Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 1 Simulating Superdiversity Bruce Edmonds Centre for Policy Modelling Manchester Metropolitan University
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 1
Simulating Superdiversity
Bruce EdmondsCentre for Policy Modelling
Manchester Metropolitan University
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 2
Acknowledgements
• This work came out of a long personal
collaboration with David Hales
• A tiny part of the “SCID” project (the
Social Complexity of Immigration and
Diversity), 2010-2016, funded by the
EPSRC under their “Complexity
Science for the Real World” call
• In conjunction with the Cathy Marsh
Institute for Social Research and the
Department for Theoretical Physics at
the University of Manchester
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 3
Aims of Talk
• To talk about agent-based social simulation, and
its place in social science
• To illustrate how social simulation might be used
to explore and illustrate issues of diversity
• To show both its power and its difficulties
• To, hopefully, inspire collaboration for the
development of this tool for understanding issues
of diversity
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 4
Caveats!
• What will be presented is an abstract simulation
• This should be treated as a kind of “thought
experiment” to suggest ideas, hypotheses etc.
• It has not been checked against any observed
data and so does not tell us about what happens
in observed processes/phenomena
• It is merely to show what sort of thing can be put
into a simulation…
• …with the hope of stimulating collaborations that
might develop a model with a better evidential
grounding from which conclusions might be drawn
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 5
Structure of Talk
1. About agent-based social simulation
2. A brief bit of historical simulation context
3. About the simulation model set-up
4. The complexity of simulation outcomes
5. How this kind of simulation might be developed
into something more serious
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 6
Agent-Based Simulation
• Is a computer program
• Much like a multi-character game, where each social actor is represented by a different “agent”
• These agents can each have very different behaviours and characteristics
• Social phenomena (such as social networks) can emerge out of the decisions and interaction of these individual agents (upwards “emergence”)
• But, at the same time, the behaviour of individuals can be constrained by “downwards” acting rules and social norms from society and peers
• No particular theoretical assumptions are needed!
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 7
System Dynamics, Statistical, or other
Mathematical modelling
Real World Equation-based Model
Actual Outcomes
Aggregated
Actual OutcomesAggregated
Model Outcomes
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 8
System Dynamics, Statistical, or other
Mathematical modelling
Real World Equation-based Model
Actual Outcomes
Aggregated
Actual OutcomesAggregated
Model Outcomes
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 9
System Dynamics, Statistical, or other
Mathematical modelling
Real World Equation-based Model
Actual Outcomes
Aggregated
Actual OutcomesAggregated
Model Outcomes
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 10
System Dynamics, Statistical, or other
Mathematical modelling
Real World Equation-based Model
Actual Outcomes
Aggregated
Actual OutcomesAggregated
Model Outcomes
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 11
System Dynamics, Statistical, or other
Mathematical modelling
Real World Equation-based Model
Actual Outcomes
Aggregated
Actual OutcomesAggregated
Model Outcomes
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 12
Individual- or Agent-based simulation
Real World Individual-based Model
Actual Outcomes Model Outcomes
Aggregated
Actual OutcomesAggregated
Model Outcomes
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 13
Individual- or Agent-based simulation
Real World Individual-based Model
Actual Outcomes Model Outcomes
Aggregated
Actual OutcomesAggregated
Model Outcomes
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 14
Individual- or Agent-based simulation
Real World Individual-based Model
Actual Outcomes Model Outcomes
Aggregated
Actual OutcomesAggregated
Model Outcomes
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 15
Individual- or Agent-based simulation
Real World Individual-based Model
Actual Outcomes Model Outcomes
Aggregated
Actual OutcomesAggregated
Model Outcomes
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 16
Individual- or Agent-based simulation
Real World Individual-based Model
Actual Outcomes Model Outcomes
Aggregated
Actual OutcomesAggregated
Model Outcomes
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 17
Individual- or Agent-based simulation
Real World Individual-based Model
Actual Outcomes Model Outcomes
Aggregated
Actual OutcomesAggregated
Model Outcomes
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 18
Individual- or Agent-based simulation
Real World Individual-based Model
Actual Outcomes Model Outcomes
Aggregated
Actual OutcomesAggregated
Model Outcomes
Agent-
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 19
What happens in ABS
• Entities in simulation are decided on
• Behavioural Rules for each agent specified (e.g. sets of
rules like: if this has happened then do this)
• Repeatedly evaluated in parallel to see what happens
• Outcomes are inspected, graphed, pictured, measured
and interpreted in different ways
Simulation
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 20
What happens in ABS
• Entities in simulation are decided on
• Behavioural Rules for each agent specified (e.g. sets of
rules like: if this has happened then do this)
