Informatik
50 Years of Social Simulation:
Why We Need Agent-Based Social
Simulation (and Why Other
Approaches Fail),
Klaus G. Troitzsch
Universität Koblenz-Landau
ESSA Summer School 2010
09/10/2015
50 Years of Social Simulation
1/152
Informatik
Outline
• Simulation from the 1960s to 2010
– historical background
– main features of some of the approaches
• system dynamics, microsimulation, discrete event analysis, sociophysics,
cellular automata
• early extensions
– some first conclusions
• Why complex social systems are even more complex than
other complex systems
– peculiarities of human social systems
– requirements for computational social science
– and how they can be met: recent extensions
09/10/2015
50 Years of Social Simulation
2/152
Informatik
From World Models to Multi-Agent Models
80000
sumFood
70000
numAll
350
300
60000
250
50000
200
40000
150
30000
100
20000
50
10000
0
5000
0
5100
5200
5300
5400
5500
5600
5700
5800
5900
6000
avNormsFood
140
avCoordsFood
120
avSubsFood
100
80
60
40
20
0
5000
09/10/2015
50 Years of Social Simulation
5100
5200
5300
5400
5500
5600
5700
5800
5900
6000
3/152
Informatik
Early Simulations
• 1963 Simulmatics
• Simulation as Science Fiction:
Simulacron 3 (1964)
– movies after this novel (“Welt am Draht”
[“World on Wire”], Reiner Werner
Fassbinder; “13th Floor”; “MATRIX”)
09/10/2015
50 Years of Social Simulation
4/152
Informatik
More Early Simulations
• Microanalytical simulation of effects of tax and transfer regulations
(since 1957)
• Club of Rome simulations by Forrester and the Meadows group (early
1970s)
• Thomas Schelling’s segregation model (1971)
• Abelson’s and Bernstein’s referendum campaign simulation (1963)
• Kirk’s and Coleman’s simulation of human behaviour in small groups
(1968)
• The Global 2000 Report to the President [Jimmy Carter], ed. Council
on Environmental Quality and U.S. Department of State (1980)
09/10/2015
50 Years of Social Simulation
5/152
Informatik
Outline
• Simulation from the 1960s to 2010
– historical background
– main features of some of the approaches
• system dynamics, microsimulation, discrete event analysis, sociophysics,
cellular automata
• early extensions
– some first conclusions
• Why complex social systems are even more complex than
other complex systems
– peculiarities of human social systems
– requirements for computational social science
– and how they can be met: recent extensions
09/10/2015
50 Years of Social Simulation
6/152
Informatik
System Dynamics
• obviously has its roots in systems of differential equations from
which it seems to differ mostly in two technical aspects:
– discrete time is used as a coarse approximation for continuous
time to achieve numerical solutions, and
– functions of all kinds, including table functions, can be used with
the help of the available tools like DYNAMO or STELLA.
• is restricted to the macro level in that it models a part of
reality (the ‘target system’) as an undifferentiated whole,
whose properties are then described with a multitude of
attributes which typically come as ‘level’ and ‘rate’ variables
representing the state of the whole target system and its
changes, respectively.
09/10/2015
50 Years of Social Simulation
7/152
Informatik
A PowerSim
example
09/10/2015
50 Years of Social Simulation
8/152
Informatik
World Models
Systems Dynamics and DYNAMO have received
public interest mainly because they were used to
build large world models:
• WORLD2 (World Dynamics, Forrester 1970)
• WORLD3 (The Dynamics of Growth in a Finite
World, Meadows et al. 1974)
• WORLD3 revisited (Beyond the Limits, Meadows et
al. 1992)
• WORLD3 (The 30-Year Update, Meadows et al.
2004)
09/10/2015
50 Years of Social Simulation
9/152
Informatik
Main Features of Forrester’s World Model (1)
09/10/2015
50 Years of Social Simulation
10/152
Informatik
Main Features of Forrester’s World Model (2)
09/10/2015
50 Years of Social Simulation
11/152
Informatik
WORLD2 complete
All these feedback loops
are, of course, tied
together by auxiliaries
and controlled by
constants not shown in the
previous diagrams.
09/10/2015
50 Years of Social Simulation
12/152
Informatik
WORLD2: Results
Prediction results of Forrester’s WORLD2 model for births, deaths and
population size
09/10/2015
50 Years of Social Simulation
13/152
Informatik
Retrodiction
Retrodiction results of Forrester’s WORLD2 model for births, deaths and
population size are obviously wrong.
09/10/2015
50 Years of Social Simulation
14/152
Informatik
Types of Validity
• With Zeigler we should distinguish between
three types of validity and three different
stages of model validation (and
development):
– replicative validity: the model matches
data already acquired from the real
system (retrodiction),
– predictive validity: the model matches
data before data are acquired from the
real system,
– structural validity: the model “not only
reproduces the observed real system
behaviour, but truly reflects the way in
which the real system operates to
produce this behaviour.”
– [Zeigler 1976:5]
09/10/2015
50 Years of Social Simulation
15/152
Informatik
Outline
• Simulation from the 1960s to 2010
– historical background
– main features of some of the approaches
• system dynamics, microsimulation, discrete event analysis, sociophysics,
cellular automata
• early extensions
– some first conclusions
• Why complex social systems are even more complex than
other complex systems
– peculiarities of human social systems
– requirements for computational social science
– and how they can be met: recent extensions
09/10/2015
50 Years of Social Simulation
16/152
Informatik
The microsimulation approach
• Microanalytic simulation models were first developed to
predict demographic processes and their consequences for
tax and transfer systems (Orcutt 1986). They consist of two
levels at least:
– the level of individuals or households (or in the rare case of
simulating enterprises, the level of enterprises)
– the aggregate level (e.g. national economy level)
• More sophisticated MSMs distinguish between the
individual and the household levels, thus facilitating
models in which persons move between households and
can found and dissolve new households (e.g. by marriage
and divorce).
09/10/2015
50 Years of Social Simulation
17/152
Informatik
… what the founding fathers said …
• “. . . in microanalytical modelling, operating
characteristics can be used at their appropriate level of
aggregation with needed aggregate values of variables
being obtained by aggregating microentity variables
generated by microentity operating characteristics”
[Orcutt 1986, p. 14].
The main advantage of this kind of procedure is that
• “available understanding about the behaviour of entities
met in everyday experience can be used ... to generate
univariate and multivariate distributions” [ibid.].
09/10/2015
50 Years of Social Simulation
18/152
Informatik
Types of micro simulation
• The classical micro simulation comes in three different
types, the first of which is most common, but does not
actually describe a (stochastic) process:
– “static micro simulation”: change of the demographic structure of
the model population is performed by reweighting the age class
according to external information;
– “dynamic micro simulation”: change of the demographic
structure of the model population is performed by ageing the
model persons individually (and by having them give birth to
new persons, and by having them die) according to life tables;
– “longitudinal micro simulation”: simulation is done on an age
cohort and over the whole life of this cohort, thus omitting a
population’s age structure (but children of the cohort members
may still be simulated).
09/10/2015
50 Years of Social Simulation
19/152
Informatik
How it proceeds
• All types of micro simulation, in contrast to many
other simulation approaches, are data driven
instead of concept driven:
– Starting from data of a population or rather a sample
from some population, normally on the nation state
level,
– this approach models the individual behaviour in
terms of reproduction, education and employment,
– simulates this individual behaviour and
– aggregates it to the population level in order to
generate predictions about the future age or
employment structure.
