NetLogo Code for SOMS-I, II, & III
SOMS-I
Reconstruction of Jerry’s OG Model Here
SOMS-II
Model with exogenously added noise
SOMS-III: Model with Endogenous Disruption
globals
[
history ;; tracks system-wide proportion of rule1 vs rule2
]
patches-own
[
new-color ;; currently, always either white or black
neighborz ;; other cells in a circle around the cell
rule1 ;; autarkical preference for rule1
rule2 ;; autarkical preference for rule2
utility1 ;; utility of cells from rule 1, function to be randomly assigned in set-up
utility2 ;; utility of cells from rule 2
utility ;; utility from chosen rule
type-random ;; used to sort into types
QK ;; Quasi-Kantian, True or False
Linear ;; Linear
MC ;; Marginal Cooperators
HCC ;; Highly Cooperative
Contrar ;; Contrarian or no? -- This is a binary {0,1}
]
to setup
clear-all
;; computes neighbors in an ellipse around each cell
ask patches
[
set neighborz ellipse-in radius-x radius-y
]
reset-ticks
end
;; this procedure sets approximately initial-density percent of the
;; cells white and the rest black; if initial-density is set at 50%
;; then about half the cells will be white and the rest black
to restart
clear-all
ask patches
[set neighborz ellipse-in radius-x radius-y]
ask patches
[ ifelse random-float 100.0 < initial-density
[ set pcolor white ]
[ set pcolor black ]
]
let total (proportion-Linear + proportion-QK + proportion-MC + proportion-HCC)
ask patches [set type-random random-float 1.0]
ask patches
[if type-random < proportion-QK / total
[set QK true
set Linear false
set MC false
set HCC false]
if type-random >= proportion-QK / total and type-random < ((total - proportion-HCC - proportion-MC) / total)
[set Linear true
set QK false
set MC false
set HCC false]
if type-random >= proportion-QK / total and type-random >= ((total - proportion-HCC - proportion-MC) / total) and type-random < ((total - proportion-HCC) / total)
[set MC true
set QK false
set Linear false
set HCC false]
if type-random >= proportion-QK / total and type-random >= ((total - proportion-HCC - proportion-MC) / total) and type-random >= ((total - proportion-HCC) / total)
[set HCC true
set QK false
set Linear false
set MC false]
]
; [ ;ifelse type-random < proportion-QK / total
;[ set QK true]
;[ set QK false
; ifelse type-random < ((total - proportion-HCC - proportion-MC) / total)
;[set Linear true]
;[set Linear false
;ifelse type-random <= ((total - proportion-HCC) / total)
; [set MC true
; set HCC false
;]
;[set MC false
; set HCC true
;]
;]
;]
;]
ask patches [assign-prefs]
reset-ticks
end
;; I set the autarkical values to be random between 0 and 10.
;; This may seem odd, since it posits no inverse relationship between the rules.
;; It is simple to check if the results hold for rule2 = 10 - rule1
;; Philosophically, there is no a priori reason to suppose that the inverse relations must hold. In fact, the rules may embody values that are simply incommensurable, in which case the utility model is out of place.
