Chapters * Title * Contents * Introduction * Place * System * Design * Using * Future * Bibliography
Sections
* Place * Constructionist * Evocative * Crystallizing * MediaFusion * Organizing * Why * Glass * Multiplayer * Reflection * MUD
Multiplayer Simulations--System Dynamics in a Social Context
How does making MarketPlace multiplayer make it possible to create interesting social simulations with just a few simple rules?
In short, the players fill in for the parts of the system that are more difficult to model formally. In the case of market
simulations, it's natural to use them to fill in for people, firms, or institutions. For instance, adding humans makes it easy
to have simulations featuring complex and idiosyncratic risk/reward tradeoffs without resorting to simple fixed heuristics or
implementing complex models of behavior. The formally stated systems dynamics model (Forrester, 1968) can then be used to do
what it does best--describe simple structures such as positive feedback loops. The humans perform the decision making and then
ask each other "Why did you do that?"--valuable opportunities for reflection. Much of the model ends up being represented
in the discussion that occurs during and after play. This informal representation lacks much of the theoretical power of a
formal representation--but a formal model of the human decision making would be so complex that it would be inaccessible to the
audience that MarketPlace is aimed at. A formal model that people don't understand is worse than no model at all.
This ability to get extra expressive mileage out of simple models has the potential to ease the user-programming problem in
future versions of social simulation environments. It's simpler to design user-programming environments to enable the creation
of simpler things. In MarketPlace it allows a simple, easily comprehended, formal model to generate a wide array of interesting
market behaviors.
An example adapted from (Krugman 1991) demonstrates the economically interesting behavior that a combination of people and a
simple formal structure can simulate that would otherwise require a complex formal structure. Imagine that we have a number of
players interested in building factories that produce machinery. Assume that factories that are adjacent to other factories are
more productive as a function of the number of nearby factories. Assume that players know this and that they can place
factories anywhere on the MarketPlace map (which starts out empty).
The players whose cluster of factories ends the biggest will be at a substantial advantage, while the value of the other
players' initial investment will fall. A player who manages to convince other players that her site is likely to end up the big
one might well make that a self-fulfilling prophecy. A site that has the most factories built on it the first turn might panic
another player to either build production at that site or change to producing a different commodity. The players' models of
the world come into play in a way that is very hard to make real in single-player environments. For example, if the players
have recently been exposed to game theory, they might assume that all the other players are going to make this move and
therefore all rush to the big site on the second turn.
Thus, a group of players playing both within and with the structure defined by a simple formal model can have experiences which
capture the dynamics of debates about trade and subsidy, the influence of public perceptions and moods on the economy, and the
very different effects of positive and negative feedback. Formal models describing the behavior of the human players would
necessarily be large and therefore inaccessible to most people. The reasons for the decisions made by the human participants
are at least potentially accessible and can lead to a wide array of discoveries about the above topics. These discoveries,
however, only come through reflection about the experiences gained in using the system.
Greg Kimberly/gregkimb@gak.com