09/10/2015
50 Years of Social Simulation
20/152
Informatik
How it proceeds
current population
with all properties
of all individuals
real process
future population
with all properties
of all individuals
sampling
representative
sample with
selected
properties
09/10/2015
projection
simulation
predicted sample
with selected
attributes updated
for all individuals
50 Years of Social Simulation
21/152
Informatik
Subprocesses
• To realise the simulation, several subprocesses
have to be modelled:
– demographic processes: ageing, birth, death,
marriage, divorce, regional mobility, household
formation and dissolution
– participation in education and employment,
employment income
– social transfers
– taxes and social security
– consumption
– wealth
09/10/2015
50 Years of Social Simulation
22/152
Informatik
Subprocesses
09/10/2015
50 Years of Social Simulation
23/152
Informatik
Structure of a typical micro simulation model
•
•
•
Initialise the individuals from an empirical data base
Link them together according to their current household structure and to
other information on networks (kinship or friendship networks, where the
latter information will usually not be available)
Then, for every simulated period
– organise the marriage market,
– and for every simulated individual
•
•
•
•
•
•
•
increase its age,
decide whether it dies,
decide whether, if it represents a woman, it gives birth to one or more children,
decide whether, if it represents a person currently married, it is divorced,
decide whether and whom it will marry,
decide whether it will move from one household to another or form a new household,
decide on transitions in education and employment, respectively
– and execute all these transitions and changes.
– Store all the data needed for the analysis and interpretation of the simulated
history and perhaps output some intermediate results.
•
Analyse and interpret the collected data, aggregate them, calculate
distributions etc.
09/10/2015
50 Years of Social Simulation
24/152
Informatik
An alternative: event orientation instead of period
orientation (1)
• Usually microsimulation proceeds in a period-oriented manner.
•
Every agent has to check in every period whether anything happens with
respect to it.
• Alternatively, the simulation could proceed from event to event, and
every event generats one or more new events:
– At the time of birth, the events “child enters school” and “mother gives
birth to another child” are scheduled for some time in the future (the
waiting time being distributed according to some frequency distribution):
• enter school
– P(tschool = tbirth+5 = 0.2),
– P(tschool = tbirth+6 = 0.5),
– P(tschool = tbirth+7 = 0.3)
• next birth
0.6
0.5
0.4
0.3
Reihe2
0.2
0.1
0
1
3
4
5
6
7
8
0.05
0.04
probability
– P(tnextbirth < tthisbrth+1 = 0.0),
– P(tthisbrth+1 < tnextbirth < tthisbrth+25 = f(tnextbirth < tthisbrth)),
– P(tnextbirth > tthisbrth+25 = 0.0)
2
0.06
0.03
0.02
0.01
24
21
22
.5
18
19
.5
15
16
.5
12
13
.5
6
7.5
9
10
.5
3
4.5
0
1.5
0
time
09/10/2015
50 Years of Social Simulation
25/152
Informatik
An alternative: event orientation instead of period
orientation (2)
• Event-oriented agent-based microsimulation makes it necessary to
look for other types of parameters than in period-oriented
microsimulation:
– instead of an age-dependent probability of giving birth to a child during
the next period (year) we need an estimate of (e.g.) the frequency
distribution of the time between the birth of the first and the second
child,
– instead of the age-dependent probability of marrying next year we need
an estimate of the frequency distribution of the time between (e.g.) the
time a person finishes school and the time when (s)he tries to find a
partner: at the time of this event (s)he will look around for partners
whose respective events are scheduled for the next very short period of
time and select the best match from them,
– instead of an age-dependent probability to die within the next period,
we need the distribution of lifetimes;
• some of these distributions are easily estimated, others are not.
09/10/2015
50 Years of Social Simulation
26/152
Informatik
Are microsimulation microentities agents?
• Agents are
√autonomous: they apply rules to beliefs and make
decisions, perhaps also plans
√reactive: they perceive stimuli from their environment
and respond to them
√proactive: they have goals which they try to achieve
¿ socially capable: they have models of their
environment and of other agents, and they can
communicate with other agents
[at least in Aparicio Diaz/Fent 2005]
09/10/2015
50 Years of Social Simulation
27/152
Informatik
UMDBS as one tool for micro simulation
• micro data base
• model
• parameters / coefficients (life tables …)
Universal Micro DataBase System UMDBS (Windows) [Sauerbier 2000,
http://www.fh-friedberg.de/sauerbier/umdbs]
09/10/2015
50 Years of Social Simulation
28/152
Informatik
Output
• tables
• graphs
• distributions (one- and
two-dimensional)
• queries
09/10/2015
50 Years of Social Simulation
29/152
Informatik
A pessimistic view
•
•
•
•
What such a micro analytical simulation model yields is in a way prediction,
but not in the strict sense.
It is the outcome of one realisation of a stochastic process whose parameters
are not exactly known but estimated on the base of more or less reliable
empirical data.
The distribution of the outcome of this stochastic process can only be
estimated (as it were, on a higher level of estimation) if a large number of
parallel runs of the same model was run; then confidence intervals can be
estimated on a Monte Carlo base.
After this time-consuming procedure we arrive at an estimate of the
distribution of, e.g., the age distribution among women ten years from now,
or of the distribution of the proportion of people over 65 with living
daughters (to nurse them in case of sickness) — but only for the one set of
parameters with which we initialised our simulation model earlier on, and
not much is then known about the sensitivity, namely the dependence of the
distribution of the outcomes of the stochastic process on slight changes on one
or several of the large number of input parameters.
09/10/2015
50 Years of Social Simulation
30/152
Informatik
… and the optimistic view
• Results of micro analytical simulation models have their
value as they show possible paths into the future,
• and Monte Carlo simulations of this type even show the
reliability of the predictions, while multiple runs of
similarly parameterised models give a first glance at the
validity of the model:
• if there is no sensitive dependence on initial conditions
then the problem of estimating parameters is not a hard
one.
• And if we happen to have a long panel or a series of
cross-sections then we can validate our model in
comparing results of simulations of past periods with the
empirical data of the same period.
09/10/2015
50 Years of Social Simulation
31/152
Informatik
Outline
• Simulation from the 1960s to 2010
– historical background
– main features of some of the approaches
• system dynamics, microsimulation, discrete event analysis, sociophysics,
cellular automata
• early extensions
– some first conclusions
• Why complex social systems are even more complex than
other complex systems
– peculiarities of human social systems
– requirements for computational social science
– and how they can be met: recent extensions
09/10/2015
50 Years of Social Simulation
32/152
Informatik
Models from Econophysics and Sociophysics
• Opinion formation or product choice
• Simple case: two alternative opinions (“yes”/
“no”) or two alternative products (“MS-DOS” /
“MacOS” or “VHS” / “Betamax”)
• Probability of choice depends on global
majorities
• Typical approach:    exp(  x)
 
   exp (  x)
x
09/10/2015
n  n
n  n
50 Years of Social Simulation
33/152
Informatik
Opinion formation in one population
• NetLogo model
09/10/2015
50 Years of Social Simulation
34/152
Informatik
Opinion formation in several disjoint
populations
• NetLogo model
09/10/2015
50 Years of Social Simulation
35/152
Informatik
Outline
• Simulation from the 1960s to 2010
– historical background
– main features of some of the approaches
• system dynamics, microsimulation, discrete event analysis, sociophysics,
cellular automata
• early extensions
– some first conclusions
• Why complex social systems are even more complex than
other complex systems
– peculiarities of human social systems
– requirements for computational social science
– and how they can be met: recent extensions
09/10/2015
50 Years of Social Simulation
36/152
Informatik
Cellular Automata
• Defining features
• Standard examples
• Social science examples
09/10/2015
50 Years of Social Simulation
37/152
Informatik
A grid of cells
09/10/2015
50 Years of Social Simulation
38/152
Informatik
Defining features
• A grid or lattice of a large number of identical
cells in a regular array
– e.g. a square
• Each cell can be in one of a (small) set of states
– e.g. ‘dead’ or ‘alive’
• Changes in a cell’s state are controlled by rules
09/10/2015
50 Years of Social Simulation
39/152
Informatik
Defining features (ii)
• The cell’s rules depend only on the state of the
cell and its local neighbours
– e.g. the immediately surrounding cells
– Consequently cells can only influence their immediate
neighbours
09/10/2015
50 Years of Social Simulation
40/152
Informatik
Defining features (iii)
• Simulated time proceeds in discrete steps
– often called steps, cycles or generations
• At each step, the state of every cell (at time
t+1) is calculated using the states of
neighbouring cells at time t.
