Agent_Zero: Toward Neurocognitive Foundations for Generative Social Science by Epstein Joshua M

Author:Epstein, Joshua M.
Language: eng
Format: epub
Publisher: Princeton University Press
Published: 2014-07-24T16:00:00+00:00

* * *

118A nice “Guide to Newcomers” is available in Axelrod and Tesfatsion, Appendix A of Tesfatsion and Judd, eds., Handbook of Computational Economics: Agent-Based Computational Economics, Volume 2. Among the closest things to a textbook on agent modeling is Railsback and Grimm (Princeton 2011). The best hands-on way to get started is to do the three excellent agent-based modeling tutorials that download with NetLogo (http://ccl.northwestern.edu/netlogo/).

119By canonical, I mean simply the base model for this development.

120A torus topology is readily available in NetLogo but would have been visually confusing for most of the runs explored here.

121NetLogo offers a wide variety of distributions from which to draw random numbers. Here, U(0, 1), the uniform distribution on the unit interval, is used.

122Update-affect is the relevant NetLogo code block. See Appendix III. My code extends Rescorla-Wagner in allowing extinction rates different than the classical model, which, of course, is an available setting.

123Properly speaking, this extinction-rate slider is a multiplier. If it is 0, there is no extinction. If it is 1.0, we obtain classical Rescorla-Wagner extinction curves. Typically, we use a value in the interval (0,1). So, this is a second extension of the original model (beyond S-curve learning), permitting yet another type of flexibility.

124Hence the adjective “spatial.”

125Later, we will interpret the set as a space of financial assets, a family of vaccines, or opportunities for unhealthy eating, over which a local relative frequency is being computed and updated.

126This is the number of orange patches over total patches within the spatial sampling radius.

127An anxiety-provoking context without question (Behrens et al., 2007).

128Lest there be any replicative or other confusion, the agent source code and NetLogo graphical output use the name disposition for net disposition. This should occasion no confusion. The NetLogo code block (see Appendix III, p. 218) governing this calculation is:

to update-disposition

ask turtle 0 [

set disposition affect + probability + [weight] of red-link 1 0 * ([affect] of turtle 1 + [probability]

of turtle 1) + [weight] of red-link 2 0 * ([affect] of turtle 2 + [probability] of turtle 2) − threshold]

ask turtle 1 [

set disposition affect + probability + [weight] of red-link 0 1 * ([affect] of turtle 0 + [probability]

of turtle 0) + [weight] of red-link 2 1 * ([affect] of turtle 2 + [probability] of turtle 2) − threshold]

ask turtle 2 [

set disposition affect + probability + [weight] of red-link 0 2 * ([affect] of turtle 0 + [probability]

of turtle 0) + [weight] of red-link 1 2 * ([affect] of turtle 1 + [probability] of turtle 1) − threshold]

end

Terms could be collected in a variety of ways, all equivalent computationally but different conceptually. This form seems expeditious for expository purposes. NetLogo’s name for a generic agent is “turtle.” I choose to imagine that this is in honor of a famous exchange between Bertrand Russell and an audience member who told Russell that the earth was supported on the back of a great turtle. Russell asked, ‘And what, pray tell, is supporting that turtle?’ The answer was immediate. “Oh, another turtle … it’s turtles all the way down.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.