• Repeatedly evaluated in parallel to see what happens
• Outcomes are inspected, graphed, pictured, measured
and interpreted in different ways
Simulation
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 21
What happens in ABS
• Entities in simulation are decided on
• Behavioural Rules for each agent specified (e.g. sets of
rules like: if this has happened then do this)
• Repeatedly evaluated in parallel to see what happens
• Outcomes are inspected, graphed, pictured, measured
and interpreted in different ways
Simulation
Specification (incl. rules)
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 22
What happens in ABS
• Entities in simulation are decided on
• Behavioural Rules for each agent specified (e.g. sets of
rules like: if this has happened then do this)
• Repeatedly evaluated in parallel to see what happens
• Outcomes are inspected, graphed, pictured, measured
and interpreted in different ways
Simulation
Specification (incl. rules)
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 23
What happens in ABS
• Entities in simulation are decided on
• Behavioural Rules for each agent specified (e.g. sets of
rules like: if this has happened then do this)
• Repeatedly evaluated in parallel to see what happens
• Outcomes are inspected, graphed, pictured, measured
and interpreted in different ways
Simulation
Specification (incl. rules)
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 24
What happens in ABS
• Entities in simulation are decided on
• Behavioural Rules for each agent specified (e.g. sets of
rules like: if this has happened then do this)
• Repeatedly evaluated in parallel to see what happens
• Outcomes are inspected, graphed, pictured, measured
and interpreted in different ways
Simulation
Specification (incl. rules)
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 25
What happens in ABS
• Entities in simulation are decided on
• Behavioural Rules for each agent specified (e.g. sets of
rules like: if this has happened then do this)
• Repeatedly evaluated in parallel to see what happens
• Outcomes are inspected, graphed, pictured, measured
and interpreted in different ways
Simulation
Specification (incl. rules)
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 26
What happens in ABS
• Entities in simulation are decided on
• Behavioural Rules for each agent specified (e.g. sets of
rules like: if this has happened then do this)
• Repeatedly evaluated in parallel to see what happens
• Outcomes are inspected, graphed, pictured, measured
and interpreted in different ways
Simulation
Specification (incl. rules)
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 27
What happens in ABS
• Entities in simulation are decided on
• Behavioural Rules for each agent specified (e.g. sets of
rules like: if this has happened then do this)
• Repeatedly evaluated in parallel to see what happens
• Outcomes are inspected, graphed, pictured, measured
and interpreted in different ways
Simulation
Representations of OutcomesSpecification (incl. rules)
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 28
In Vitro vs In Vivo Analogy
• In biology there is a well established distinction between what happens in the test tube (in vitro) and what happens in the cell (in vivo)
• In vitro is an artificially constrained situation where some of the complex interactions can be worked out…
• ..but that does not mean that what happens in vitrowill occur in vivo, since processes not present in vitrocan overwhelm or simply change those worked out in vivo
• One can (weakly) detect clues to what factors might be influencing others in vivo but the processes are too complex to be distinguished without in vitroexperiments or observation
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 29
The Micro-Macro Link
• How do the tendencies, abilities and observed behaviour of individuals…
• …relate to the measured aggregate properties of society?
• Social Embedding etc. implies this link is complex
• Averaging assumptions (a general tendency + random noise) do not capture non-linear interaction
• This is often two-way, with society constraining and framing individual action as well as individual constituting society in an emergent fashion
• Somewhat-persistent, complicated meso-level structures mediate these effects – these might be key to understanding this
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 30
Micro-Macro Relationships
Micro/
Individual data Qualitative, behavioural, social psychological data
Theory,
narrative
accounts
Social, economic surveys; Census Macro/
Social data
Simulation
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 31
Micro-Macro Relationships
Micro/
Individual data Qualitative, behavioural, social psychological data
Theory,
narrative
accounts
Social, economic surveys; Census Macro/
Social data
Simulation
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 32
Simulations can be very complex
• Simulations can be complicated, with lots of detail happing simultaneously to many agents in parallel
• This is the point of agent-based simulation, since it allows us to track complicated processes that we could not hold in our mind
• There may be emergent phenomena – patterns that appear at the macro level that are not obviously ‘built into’ the structure but result from the processes at the micro level
• As well as constraints from the population and surrounding agents on behaviour of individuals
• This makes the simulations difficult to understand
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 33
Understanding Simulations
• Although complex, simulation outcomes can be
inspected in multiple ways at any level of detail
• Any number of experiments on the simulation can
be performed to test understandings
• Population of agents can be measured just as
people can be, (but all of them and without error)
• However other ways can be more helpful, e.g.