09/10/2015
50 Years of Social Simulation
41/152
Informatik
Famous examples
• The Game of Life
– rules:
• a ‘living’ cell remains alive if it has 2 or 3 living neighbours,
otherwise it dies
• a ‘dead’ cell stays dead unless it has exactly 3 living
neighbours, when it bursts into life.
09/10/2015
50 Years of Social Simulation
42/152
Informatik
A Life sequence
09/10/2015
50 Years of Social Simulation
43/152
Informatik
The Game-Of-Life Glider
09/10/2015
50 Years of Social Simulation
44/152
Informatik
Neighbourhoods
• von Neumann neighbourhood
North
East
South
West
• Moore neighbourhood
North
North-east
East
South-east
South
South-west
West
North-west
09/10/2015
50 Years of Social Simulation
45/152
Informatik
The universe
Right neighbour is left edge cell
Bottom neighbour is
top edge cell
09/10/2015
50 Years of Social Simulation
46/152
Informatik
Spreading gossip
09/10/2015
50 Years of Social Simulation
47/152
Informatik
Majority rule
Starting configuration:
50% random ‘on’
Rule:
‘on’ if 5 or more Moore neighbours
and self are ‘on’,
‘off’ if 5 or more Moore
neighbours and self are ‘off’
Result: stable blocks of
‘on’ and ‘off’ form
09/10/2015
50 Years of Social Simulation
48/152
Informatik
The effect of individual differences
At start
Later
Rule: majority rule with uniform random threshold variation
(if 4 neighbours on and 4 off, new state is either on or off at random)
09/10/2015
50 Years of Social Simulation
49/152
Informatik
Outline
• Simulation from the 1960s to 2010
– historical background
– main features of some of the approaches
• system dynamics, microsimulation, discrete event analysis, sociophysics,
cellular automata
• early extensions
– some first conclusions
• Why complex social systems are even more complex than
other complex systems
– peculiarities of human social systems
– requirements for computational social science
– and how they can be met: recent extensions
09/10/2015
50 Years of Social Simulation
50/152
Informatik
Extensions to basic cellular automata
• Migration models
– Actors can move around the grid
• Larger neighbourhoods
– Transitions depend on more than the immediate
neighbours
• More complex rules
– e.g. rules involving memory
09/10/2015
50 Years of Social Simulation
51/152
Informatik
Migration models
•
•
•
•
Agents can move around the grid
Rules determine when and where they move to
Agents must be distinguished from cells (locations)
Agents can only move to a vacant space on the grid
09/10/2015
50 Years of Social Simulation
52/152
Informatik
An example:
segregation
• Suppose that (e.g. in the US) there was a threshold of
‘tolerance’, so that white people are content so long as at
least 3/8 of their neighbours are also white (i.e. less than a
majority), the rest being black
• If less than 3/8th are white, they move to a
neighbourhood where they are content with the ratio
• And the same applies to black people in reverse
09/10/2015
50 Years of Social Simulation
53/152
Informatik
An example
• Thomas Schelling proposed a theory† to explain the
persistence of racial segregation in an environment of
growing tolerance
• He proposed: If individuals will tolerate racial diversity,
but will not tolerate being in a minority in their locality,
segregation will still be the equilibrium situation
†Schelling,
Thomas C. (1971) Dynamic Models of Segregation.
Journal of Mathematical Sociology 1:143-186.
09/10/2015
50 Years of Social Simulation
54/152
Informatik
A segregation model
• Grid 50 by 50
• 1500 agents, 1050 green, 450 red
– so: 1000 vacant patches
• Each agent has a tolerance
– A green agent is ‘happy’ when the ratio of greens to reds in its
Moore neighbourhood is more than its tolerance
– and vice versa for reds
09/10/2015
50 Years of Social Simulation
55/152
Informatik
Aggregation
• Randomly allocate reds and greens to patches
• With a tolerance of 40%:
– An agent is happy when more than 3/8 ( = 37.5%) of
its neighbours are of the same colour
• Then the average number of neighbours of the
same colour is 58% (about 5)
• And about 18% of the agents are unhappy
09/10/2015
50 Years of Social Simulation
56/152
Informatik
At the start
09/10/2015
50 Years of Social Simulation
57/152
Informatik
Tipping
• Unhappy agents move along a random walk to a patch
where they are happy
• Emergence is a result of ‘tipping’
– If one red enters a neighbourhood with 2 reds already there, a
previously happy green will become unhappy and move
elsewhere, either contributing to a green cluster or possibly
upsetting previously happy reds and so on…
09/10/2015
50 Years of Social Simulation
58/152
Informatik
Emergence
•
•
Values of tolerance above 30% give
clear display of clustering: ‘ghettos’
Even though agents tolerate 30% of
their neighbours being of the other
colour in their neighbourhood, the
average percentage of same-colour
neighbours is typically 75 - 80% after
everyone has moved to a satisfactory
location (risen from 58% before
relocations)
09/10/2015
50 Years of Social Simulation
59/152
Informatik
Outline
• Simulation from the 1960s to 2010
– historical background
– main features of some of the approaches
• system dynamics, microsimulation, discrete event analysis, sociophysics,
cellular automata
• early extensions
– some first conclusions
• Why complex social systems are even more complex than
other complex systems
– peculiarities of human social systems
– requirements for computational social science
– and how they can be met: recent extensions
09/10/2015
50 Years of Social Simulation
60/152
Informatik
Views on simulation can be quite different
• Sugarscape:
– the question “can you explain it?” is interpreted as “can you grow it?”, and
– “a given macrostructure [is] ‘explained’ by a given microspecification when the
latter’s generative sufficiency has been established.”
• [Epstein and Axtell 1996:177]
• Microanalytical simulation:
– starts from a large collection of observational data on persons and households
and the population as a whole,
– is initialised with empirical estimates of transition probabilities, age-specific
birth and death rates and so on,
– tens of thousands of software agents are created with data from real world
people.
– And all this aims at predicting something like the age structure or kinship
networks of this empirical population in the far future
09/10/2015
50 Years of Social Simulation
61/152
Informatik
Simulation as a thought experiment
• Simulation may be seen as a thought experiment which is carried out
with the help of a machine, but without any direct interface to the
target system: We try to answer a question like the following.
• Given our theory about our target system holds (and given our theory is
adequately translated into a computer model), how would the target
system behave?
• The latter has three different meanings:
– Which kinds of behaviour can be expected under arbitrarily given
parameter combinations and initial conditions?
– Which kind of behaviour will a given target system (whose parameters and
previous states may or may not have been precisely measured) display in
the near future?
– Which state will the target system reach in the near future, again given
parameters and previous states which may or may not have been precisely
measured?
09/10/2015
50 Years of Social Simulation
62/152
Informatik
Qualitative prediction
• This is either the prediction
– which modes of behaviour are possible for a given type of systems or
– which of several possible modes of behaviour a particular target system will
have in the near future,
• provided the theory we have in mind holds for this kind of target systems
or for this particular target system.
– Will this system stabilize or lock in (and in which of several stable states will it
do so), will it go into more or less complicated cycles, will it develop chaotic
behaviour (such that long-time quantitative predictions are impossible)?
– Will this system display some emergent structures like stratification,
polarization, or clustering?
• Note: Most quantitative social simulation aims only at qualitative
prediction. And: Most qualitative prediction is done by quantitative
simulation.
09/10/2015
50 Years of Social Simulation
63/152
Informatik
Qualitative predictions
80000
sumFood
70000
numAll
350
300
60000
250
50000
200
40000
150
30000
100
20000
50
10000
0
5000
0
5100
5200
5300
5400
5500
5600
5700
5800
5900
6000
avNormsFood
140
avCoordsFood
120
avSubsFood
100
80
60
40
20
0
5000
09/10/2015
5100
5200
5300
5400
50 Years of Social Simulation
5500
5600
5700
5800
5900
6000
64/152
Informatik
Quantitative prediction
• This is the prediction
– which state the system will reach after some time, given we know
its actual state precisely enough.
– which state the system will acquire if we change parameters in a
certain manner, i.e. if we control parameters to reach a given
goal.
