The agents in a computational model need not be given much knowledge of one another, or of the model's working. They lack just as we in our world lack, any direct lines to the gods. They do not 'know the model' they are in, and need to struggle to learn what is going on. Agents like this might strike Austrian economists as somewhat interesting. We would not expect the virtual agents of such computational models ever to be terribly creative creatures, by our standards. They could be interpreted as creative for their world, however, in the sense that they will experiment with modes of behavior they stumble upon, learn from their experience and adopt the ones that seem to work. The set of all possible modes of behavior that are there to stumble into will be available only to the gods. This seems similar, in the relevant respect, to the circumstance in which we mortals find ourselves within our world. pp 554-55.That is a paragraph from the chapter "Austrian models? Possibilities of evolutionary computation" by Don Lavoie, in The Elgar companion to Austrian economics, edited by Peter Boettke (1994). Lavoie is obviously referring to Agent Based Models (ABM).
Lavoie was a professor of Austrian economics, among other subjects, at GMU and influenced several important economists. He was interested in different approaches to understand society. The paragraph above is one example which tells about his interest in evolutionary computation. He was also interested in culture and economics, hermeneutics, among other fields.
ABMs have been around for several years. One example is the Sugarscape model by Epstein and Axtell. Currently important macroeconomists are exploring those models to understand outcomes and dynamics that can not be explored by using standard macroeconomic models that usually assume homogeneous agents and equilibrium. See here an example of a project to develop an ABM for an economic crisis. Rajiv Sethi has a very interesting post on ABM vs Dynamic Stochastic General Equilibrium (DSGE) models.