– Using different visualisations of the population
– Looking at social networks
– Following individual agents and generating their
‘stories’
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 34
Historical Context 1: Schelling’s
Segregation Model
Schelling, Thomas C. 1971.
Dynamic Models of
Segregation. Journal of
Mathematical Sociology 1:143-
186.
Rule: each iteration, each dot
looks at its 8 neighbours and if,
say, less than 30% are the
same colour as itself, it moves
to a random empty square
This was a kind of counter
example – it showed that
segregation could emerge with
low levels of ethnocentrism
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 35
Historical Context 2: Axelrod’s Model
of Cultural Change
Axelrod, R (1997) The
dissemination of culture - A
model with local convergence
and global polarization.
Journal of Conflict
Resolution, 41(2):203-226.
Rule: each iteration, each
patch picks a neighbour, if is
sufficiently similar copy one
of their ‘values’
Increasing sized patches
appear different from each
other but uniform inside.
Colours above are a summary,
ethnicity of patches represented
as a string of values
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 36
Historical Context 3: Hammond and
Axelrod’s Model of Ethnocentrism
Hammond, RA. & Axelrod, R.
(2006) The Evolution of
Ethnocentrism. Journal of Conflict
Resolution, 50(6):1-11.
Rules: Colours are different
ethnicities: circles cooperate with
same color, squares defect with
same color, filled-in shapes
cooperate with different color,
empty shapes defect with
different color.
If new agents inherit from parents
(with some mutation) then
ethnocentrism evolves over time
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 37
This model aims to…
• …go beyond that of a few, pre-defined sets, but
rather allows groupings to emerge and dissolve
• That does not pre-determine what constitutes an
individual’s “in-group” but lets this develop
• That takes seriously the heterogeneity of people
• But also how behaviour and groupings result from
the social embedding of those individuals within
their social environment as a result of their
individual experience and interactions
• To be a starting point for the development of a
more serious model of these issues
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 38
Agents in the model have:
• 2 continuous characteristics: their ethnic tag,
and a cultural tag – only difference is that
cultural tag can be changed! No hard-wired
link to behaviour.
• Behaviour is specified as to which action (out
of 3 possible) an agent takes towards: (a) a
member of its in-group (b) a non-member
– 3 possible actions no nothing (Sit), donate
altruistically (Donate), harm other (Fight)
• 2 numbers to determine the extent of their
ethnic- and cultural-tolerance
• Their score in current round of interactions
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 39
Agents in the model have:
• 2 continuous characteristics: their ethnic tag,
and a cultural tag – only difference is that
cultural tag can be changed! No hard-wired
link to behaviour.
• Behaviour is specified as to which action (out
of 3 possible) an agent takes towards: (a) a
member of its in-group (b) a non-member
– 3 possible actions no nothing (Sit), donate
altruistically (Donate), harm other (Fight)
• 2 numbers to determine the extent of their
ethnic- and cultural-tolerance
• Their score in current round of interactions
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 40
The meaning of actions
Before the rounds all agents have a score of 0
In the rounds of the interaction phase when paired
• “Bit” (do nothing) no change is made
• “Donate” the agent transfers value to the other at
a cost to itself (value received 0.2 value cost by
sender is 0.1 here)
• “Fight” the agent subtracts value from the other at
a cost to itself (value lost 1.0 value cost by sender
is 0.1 here)
Outcome: an agent may imitate (mutable)
characteristics from one with a higher score
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 41
In- and Out-group
• Agents can behave differently towards other
agents, depending on whether other is in their in-
group or not (any of the 3 actions can be their
behaviour to in-group and to out-group)
• Key rule for in-group: the difference in cultural
characteristics is less than their cultural tolerance
AND if the difference in ethnic characteristics is
less than their ethnic tolerance
• Note this is not symmetric: A may consider B as
part of their in-group but not vice versa (e.g.