• Here it is only possible to calculate trajectories starting from the
measured initial state of the target system and using the
parameters of the target system (which, too, must have been
measured or adequately estimated beforehand).
• Quantitative prediction is the field of microanalytic simulation
models which are very often used for prediction in demography
and policy making.
09/10/2015
50 Years of Social Simulation
65/152
Informatik
A quantitative prediction
09/10/2015
50 Years of Social Simulation
66/152
Informatik
Another quantitative prediction
09/10/2015
50 Years of Social Simulation
67/152
Informatik
Quantitative prediction: problems
• Two additional problems have to be kept in mind
here:
– If sensitivity analysis has yielded the result that the
trajectory of the system depends sensitively on initial
conditions and parameters, then quantitative prediction
may not be possible at all (which is a very valuable result!).
– And if the model is stochastic, then only a prediction in
probability is possible, i.e. confidence intervals can be
estimated from a large number of stochastical simulation
runs with constant parameters and initial conditions.
09/10/2015
50 Years of Social Simulation
68/152
Informatik
A first conclusion …
• It should have become clear by now that social
science simulation has at least two very different
types of purposes.
– One of them might be called explanatory — this includes
also teaching —, while
– the other comprises different types of prediction and
prescription, including parameter estimation, retrodiction,
and decision making.
• In most cases, the explanatory type of simulation —
exploring would-be worlds [Casti 1996] — has to be
done before the prediction and prescription type of
simulation can be accessed.
09/10/2015
50 Years of Social Simulation
69/152
Informatik
Outline
• Simulation from the 1960s to 2010
– historical background
– main features of some of the approaches
• system dynamics, microsimulation, discrete event analysis, sociophysics,
cellular automata
• early extensions
– some first conclusions
• Why complex social systems are even more complex than
other complex systems
– peculiarities of human social systems
– requirements for computational social science
– and how they can be met: recent extensions
09/10/2015
50 Years of Social Simulation
70/152
Informatik
The whole is greater than the sum of its parts
• … the history of an aggregate is the union of the
histories of its members …
• … the history of the whole [system] differs from
the union of the histories of its parts …
• … an accurate version of the fuzzy slogan of
holistic metaphysics, namely The whole is greater
than the sum of its parts.
[Bunge 1979:4]
09/10/2015
50 Years of Social Simulation
71/152
Informatik
Emergence and emergent properties
• emergent properties (of a system) are properties
that cannot belong to the parts (elements of the
composition) of the same system
• they come into being through emergent
processes: things unconnected initially (forming
“aggregates”) begin to interact with the effect of
self-assembly: the aggregate becomes a system
with properties which none of its parts has
09/10/2015
50 Years of Social Simulation
72/152
Informatik
Social systems: what is special about them?
• social systems, unlike most others, consist of
– elements that can interact symbolically (not by
pheromones, but by words, for instance)
– elements that can take over different roles in different
contexts
– elements that can belong to different systems
(including: systems of different kinds) at the same time
09/10/2015
50 Years of Social Simulation
73/152
Informatik
Human social systems: objects of economics
and social science
• are among the most complex systems in our
world
• consist of human actors which
–
–
–
–
–
09/10/2015
are autonomous
interact in numerous different modes
take on different roles even at the same time
are conscious of their interactions and roles
communicate in symbolic languages even about the
counterfactual
50 Years of Social Simulation
74/152
Informatik
Complex systems
Physical systems consist of Living systems consist of
Human social systems
consist of
particles which
• obey natural laws
• interact only in a few
different modes
• have no roles
living things which
• are partly autonomous
• interact in several
different modes
• can play different roles
• are not conscious of
their interactions
• are only partly
conscious of their roles
and interactions (but
not all are at all)
• communicate only in a
very restricted manner
(and never about
counterfactuals)
human actors which
• are autonomous
• interact in numerous
different modes
• take on different roles
even at the same time
• are conscious of their
interactions and roles
• do not communicate
09/10/2015
• communicate in
symbolic languages
even about the
counterfactual
50 Years of Social Simulation
75/152
Informatik
Fields and forces
Physical particles interact
with the help of
Living things additionally
interact with the help of
• (a small number of
different) forces
• fields which can change
due to the movements of
particles
• chemical substances and
their concentration
gradients
• by sounds (halfway
symbolic, very restricted
lexicon)
• by observing each other
and predicting next
moves
09/10/2015
Human actors additionally
interact with the help of
• by sounds and graphical
symbols (symbolic,
unrestricted lexicon, also
referring to absent or nonexisting things, e.g. unicorns
and angels)
• by observing each other,
predicting next moves and
deriving regularities from
what they observed (but
they can also learn about
regularities from others)
50 Years of Social Simulation
76/152
Informatik
Adaptation
• many systems can adapt to their environment
• finding a local minimum of some potential or a
concentration maximum, following a
concentration gradient
• adaptation of a population of systems via
evolution (“blind watchmaker” metaphor)
• adaptation via norm learning
• mutual adaptation via norm emergence and
norm innovation
09/10/2015
50 Years of Social Simulation
77/152
Informatik
Decision making
• in physical particles: according to natural laws or
probabilistic (no decision making in any
reasonable sense of the word)
• in animals: instinct (mechanisms not well
understood)
• in humans: after deliberation of different possible
outcomes of different action alternatives,
boundedly rational, often after discussion among
groups of actors
09/10/2015
50 Years of Social Simulation
78/152
Informatik
Emergence
• definable as the supervenience of characteristics
of a system that cannot be owned by the parts of
this system
– atoms and molecules have a velocity, but no
temperature, the gas or fluid or solid body has a
temperature
– families have places where they live, but they do not
have a degree of segregation (but the city has)
– voters have attitudes, but no attitude distribution (the
electorate has)
09/10/2015
50 Years of Social Simulation
79/152
Informatik
Emergence, immergence and second-order
emergence
• emergence of order in slime moulds works via
the concentration gradient of some chemical
substance
• emergence of an attitude distribution (e.g.
