Virtual Markets

Computing power is advancing at a fairly rapid clip. Eventually, we may in fact have enough power to begin computer simulations of virtual people in virtual markets. What could this mean economically if we were able to run these types of simulations? I think that simulations of both economic and political policies could first be run on a computer before being enacted in the real world. This might result in an optimization of economic activity. It might also allow a society to avoid bubbles or recessions. How does one successfully model human behavior to a degree both computationally efficient and accurate?

The New York Times had a recent interesting article about using virtual markets populated with virtual agents to mimic market dynamics. Few economist have tried to do this so far, but I think in the future it will increasingly come into use.

The idea is to populate virtual markets with artificially intelligent agents who trade and interact and compete with one another much like real people. These “agent based” models do not simply proclaim the truth of market equilibrium, as the standard theory complacently does, but let market behavior emerge naturally from the actions of the interacting participants, which may include individuals, banks, hedge funds and other players, even regulators. What comes out may be a quiet equilibrium, or it may be something else.

The Pentagon currently wants to "realistically replicate human behavior and frailties" to improve cyberwarfare tactics. So there is already a push in this direction by our government

"Provide robust technologies to emulate human behavior on all nodes of the range for testing all aspects of range behavior."

"Replicants will produce realistic chain of events between many users without explicit scripting behavior."

This is an article bringing up the potential problems inherent in running a computer simulation with virtual agents (such as observer bias). Here is another interesting short article talking about virtual reality being used as an economic laboratory. They bring up the fact that people often do not act the way that economists expect people to.

"One problem that economists face is the gnawing fact that humans often don't act rationally, potentially undermining many basic theories. It would be nice if economic theories could be tested in a lab setting. Professor Edward Castronova has made a career studying the economics of virtual worlds, MMORPGS like Ultima Online and Everquest."

"Theoretically, these worlds could act as labs to test economic models, like socialism, the third way, or a flat tax scheme. As the games get more advanced, perhaps they could be used to study more complex concepts like unionization, school vouchers, and single-payer healthcare. The experience of games designers could help test for the unintended consequences of regulations, something that would be a help in public policy."

A simulation is only as good as it reliably models a virtual agent within that model. Virtual agents would have to successfully model human psychology. Perhaps aspects of general intelligence (IQ) might be necessary to get a better simulation. Multiple simulations could then be run with different variables until a model that worked reasonably well was chosen. Then it could be used to test out a variety of regulatory/economic policies.