

Nick, hi,

I think a lot of people agree with your take on a very heavy emphasis
on a somewhat narrow view of game theory.  

Though it is not without its difficulties, a book I have found
valuable is Martin Shubik's "Game Theory in the Social Sciences".
This is written more from the perspective of a RAND cold-war game
theorist, and has an entirely different feel.  Rather than try to
convince you that a 2x2 payoff matrix yields compelling wisdom about
emotionally charged social problems, the Shubik book is more in the
spirit of von Neumann and Morgenstern.  This crowd views game theory
as the proper framework to make explicit ones assumptions about what
an agent knows and what he is allowed to do.  As a result, there is a
lot more emphasis on cooperative versus non-cooperative gaming as a
procedural distinction, and on the relations between extensive-form,
strategic-form, and coalitional-form specifications of games (the 2x2
payoff matrices are a tiny subset of games in strategic form).
Shubik's book tends to emphasize how one can get any answer as a
solution -- sometimes more than I have a taste for -- where I would
prefer more emphasis on validation, but it is a better survey of the
richness of game theory as a tool for talking carefully about agency
and strategic dynamics than some of the more visible recent books.

One also sees a wider variety of applications, including voter
problems and the valuation of the bargaining "power" of an individual
in a setting where coalitions are invested with actual strategic
choices.  Shubik's big interest, of course, is the conceptualization
and modeling of money, and that is heavily emphasized, but a nice
side-effect is the treatment of "special" players like governments.  

Of course, because of its era and the author's interests, it does
leave out developments that are important in thinking about games.
One is the "behavioral economics" angle as an input to parametrize
valuations.  The other is the commitment to evolutionary models of
error and updating, which could be viewed as either the combination of
game theory with history-dependent monte-carlo, or the recursive
specification of game dynamics where the original strategic form
required that everything be laid out explicitly in advance.  

Hope this is useful to balance out the discussion of what game
theorists think game theory is.  

Eric

[Jochen Fromm-3]

> "It seems to be a compilation of dozens and dozens of games
> with groovy names and silly stories. Is this what game theory
> IS when one gets close to it?"

> "..the game I have spent most time thinking about ... Tragedy of
> the Commons type games lke PD games.are actually a narrow category
> of games, Simultaneous, symetrical, two player games."

> "..within the scope of simultaneous symmetrical two player games,  
> are there a zillion games that differ only in subtle changes in
> their payoff tables AND in their groovy names and silly stories?  
> Could all of this be collapsed into a 4d space (one dimension for
> each value in a 2x2 table and the space analysed?  The goal would
> be to identify interesting regions in this space."

Yes, I am afraid that really comes close to what game theory is.
And yes, most of the games from game theory resemble games played
by children and are very primitive (simultaneous and symetrical two
player games). Wikipedia gives a good overview over the most popular
"games": the prisoner's dilemma, the hawk-dove game, the minority game,
the ultimatum game, the rock-paper-scissor game, the matching pennies game,
see http://en.wikipedia.org/wiki/List_of_games_in_game_theory
Since game theory was invented to study strategic situations in the cold
war, it appears outdated to many scientists. In a certain sense game
theory has failed dramatically to predict and prevent the nuclear arms
race in the cold war. A huge pile of atomic weapons that still exists
today on both sides is certainly not the result of an optimal strategy.
Or is it ?

All these simple strategic games with 2x2 payoff matrix and binary
behavior are well known, and I doubt that there are still many
spectacular insights that can be gained by studying them, especially
for psychology. The behavior of real agents is more complex than
binary strategies suggest, whereas the reasoning capabilities of humans
are certainly less complex than those assumed in many strategic games,
which do not really help us to find the best strategy in a real life
situation. Reality is infinitely more complex than the most sophisticated
strategic game, and questions like "What will the other player do if he
knows that I know that he knows etc." are less important than short-term
decisions on a gut level. Such an infinite regress happens more in
Philosophy, and less in daily life.

Nevertheless, the simplicity of strategic games in game theory is
striking compared to general agent-based models. According to my
personal experience, the most difficult thing in agent-based modelling
is the extensive search to find simple models with complex results.
It is easy to create complex micro-rules with simple results,
where the experimental results are more confusing than enlightening
and in the end only confirm the setup of the experiment itself.
In the strategic games from game theory the problem to find simple
models with complex results is probably even harder: there is a smaller
number of models since they are already as simple as possible, and the
classic models have now been examined for more than fifty years.

However, the question if there are a zillion games that differ only
in subtle changes is interesting. I think this is the right way to
deal with complexity and diversity: to identify interesting regions in
the huge space of possibilities and to classify them according to their
common properties. As long as we consider closed systems with a few states,
we can not expect arbitrary complex and marvellous structures. History
also shows us that whenever you put some autonomous agents together,
the only thing you get for free in the short-term is constant conflicts,
fights and wars. Complex systems usually do not arise from simple
interactions between agents, only disorder and chaos or simple spatial
patterns like stripes, heaps, grids or simple networks. Everything else
requires a very long process of evolution or a sophisticated, deliberate
design. So it should be possible to group the different games into a
finite number of meaningful classes, alhough I think that the basic
classes of strategic two player games have already been found.

-J.