polarisation of voter attitude during an election
campaign) works via communication, persuasion
and publication of opinion poll results (as
humans have no “objective” measuring
instrument for attitude “gradients”)
09/10/2015
50 Years of Social Simulation
80/152
Informatik
Outline
• Simulation from the 1960s to 2010
– historical background
– main features of some of the approaches
• system dynamics, microsimulation, discrete event analysis, sociophysics,
cellular automata
• early extensions
• Why complex social systems are even more complex than
other complex systems
– peculiarities of human social systems
– requirements for computational social science
– and how they can be met: recent extensions
09/10/2015
50 Years of Social Simulation
81/152
Informatik
Outline
• Simulation from the 1960s to 2010
– historical background
– main features of some of the approaches
• system dynamics, microsimulation, discrete event analysis, sociophysics,
cellular automata
• early extensions
– some first conclusions
• Why complex social systems are even more complex than
other complex systems
– peculiarities of human social systems
– requirements for computational social science
– and how they can be met: recent extensions
09/10/2015
50 Years of Social Simulation
82/152
Informatik
Micro and macro level
• “sociological phenomena penetrate into us by
force or at the very least by bearing down more
or less heavily upon us” [Durkheim 1895]
macro cause
macro effect
“upward
causation”
“downward
causation”
micro cause
micro effect
[Coleman 1990]
09/10/2015
50 Years of Social Simulation
83/152
Informatik
Micro and macro level
•
“sociological phenomena penetrate into us by force or
at the very least by bearing down more or less heavily
upon us” [Durkheim 1895]
macro cause
“downward
causation”
micro cause
macro effect
“upward
causation”
• both interpretations can
micro effect
[Coleman 1990]
be applied to physical
and to social systems
• both interpretations can be applied
 to physical systems
o
macro cause = field, “downward causation” = force, micro effect = movement,
“upward causation” = field change
 to social systems
o
09/10/2015
macro cause = “social field”, social norms, “downward causation” = immergence,
micro effect = norm adoption, “upward causation” = norm innovation
50 Years of Social Simulation
84/152
Informatik
Micro and macro level
•
“sociological phenomena penetrate into us by force or
at the very least by bearing down more or less heavily
upon us” [Durkheim 1895]
macro cause
• but the difference is:
“downward
causation”
micro cause
macro effect
micro effect
“upward
causation”
[Coleman 1990]
 in physical systems
o the effect of the “downward causation” is transitory, as is the effect of
the “upward causation” as there is usually no memory on either level
 in social systems
o the effect of the “downward causation” lasts for a long time, it changes
the state of the micro entity forever, as it is stored symbolically in his or
her memory, and the effect of the “upward causation” also lasts for a
long time, as there is a long-term memory in society (folklore, libraries
codes of law …)
o the “downward causation” takes only effect after being interpreted by
the individual, and this interpretation is dependent of his or her past
09/10/2015
50 Years of Social Simulation
85/152
Informatik
Outline
• Simulation from the 1960s to 2010
– historical background
– main features of some of the approaches
• system dynamics, microsimulation, discrete event analysis, sociophysics,
cellular automata
• early extensions
– some first conclusions
• Why complex social systems are even more complex than
other complex systems
– peculiarities of human social systems
– requirements for computational social science
– and how they can be met: recent extensions
09/10/2015
50 Years of Social Simulation
86/152
Informatik
Markets
 Many agents trading with each other
 Each trying to maximise its own welfare
 Neo-classical economics assumes that markets are at equilibrium,
where the price is such that supply equals demand
 But with agents, we can model markets in which the price varies
between localities according to local supply and demand
09/10/2015
50 Years of Social Simulation
87/152
Informatik
Example: Sugarscape
• Agents located on a grid
• Trade with neighbours
• Two commodities: sugar and spice. All agents consume
both these, but at different rates
• Each agent has its own welfare function, relating its
relative preference for sugar or spice to the amount it has
‘in stock’ and the amount it needs
09/10/2015
50 Years of Social Simulation
88/152
Informatik
Agent strategies
• An agent moves to the cell it prefers that is within its
range of vision to replenish sugar and spice stocks
• But can also trade (barter) with other neighbouring agents
• Agents trade at a price negotiated between them when
both would gain in welfare
09/10/2015
50 Years of Social Simulation
89/152
Informatik
Example: Sugarscape
09/10/2015
50 Years of Social Simulation
90/152
Informatik
Results
 The expected market clearing price emerges from the
many bilateral trades (but with some remaining
variations)
 The quantity of trade is less than that predicted by
neoclassical theory
- since agents are unable to trade with others than their neighbours
 The effect of trade is to make the distribution of wealth
(measured in sugar) more unequal
09/10/2015
50 Years of Social Simulation
91/152
Informatik
Outline
• Simulation from the 1960s to 2010
– historical background
– main features of some of the approaches
• system dynamics, microsimulation, discrete event analysis, sociophysics,
cellular automata
• early extensions
– some first conclusions
• Why complex social systems are even more complex than
other complex systems
– peculiarities of human social systems
– requirements for computational social science
– and how they can be met: recent extensions
09/10/2015
50 Years of Social Simulation
92/152
Informatik
Lake Anderson revisited
Original model,
System
Dynamics style
09/10/2015
Variant 1 with strategies applied within
the model
50 Years of Social Simulation
Variant 2 with
feedbacks on several levels
93/152
Informatik
Anderson‘s model: variables
•
The behaviour of the lake is described in a number of equations for the following
main “level” variables:
– nutrient: the amount of fertiliser and other nutrients in the lake, increased by fertiliser
discharge, by respiration and decay of the biomass, and decreased by the growth of
the biomass,
– biomass: the amount of algae in the lake, increased by their growth, and decreased
by their death rate, by respiration and, possibly, by harvesting algae,
– detritus: the amount of sediment at the bottom of the lake, increased by dying
algae, and decreased by detritus decay and, possibly, by the dredging the lake ground,
– oxygen: the concentration of oxygen available to the algae for their metabolism; this
level variable is composed of two parts, the epilimnion oxygen concentration (which is
considered to be constant because oxygen is always replenished from the air above the
lake surface) and the hypolimnion oxygen concentration which is increased by the
diffusion of oxygen from the epilimnion into the hypolimnion, and decreased by the
oxygen consumption (due both to the respiration of the algae and to the detritus
decay process) and, possibly, by artificial aeration.
09/10/2015
50 Years of Social Simulation
94/152
Informatik
Anderson‘s model: policies
•
Anderson describes five policies to avoid eutrophication of the lake:
– applying algicides: the application of algicides can increase the natural death rate
of the algae,
– dredging the detritus: the detritus can be dredged from the ground of the lake,
which results in a decrease of nutrient (which otherwise would have been produced
from the detritus naturally) and in an increase in the hypolimnial oxygen concentration
(because less oxygen is consumed in the detritus process),
– aeration of the lake: oxygen can be bubbled into the water of the lake, thus
increasing the hypolimnial oxygen concentration,
– harvesting algae: biomass can be harvested, thus decreasing the biomass (and, in
consequence, its oxygen comsumption and its conversion into detritus); the harvested
biomass can be used for agricultural purposes,
– reducing nutrient (fertilizer) discharge into the lake: Anderson suggests an
artificial discharge of fertiliser into the lake which is ten times the natural discharge of
nutrient from outside the lake at the beginning of most of his simulation runs;
moreover he suggests a yearly increase of the artificial fertiliser discharge of two per
cent if no specific measures are taken.
09/10/2015
50 Years of Social Simulation
95/152
Informatik
Extensions
• In the original model, these policies are applied
by the experimenter;
• in extended models,
– one or more simulated “governments” or
– other agents/agencies under the control (tax
reduction, fines, …) of local authorities
• perform the task to apply strategies to avoid or
fight eutrophication.
09/10/2015
50 Years of Social Simulation
96/152
Informatik
Environmental protection
Farm
Farm
Farm
Farm
tourists
levying taxes
smelling
industry
booking /
cancellation
voting
tourist board
information
09/10/2015
government
agency
lobbying
50 Years of Social Simulation
97/152
Informatik
Outline
• Simulation from the 1960s to 2010
– historical background
– main features of some of the approaches
• system dynamics, microsimulation, discrete event analysis, sociophysics,
cellular automata
• early extensions
– some first conclusions
• Why complex social systems are even more complex than
other complex systems
– peculiarities of human social systems
– requirements for computational social science
– and how they can be met: recent extensions
09/10/2015
50 Years of Social Simulation
98/152
Informatik
Another example:
Co-ordination and sustainability
• Agents who move in a world much like
Sugarscape [Epstein/Axtell 1996], feed there,
reproduce and perhaps communicate.
• Some agents act as co-ordinators for others: coordinators and subordinates co-operate,
informing each other about resources.
09/10/2015
50 Years of Social Simulation
99/152
Informatik
Example continued ...
• Co-ordinators gather information about
available resources from subordinates, forward it
as hints to other subordinates and receive a
contribution from successful subordinates.
• Resourcs grow on fields, spread to neighbouring
fields, and are consumed.
09/10/2015
50 Years of Social Simulation
100/152
Informatik
Example continued …
• If a field is exhausted by harvesting, new crops
can grow if seed is spread on it.
• An agent can harvest all or part of the crop in
the field (the latter acts in a sustainability mode).
• The simulation programme allows for numerous
parameter changes.
09/10/2015
50 Years of Social Simulation
101/152
Informatik
Result
• One of the simulation results is that an agent
society with co-ordination is more likely to
achieve sustainability than a society with isolated
agents.
09/10/2015
50 Years of Social Simulation
102/152
Informatik
The model
Circles and triangles: agents
: co-ordinators
: subordinates
: independent agents
colour shade of agents:
degree of saturation
black: dead
colour shade of fields:
amount of resources
09/10/2015
50 Years of Social Simulation
103/152
Informatik
Agents can ...
... feed on their individual supply,
... die (either from hunger or from old age),
... recognise the state of neighbouring cells (resources, agents) and store it in
their memories,
... estimate the results of possible
actions,
... decide which to apply,
and finally
... act.