to assign-prefs
set rule1 random-float 10
set rule2 random-float 10
if rule1 < Contrarians
[set Contrar true]
end
to go
ask patches [ pick-new-color ]
ask patches [ set pcolor new-color ]
add-noise
tick
end
to pick-new-color ;; patch procedure
let activator count neighborz with [pcolor = white]
let inhibitor count neighborz with [pcolor = black]
let weight1 (activator / count neighborz)
let weight2 (inhibitor / count neighborz)
if Linear = true
[
set utility1 weight1 * rule1
set utility2 weight2 * rule2
]
if QK = true
[ifelse (weight1) > .1 and (weight1) < .5
[set utility1 ((weight1 - .1) / .4) * rule1
ifelse (weight2) > .1 and (weight2) < .5
[
set utility2 ((weight2 - .1) / .4) * rule2
]
[
ifelse weight2 < .1
[set utility2 0]
[set utility2 rule2]
]
]
[ifelse weight1 < .1
[set utility1 0]
[set utility1 rule1
ifelse weight2 > .1 and weight2 < .5
[
set utility2 ((weight2 - .1) / .4) * rule2
]
[
ifelse weight2 < .1
[set utility2 0]
[set utility2 rule2]
]
]
]
]
if MC = true
[ifelse (weight1) >= .3 and (weight1) <= .8
[set utility1 ((weight1 - .3) / .5) * rule1
ifelse (weight2) > .3 and (weight2) < .8
[
set utility2 ((weight2 - .3) / .5) * rule2
]
[
ifelse weight2 < .3
[set utility2 0]
[set utility2 rule2]
]
]
[ifelse weight1 < .3
[set utility1 0]
[set utility1 rule1
ifelse weight2 >= .3 and weight2 <= .8
[
set utility2 ((weight2 - .3) / .5) * rule2
]
[
ifelse weight2 < .3
[set utility2 0]
[set utility2 rule2]
]
]
]
]
if HCC = true
[ifelse (weight1) >= .35
[set utility1 ((weight1 - .35) / .65) * rule1
ifelse (weight2) >= .35
[set utility2 ((weight2 - .35) / .65) * rule2
]
[set utility2 0]
]
[
set utility1 0
ifelse weight2 >= .35
[
set utility2 ((weight2 - .35) / .65) * rule2
]
[set utility2 0]
]
]
;; The following code says that patches will switch to the patch with higher utility -- unless they are contrarian and there is 'rule hegemony'
;; in which case, they will switch to the minority color
ifelse Contrar = true
[ifelse weight1 < HegThresh
[set new-color white]
[ifelse weight2 < HegThresh
[set new-color black]
[ifelse utility1 >= utility2
[ set new-color white ]
[ set new-color black ]]]]
[ifelse utility1 >= utility2
[ set new-color white ]
[ set new-color black ]]
ifelse new-color = white
[set utility utility1]
[set utility utility2]
;; tie goes to rule 1, following Jerry, perhaps rule 1 is simpler
end
;;Older code trying to incorporate Rebellion as a separate procedure from new-color calculation
;to rebel
; let activator count neighborz with [pcolor = white]
; let inhibitor count neighborz with [pcolor = black]
; let weight1 (activator / count neighborz)
; let weight2 (inhibitor / count neighborz)
; if Contrar = true and weight1 < .2
; [
; set pcolor white
; ]
; if Contrar = true and weight2 < .2
; [
; set pcolor black
; ]
;end
to add-noise
ifelse randomshock = true
[
ask n-of (random-poisson noise-level) patches
[ifelse pcolor = black
[set pcolor white]
[set pcolor black]
]
]
[
ask n-of noise-level patches
[ifelse pcolor = black
[set pcolor white]
[set pcolor black]
]
]
end
;; procedures for defining elliptical neighborhoods
;; why x-radius and not radius-x? radius-x is a global variable, the brackets define 2 local variables (x-rad and y-rad)
;;whose value is determined by the operator 'with,' which takes all patches with x-radius, y-radius values satisfying the inequality
;;the inequality is just the formula for an elipse with center at the current patch ('myself')
;; since we're taking max and considering the radius, I think we end up with the neighborhood being circular, rather than eliptical
;; looking at the actual interface, though, this does not seem true. Odd.
to-report ellipse-in [x-radius y-radius] ;; patch procedure
report patches in-radius (max list x-radius y-radius)
with [1.0 >= ((xdistance myself ^ 2) / (x-radius ^ 2)) +
((ydistance myself ^ 2) / (y-radius ^ 2))]
end
;; The following two reporter give us the x and y distance magnitude.
;; you can think of a point at the tip of a triangle determining how much
;; "to the left" it is from another point and how far "over" it is from
;; that same point. These two numbers are important for computing total distances
;; in elliptical "neighborhoods."
;; Note that it is important to use the DISTANCEXY primitive and not
;; just take the absolute value of the difference in coordinates,
;; because DISTANCEXY handles wrapping around world edges correctly,
;; if wrapping is enabled (which it is by default in this model)
;; NB distance xy works by comparing the distance between current patch to the coordinate following 'distancexy'
;;so, here, we first consider the point that lets x differ by taking the x-coordinte of 'other-patch', but holds y-coordinate fixed by simply taking pycor
to-report xdistance [other-patch] ;; patch procedure
report distancexy [pxcor] of other-patch
pycor
end
to-report ydistance [other-patch] ;; patch procedure
report distancexy pxcor
[pycor] of other-patch
end
; Copyright 2003 Uri Wilensky.
; See Info tab for full copyright and license.
Use it, cite me
Comments/descriptions, etc.