because B is less tolerant of deviation)
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 42
Illustration of characteristics,
tolerances and in-group
Range o
f cultura
l chara
cte
ristics
Range of ethnic characteristics
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 43
Illustration of characteristics,
tolerances and in-group
Range o
f cultura
l chara
cte
ristics
Range of ethnic characteristics
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 44
Illustration of characteristics,
tolerances and in-group
Range o
f cultura
l chara
cte
ristics
Range of ethnic characteristics
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 45
Illustration of characteristics,
tolerances and in-group
Range o
f cultura
l chara
cte
ristics
Range of ethnic characteristics
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 46
Illustration of characteristics,
tolerances and in-group
Range o
f cultura
l chara
cte
ristics
Range of ethnic characteristics
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 47
Illustration of characteristics,
tolerances and in-group
Range o
f cultura
l chara
cte
ristics
Range of ethnic characteristics
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 48
Illustration of characteristics,
tolerances and in-group
Range o
f cultura
l chara
cte
ristics
Range of ethnic characteristics
Ethnic tolerance
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 49
Illustration of characteristics,
tolerances and in-group
Range o
f cultura
l chara
cte
ristics
Range of ethnic characteristics
Cultural tolerance
Ethnic tolerance
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 50
Illustration of characteristics,
tolerances and in-group
Range o
f cultura
l chara
cte
ristics
Range of ethnic characteristics
Cultural tolerance
Ethnic tolerance
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 51
Illustration of characteristics,
tolerances and in-group
Range o
f cultura
l chara
cte
ristics
Range of ethnic characteristics
Cultural tolerance
Ethnic tolerance
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 52
Illustration of characteristics,
tolerances and in-group
Range o
f cultura
l chara
cte
ristics
Range of ethnic characteristics
Cultural tolerance
Ethnic tolerance
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 53
Illustration of characteristics,
tolerances and in-group
Range o
f cultura
l chara
cte
ristics
Range of ethnic characteristics
Cultural tolerance
Ethnic tolerance
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 54
Illustration of characteristics,
tolerances and in-group
Range o
f cultura
l chara
cte
ristics
Range of ethnic characteristics
Cultural tolerance
Ethnic tolerance
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 55
A visualisation of a population
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 56
A visualisation of a population
Each
rectangle
represents
an individual
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 57
A visualisation of a population
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 58
A visualisation of a population
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 59
A visualisation of a population
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 60
A visualisation of a population
Projections to 1D
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 61
A visualisation of a population
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 62
A visualisation of a population
Cultural
Picture
only
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 63
A visualisation of a population
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 64
A visualisation of a population
Ethnic
picture
only
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 65
A visualisation of a population
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 66
A visualisation of a population
FS
“Fight”
in-
group
“Sit” with
out-
group
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 67
A visualisation of a population
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 68
Behaviour Rules
• (Interaction) Several times for each agent:
– agent paired with other (in the same patch)
• If other is in its in-group: do in-group action to it
• If other is not in its in-group: do out-group action to it
• (Imitation) Several times for each agent:
– agent paired with other (in the same patch)
• If other agent has a better score than self: imitate all that
agent’s characteristics except ethnicity
• (Noisy change) For each agent:
– with a small probability randomly change strategy
– with a small probability randomly change tolerances
– with a small probability randomly change cultural value
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 69
Pairing
Is biased in both interaction and imitation phases
• A parameter can be set so as to make it more
likely an agent will be paired from another in its in-
group during the interaction phase (here 50% of
the time from own group 50% at random)
• Another parameter controls how likely an agent is
to be paired with another of its own group during
the imitation phase (here 10% of the time from
own group, 90% at random)
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 70
Summary of Model
• Agents have their own behaviours (Sit, Donate,
Fight), different for in- and out-groups
• They have their own definitions of their in-group
• Ethnic characteristic is fixed, but cultural value
characteristic may change
• Model goes through interaction, imitation and
noisy change phases
• No initial correlation between ethnic, cultural
values and behaviours (behaviours are always
random at the start)
• Key process: imitation of an agent doing better
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 71
Example run 1
• Only one patch
• 200 agents
• A continuous range of ethnic characteristics
• Initially random ethnic and cultural characteristics
• Initially wide tolerances
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 72
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 73
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 74
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 75
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 76
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 77
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 78
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 79
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 80
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 81
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 82
Emergent
cooperative
group based on
culture
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 83
Graphs of example run 1
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 84
Graphs of example run 1
‘Waves’ of
group-based
cooperation
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 85
Graphs of example run 1
‘Waves’ of
group-based
cooperation
Cultural
distinctions
emerging
but not
increasing
ethnic ones
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 86
One cooperative dynamic found
One of the dynamics found in this model is:
1. A group of mutual cooperators happens to form
2. These do very well by mutually donating to each
other and hence increasing their score a lot
3. Other agents imitate these, ‘joining’ their group
and copying their cooperative strategy
4. So the group grows quickly
5. After a while one agent in the group changes its