09/10/2015
50 Years of Social Simulation
104/152
Informatik
Needs and actions
actions
needs
survival
wealth
breeding
influence
curiosity
gather
X
X
D
D
D
move
X
X
D
D
X
breed
X
X
X
D
D
start co-ordinating
X
X
D
X
X
end co-ordinating
X
X
D
X
X
subordinate
X
X
D
D
D
unsubordinate
X
X
D
D
D
rest
D
D
D
D
D
09/10/2015
09/10/2015
Klaus G. Troitzsch:
Simulation in Sociology
50 YearsComplex
of SocialSystems
Simulation
105/152
Informatik
Decision making
weight
needs
j
actions
sum
i=1
i=2
i=3
D
E
F
1 j=1
0.7
0.4
0.6
0.8
2 j=2
0.3
0.9
0.6
0.3
D..F
3 j j satij
0.55 0.60 0.65
4 j j satij - mini i j satij
0.00 0.05 0.10
5 P(i)
0.00 0.333 0.666 1.00
0.15
Actions are taken with a certain probability which depends on the
degree to which an action satisfies a need and the weights of the
needs for a particular agent.
09/10/2015
09/10/2015
Klaus G. Troitzsch:
Simulation in Sociology
50 YearsComplex
of SocialSystems
Simulation
106/152
Informatik
Simulation results
• Populations of isolated agents die out soon,
• those with co-ordinators and subordinates
survive for a long time,
• and we find Lotka-Volterra cycles.
09/10/2015
50 Years of Social Simulation
107/152
Informatik
A society with co-ordinators survives for a longer time,
60000
sumFood flat
50000
sumFood co-ordinated
40000
30000
20000
10000
0
0
20
40
60
80
100
120
140
160
180
200
350
numAll flat
300
numAll co-ordinated
250
200
150
100
50
0
0
09/10/2015
20
40
60
80
100
120
140
50 Years of Social Simulation
160
180
200
108/152
Informatik
... the population with isolated agents dies out.
60000
sumFood flat
50000
sumFood co-ordinated
40000
30000
20000
10000
0
excessive
exploitation of
resources
90
0
20
40
60
80
100
120
140
160
180
80
avAllFood flat
70
avAllFood co-ordinated
200
60
50
40
survival
30
reluctant
exploitation
20
10
extinction
0
0
09/10/2015
20
40
60
80
100
120
140
50 Years of Social Simulation
160
180
200
109/152
Informatik
Become self-employed, when times are
getting better!
80000
sumFood
70000
numAll
350
300
60000
250
50000
200
40000
150
30000
100
20000
50
10000
0
0
5000
5100
5200
5300
5400
5500
140
5600
5700
avNormsFood
5800
5900
6000
5800
5900
6000
avCoordsFood
120
avSubsFood
100
80
bad times
60
40
good times
good times
20
0
5000
09/10/2015
5100
5200
5300
5400
5500
5600
5700
50 Years of Social Simulation
110/152
Informatik
Conclusions drawn from complex antecedents
• Conclusion from a complex set of simple
assumptions:
• Co-ordination and subordination in this artificial
agent society facilitate sustainability of resources.
09/10/2015
50 Years of Social Simulation
111/152
Informatik
replicative validity: the model matches data already acquired from the real
system (retrodiction),
predictive validity: the model matches data before data are acquired from the
real system,
• Our conclusion is unlikely to ever be validated
empirically:
• real-world human societies have an
overwhelmingly more complex structure of coordination and subordination than the simple
artificial society of our model.
09/10/2015
50 Years of Social Simulation
112/152
Informatik
replicative validity: the model matches data already acquired from the real system
(retrodiction),
predictive validity: the model matches data before data are acquired from the real
system,
• Indigenous societies, however, show some aspects of the
behaviour of our simulation model:
• In a society of herdsmen and farmers in Western Africa,
decisions which rest on friendship networks (“friend-priority”
decisions) proved to be much more effective then decisions
which were made on pure cost deliberations (“cost priority”
decisions).
•
09/10/2015
50 Years of Social Simulation
[Rouchier et al. 2000, 2001:189].
113/152
Informatik
structural validity: the model “not only reproduces the observed real system
behaviour, but truly reflects the way in which the real system operates to produce this
behaviour.”
• In this respect, the multi-agent model is superior
to simpler mathematical models such as
– a Lotka-Volterra process,
• either deterministically on the macro level
– dx/dt = a x – b x y
– dy/dt = c x y – d y
• or stochastically on the micro level
– pb1(n1, n2) = α n1
– pd1(n1, n2) = γ n1 n2
09/10/2015
pb2(n1, n2) = β n1 n2
pd2(n1, n2) = δ n2
50 Years of Social Simulation
114/152
Informatik
Outline
• Simulation from the 1960s to 2010
– historical background
– main features of some of the approaches
• system dynamics, microsimulation, discrete event analysis, sociophysics,
cellular automata
• early extensions
– some first conclusions
• Why complex social systems are even more complex than
other complex systems
– peculiarities of human social systems
– requirements for computational social science
– and how they can be met: recent extensions
09/10/2015
50 Years of Social Simulation
115/152
Informatik
Flávia F. Feitosa et al.: MASUS: A Multi-Agent
Simulator for Urban Segregation, ESSA 2009, paper
30
Flávia da Fonseca Feitosa: Urban Segregation as a
Complex System. An Agent-Based Simulation
Approach, Diss. Geogr. Bonn 2010
Urban Development: MASUS
09/10/2015
50 Years of Social Simulation
116/152
Informatik
Urban Development: MASUS
09/10/2015
50 Years of Social Simulation
117/152
Informatik
Urban Development: MASUS
São José dos Campos, São Paulo, Brasilien 1991-2000
is the percentage of similar households in the neighbourhood
09/10/2015
50 Years of Social Simulation
118/152
Informatik
Outline
• Simulation from the 1960s to 2010
– historical background
– main features of some of the approaches
• system dynamics, microsimulation, discrete event analysis, sociophysics,
cellular automata
• early extensions
– some first conclusions
• Why complex social systems are even more complex than
other complex systems
– peculiarities of human social systems
– requirements for computational social science
– and how they can be met: outlook on a new approach
09/10/2015
50 Years of Social Simulation
119/152
Informatik
What this simulation is about
• This simulation is about the emergence and
immergence of norms.
• Our example is taken from everyday life: a
scenario with children crossing a street between
two playgrounds and with car drivers using this
street, both of whom learn to avoid collisions to
invent traffic rules and to respect them.
09/10/2015
50 Years of Social Simulation
120/152
Informatik
Emergence in the Loop: EMIL
• Istituto di Scienze e Tecnologia della
Cognizione − Consiglio Nazionale delle
Ricerche, Rome, Italy
• Universität Bayreuth, Germany
• The University of Surrey, United Kingdom
• Universität Koblenz−Landau, Germany
• Manchester Metropolitan University, United
Kingdom
• AITIA International Informatics Inc.,
Budapest, Hungary
09/10/2015
50 Years of Social Simulation
121/152
Informatik
Micro and macro level
•
“sociological phenomena penetrate into us by force or
at the very least by bearing down more or less heavily
upon us” [Durkheim 1895]
macro cause
“downward
causation”
micro cause
• both interpretations can
be applied to physical
and to social systems
• both interpretations can be applied
macro effect
micro effect
“upward
causation”
[Coleman 1990]
 to physical systems
o
macro cause = field, “downward causation” = force, micro effect = movement,
“upward causation” = field change
 to social systems
o
09/10/2015
macro cause = “social field”, social norms, “downward causation” = immergence,
micro effect = norm adoption, “upward causation” = norm innovation
50 Years of Social Simulation
122/152
Informatik
Micro and macro level
•
“sociological phenomena penetrate into us by force or
at the very least by bearing down more or less heavily
upon us” [Durkheim 1895]
macro cause
• but the difference is:
“downward
causation”
micro cause
macro effect
micro effect
“upward
causation”
[Coleman 1990]
 in physical systems
o the effect of the “downward causation” is transitory, as is the effect of
the “upward causation” as there is usually no memory on either level
 in social systems
o the effect of the “downward causation” lasts for a long time, it changes
the state of the micro entity forever, as it is stored symbolically in his or
her memory, and the effect of the “upward causation” also lasts for a
long time, as there is a long-term memory in society (folklore, libraries,
codes of law …)
o the “downward causation” takes only effect after being interpreted by
the individual, and this interpretation is dependent of his or her past
09/10/2015
50 Years of Social Simulation
123/152
Informatik
Immergence and second-order emergence
“I must
don’tnot
likecross
yourthe street
A:A:You
here, B!”
when I smoking
am approaching
in my car, B!
• norm-invocation messages
• motivate individual agents to change the rules
(B abstains
fromsmoking
crossing
(B abstains
from
controlling their actions
thein
street
when A is approaching
the presence
of A.)
with her car.)
• if this happens often enough, “sociological phenomena penetrate
into us by force or at the very least by bearing down more or less
heavily upon us” [Durkheim 1895]
•
… and we have programmed
and as a consequence,
these
norm invocations
–
something
much
like this
and the resulting behaviour – occur more and
in
an
agent-based
simulation
syste
more often and become a “sociological
(not(not
onlyonly
B, but
B, but
others,
others,
too,too,
abstain
abstain
from
from
crossing
smoking,
streets,
phenomenon”
not
not only
only in
in the
the presence
presence of
of A’s
A, but
car, also
but in
onmost
otherother
occasions.)
cases.)
09/10/2015
50 Years of Social Simulation
124/152
Informatik
Agent activities
• “child” agents can
– observe
– move
– admonish
• “car driver” agents can
–
–
–
–
–
–
09/10/2015
observe
stop
slow down
speed up
admonish
honk the horn
50 Years of Social Simulation
125/152
Informatik
Theoretical Framework
• Inter-agent communication uses a message concept,
triggering the processing of events and
corresponding actions
• Agents can learn (form normative beliefs into their
minds):
 own experience (reinforcement strategies)
 observation of other agents’ experience
(imitation)
 listening to other agents’ reports of their
experiences (normative learning)
09/10/2015
50 Years of Social Simulation
126/152
Informatik
Theoretical Framework
• Learning capabilities of a normative agent:
• own experience (reinforcement strategies)
• “Pedestrian experiences a near-collision with a car because of not using
the striped area for crossing a street.”
• observation of other agents’ experience (imitation)
• “Pedestrian or car driver observe a near-collision between another
pedestrian and another car because this pedestrian did not use the
striped area for crossing a street.”
• listening to other agents’ reports of their experiences
(normative learning)
• “One pedestrian tells another pedestrian: ‘You should use the striped
area for crossing a street!’”
•  Norm-invocation (messages)
•  Necessity of observer agents
09/10/2015
50 Years of Social Simulation
127/152
Informatik
Architecture of a Normative Agent
Basics
• Agents perceive events within the
environment in which they are situated and influence
the environment by corresponding actions
“Car Driver: Collision with a pedestrian”
 Environmental events
• Introduction of events which allow the assessment of
(environmental) events by positive/negative
valuations or sanctions ( normative learning)
“You should use the striped area for crossing a street!”
 Norm-invocation events
09/10/2015
50 Years of Social Simulation
128/152
Informatik
Basic Structures: Messages
Modals
- messages orginated from an
agent‘s perception
(assertion, behavior, …)
 Environmental messages
- messages received by
notifications from other agents
(valuation, sanction, …)
 Norm-invocation messages
 (Norm-oriented)
agent behavior
 Norm formation
Sender Recipient Modal Content Time Stamp
09/10/2015
50 Years of Social Simulation
129/152
Informatik
Basic Structures: Initial Rules
• Initial Rule Base: Describing the basic behavioural elements,
constituting the
seeds for more
complex rules
emerging from the
simulation process.





09/10/2015
50 Years of Social Simulation
„slow
down“
Environmental
„accelerate“
actions
„stop“
Norm-Invocation
„admonish“
actions
„honk the
horn“
130/152
Informatik
EMIL-S•
the first (simulated) minute (20 children, random cars
– children and cars run into each other, near-collision is interpreted as norm
invocation (“You have to stop when I am stepping on the street!”, “You
must not step on the street when I am around with my car!”)
•
several (simulated) minutes later (again 20 children, random cars)
– children have learnt that they have to use the striped area for street
crossing, car drivers have learnt that they are expected (obliged) to slow
down or stop in front of the striped area (which has emerged into an
institution after the first successful norm learning happened there) when
there are children visible in the neighbourhood
•
the same, some children have not learnt that the striped area is
something special
– some children still do not use the striped area but stop for an approaching
car
•
the same with perception sectors (only four children)
– approaching the street, children enlarge their perception area;
approaching the striped area, cars enlarge their perception area
09/10/2015
50 Years of Social Simulation
131/152
Informatik
Step
How EMIL-S works:
an overview of
our agent
architecture
Structure
1. E2 occurs in
ENV and is
added to EB
Environment
(ENV)
Event Board
Entry (EB)
2. Rule for E2
is retrieved
from IRB
3. Current ES is
saved in EB
E2
E2
1 .0
G2
0.33
A3
0.33
A4
0.33
A5
Environmental
State (ES)
E2
1 .0
G2
0.33
A3
0.33
A4
0.33
A5
0.5
1 .0
G1
0.5
Structure
A1
A2
6. Negative
valuation for
A4 arrives
1 .0
G2
0.33
A3
0.33
A4
0.33
A5
A1 A2 A3 A4 A5
0 0 0 0 0
E2
1 .0
G2
0.33
A3
0.33
A4
0.33
A5
7. Rule for NF is created from EB
entry
A1 A2 A3 A4 A5
0 0 0 10 0
8. Valuation is added to NF entry
and probability of valuated
action is adapted
Ne
gV
Environment
(ENV)
A4
al(
)
Event Board
Entry (EB)
E2
A1 A2 A3 A4 A5
0 0 0 0 0
E1
Step
5. ES is
adapted
A4
E2
A1 A2 A3 A4 A5
0 0 0 0 0
Initial Rule
Base (IRB)
4. Action is
selected and
sent to ENV
E2
1 .0
G2
0.33
A3
0.33
A4
0.33
A5
Normative
Frame Entry
(NF)
09/10/2015
R ule E 2 S 000000.33 A3
1 .0
G2
E2
0.33
0.33
50 Years of Social Simulation
A4
A5
A1 A2 A3 A4 A5
0 0 0 0 0
R ule E 2 S 000000.42 A3
1 .0
G2
E2
0.1 6
A4
0.42
A5
A1 A2 A3 A4 A5
0 0 0 0 0
132/152
NegVal(A4)
A1 A2 A3 A4 A5
0 0 0 0 0
Informatik
Other applications
• Agent-based modelling can also be applied to
politically relevant scenarios:
– emergence of loyalties within criminal organisations
and collusion between criminals and their victims: the
example of extortion rackets
– emergence of trust (and of mechanisms justifying
trust) in online transactions between sellers,
intermediaries and buyers
– ethnic conflicts: the emergence of consciousness of
belonging to a certain group
– emergence of practices in microfinance
09/10/2015
50 Years of Social Simulation
133/152
Informatik
Dynamics of Legality and Illegality: Agents
• DyLeg agents will be
– members of criminal or terrorist organisations,
– members of organisations which fight such
organisations,
– victims of such organisations
– supporters of such organisations and
– others who are something like a reservoir for the other
four breeds.
09/10/2015
50 Years of Social Simulation
134/152
Informatik
Dynamics of Legality and Illegality: Behaviours
• DyLeg agents will have to be able
– to influence one another, learn, etc.;
– to discriminate between social norms and coercive
requests;
– to distinguish revenge from normative sanction;
– to perceive not only those messages which were sent
to them individually, i.e. to listen to communication
surviving is always allowed and never
between others, and
commanded, paying extortion money is forbidden
in civil society but commanded in gangland,
– to cope with conflicting goals whistle-blowing
is allowed or even commanded by
civil society and forbidden in gangland.
• (e.g. surviving and not paying extortion money, surviving
and whistle blowing.)