strategy or group and so gains from
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 87
Example run 2
• Only one patch
• 200 agents
• 3 differentiated ethnicities
• Initial cultures correlated with ethnicity
• Initially small tolerances
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 88
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 89
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 90
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 91
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 92
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 93
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 94
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 95
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 96
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 97
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 98
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 99
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 100
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 101
Graphs of example run 1
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 102
Cooperation in run 2
• Cooperation does occur, with the strategy to
cooperate being imitated
• But cooperation is defined by culture AND
ethnicity
• However no lasting purely ethnically-based
cooperation lasts in this model
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 103
Example run 3
• four patches
• 100 agents per patch
• 6 differentiated ethnicities
• Initial cultures and space correlated with ethnicity
(so one majority and minority ethnicity in each
patch)
• Initially small tolerances
• No migration – 4 independent patches
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 104
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 105
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 106
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 107
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 108
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 109
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 110
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 111
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 112
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 113
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 114
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 115
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 116
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 117
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 118
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 119
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 120
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 121
A Patch
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 122
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 123
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 124
Each ‘spoke’
is a group of
culturally
identical
agents
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 125
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 126
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 127
Colours indicate
behaviour,
shape is
ethnicity
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 128
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 129
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 130
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 131
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 132
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 133
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 134
Graphs of run 3
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 135
Graphs of run 3
Low
cooperative
dynamics
some
aggressive
action
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 136
Example run 4
• four patches
• 100 agents per patch
• 4 differentiated ethnicities
• Initial cultures and space correlated with ethnicity
(so one majority and minority ethnicity in each
patch)
• Initially small tolerances
• Migration at low rates (0.5%) and comparison
between agents on other patches also at low
rates (1%)
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 137
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 138
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 139
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 140
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 141
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 142
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 143
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 144
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 145
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 146
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 147
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 148
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 149
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 150
Graphs of example run 4
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 151
Graphs of example run 4
Good
cooperative
dynamics but
presence of
aggressive
strategy but
unexpressed
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 152
Example run 5
• 5x5 patches
• 20 agents per patch
• 5 differentiated ethnicities
• Initial cultures and space correlated with ethnicity
(so one majority on each patch)
• Initially small tolerances
• Migration at low rates (0.5%) and comparison
between agents on other patches also at low
rates (1%)
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 153
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 154
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 155
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 156
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 157
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 158
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 159
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 160
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 161
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 162
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 163
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 164
Graphs for run 5
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 165
Graphs for run 5
Cooperative
dynamics but
much greater
variety of
behaviours
and more
expressed
aggression
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 166
Summary of model
In this model…
• Groups, in-groups etc. all ‘fuzzy’ and only identifiable from patterns and processes observed
• Cultural groups strongly emerged even when enthicities and cultures separated to start with
• Groups are dynamic, new ones forming, growing, decaying all the time
• Cooperation maintained despite ‘selfish’ motivation to ‘defect’ and be a parasite
• Sometimes ethno-cultural groups
• Migration between patches promotes cooperation
• The more patches and the smaller the numbers on each patch (also the lower the migration) the greater the variety of behaviours and the more expressed agressive actions there were
Simulating Superdiversity, Bruce Edmonds, Birmingham, January 2017. slide 167
The End
The Centre for Policy Modelling:
http://cfpm.org
These slides will be available at: http://slideshare.net/BruceEdmonds
Ad for Workshop!
Beyond Schelling and Axelrod:
Computational Models of
Ethnocentrism and Diversity
Manchester
June 7-8th
“Ethnosim2017”