09/10/2015
50 Years of Social Simulation
135/152
Informatik
EMIL-S features in DyLeg
• Agents can influence each other
– by direct (reporting) communication,
– by norm invocation and
– even physically,
• they make a difference between learning
– by explicit norm invocation (“forbidden”) and
– by direct experience or report from others (“dangerous”)
• (i.e. it is not necessary for them to experience or consider direct
negative or positive consequences of their behaviour once they
have been told that something is forbidden or commanded).
09/10/2015
50 Years of Social Simulation
136/152
Informatik
Systems of systems in DyLeg
cosa nostra
mandamento
famiglia
• DyLeg also necessitates a multi-level (not just
two-level) model:
cosa nostra, ‘ndrangheta, police, political party
– Besides the agents of different types, different kinds
and different levels of organisations have to be
modelled, where agents may be members of different
organisations at the same time
• (e.g. a member of a criminal organisation and an undercover
agent)
– and organisations might work differently (have
different norms) in different cultural contexts
– a requirement that is also easily fulfilled in EMIL-S.
09/10/2015
50 Years of Social Simulation
137/152
Informatik
Learning strategies
• Reinforcement learning: increase propensities of
successful strategies, decrease propensities of
unsuccessful strategies
• Recombination of event-action trees (similar to
crossover in genetic algorithms, but without
survival and selection over generations of
agents): learning to react to new events by
changing the structure of one or more eventaction trees
09/10/2015
50 Years of Social Simulation
138/152
Informatik
Changing the structure of event-actions trees
• copy, prune and graft …
E1: fellow has
betrayed other fellow
G1
G2
A11: blame him A:12 injure him
A:21 .. A:21 ..
E2: publican refuses
to pay extortion
G3
A:31 destroy his A:32 shoot
apartment
him
E1: fellow has
betrayed other fellow
G3
A:31 destroy his A:32 shoot
him
apartment
•
G2
G4
A41:..
A:42 ..
E2: publican refuses
to pay extortion
G1
A:21 .. A:21 .. A11: blame him A:12 injure him
G4
A41:..
A:42 ..
In the end, this means that actions of groups G1, G3, G4 and G2 can be
considered in both situations
09/10/2015
50 Years of Social Simulation
139/152
Informatik
Messages and their interpretations
• In the current examples, messages are still simple, but they are
already interpreted:
•
the message “Don’t deliver me into the hands of police even when you get caught!” is interpreted
as “this is a situation where both of us are in danger of being caught, and if one has a chance to
escape the other should do whatever possible that this escape is successful, as it is important that at
least one of us can escape and tell the others …”
• Unlike agents in gradient and pheromone metaphor models,
both sender and receiver of messages are “free” to make their
choices.
• Choices will depend on a long individual history.
–
–
–
09/10/2015
Whether a person gets infected by a virus and how severe the infection will be also depends
on a long individual history, but the outcome in this case is one-dimensional!
One can successfully vaccinate a person against her will to protect her from smallpox, but a
“vaccination” to protect a person from being infected with terrorism against his will is in vain
Choices made with a variety of decision trees are much more polymorphic.
50 Years of Social Simulation
140/152
Informatik
Can computational social science contribute to a
better understanding of complex social systems?
• Computational social science aims at
understanding the adaptive behaviour of humans
and systems of humans.
• Simulation is one way to improve [the
communication of] our understanding (to make
Adam Smith’s invisible hand visible) as a
simulation is analytically narrative and – in
contrast to verbal theory – produces data of the
same kind as the real world.
• And we can look into the minds of software agents.
09/10/2015
50 Years of Social Simulation
141/152
Informatik
References: General
•
•
•
•
•
•
•
•
Anderson, Jay M. The Eutrophication of Lakes, in: Dennis and Donnella Meadows: Toward Global
Equilibrium,Cambridge MA (Wright Allen) 1973, pp. 171–140
Bunge, Mario (1979). Treatise on Basic Philosophy. Volume 4: Ontology II: A World of Systems.
Dordrecht/Boston: Reidel
Carley, Kathleen M., Michael Prietula, eds. (1994): Computational Organization Theory.
Hillsdale/Hove: Lawrence Erlbaum
Carpenter, Stephen, and William Brock and Paul Hanson: Ecological and Social Dynamics in
Simple Models of Ecosystem Management. In Conservation Ecology 3 (2):4 1999, URL:
http://www.consecol.org/vol3/iss2/art4
de Sola Pool, Ithiel, and Robert Abelson. The Simulmatics Project. Public Opinion Quarterly 25,
1961, 167-183
Epstein, Joshua M., and Robert Axtell. Growing Artificial Societies. Social Science from the Bottom
Up. Cambridge, Mass., London: MIT Press, 1996
Forrester, Jay W. World Dynamics. Cambridge, Mass., London: MIT Press 1971
König, Andreas, Michael Möhring and Klaus G. Troitzsch. Agents, Hierarchies and Sustainability, in:
Billari, Francesco, and Alexia Prskawetz. Agent-Based Computational Demography. Berlin:
Physica 2003
09/10/2015
50 Years of Social Simulation
142/152
Informatik
References
•
•
•
•
•
Meadows, Dennis L., William W.,Behrens III, Donnella H. Meadows, Roger F. Naill, Jørgen Randers, and Erich
K.O. Zahn, (1974). Dynamics of Growth in a Finite World. Cambridge: MIT Press.
Meadows, Donnella H., Dennis L. Meadows, and Jørgen Randers (1992). Beyond the Limits. Post Mills: Chelsea
Green.
Meadows, Donnella H., Dennis L. Meadows, and Jørgen Randers(2004). The Limits to Growth: The 30-Year
Update . Post Mills: Chelsea Green.
Möhring, M. & Troitzsch, K.G. (2001). Lake anderson revisited. Journal of Artficial Societies and Social
Simulation, 4/3/1, http://jasss.soc.surrey.ac.uk/4/3/1.html.
Rouchier, Juliette, François Bousquet, Mélanie Requier-Desjardins, Martine Antona: A multi-agent model for
describing transhumance in North Cameroon: comparison of different rationality to develop a routine.
Journal of Economic Dynamics and Control, 2001, 25: 527-559.
09/10/2015
50 Years of Social Simulation
143/152
Informatik
References: General (continued)
•
•
•
•
Rouchier, Juliette, François Bousqet, Olivier Barreteau, Christophe LePage, Jean-Luc Bonnefoy:
Multi-Agent Modelling and Renewable Resources Issues: The Relevance of Shared Representations
for Interacting Agents, in: Moss, Scott, and Paul Davidsson: Multi-Agent-Based Simulation,
Springer, Berlin 2000 (LNAI 1979), pp. 181–197
Schelling, Thomas. Dynamic Models of Segregation. Journal of Mathematical Sociology 1971 (1),
143—186
Troitzsch, Klaus G. Multi-agent systems and simulation: a survey from an application perspective.
In Adelinde Uhrmacher and Danny Weyns, editors, Agents, Simulation and Applications, pages 2.1–
2.23. Taylor and Francis, London, 2008. to appear.
Zeigler, Bernard P. Theory of modelling and simulation. Malabar: Krieger 1985 (Reprint, originally
published: New York: Wiley 1976)
09/10/2015
50 Years of Social Simulation
144/152
Informatik
Further reading: System dynamics
• Mario Bunge. Ontology II: A world of systems. Treatise on basic philosophy, vol. 4. Reidel, Dordrecht, Boston, London, 1979.
• Diether Craemer. Mathematisches Modellieren dynamischer Vorgänge. Eine Einf ührung in die Programmiersprache
DYNAMO. Teubner, Stuttgart, 1985.
• Manfred Eigen and Peter Schuster. The Hypercycle. A Principle of Natural Self-Organization. Springer, Berlin, Heidelberg, New
York, 1979.
Jay W. Forrester. World Dynamics. MIT Press, Cambridge, Mass., London, 1971.
Jay W. Forrester. Principles of Systems. MIT Press, Cambridge, Mass., London, 1968, 2nd preliminary edition 1980.
Robert A. Hanneman. Computer-Assisted Theory Building. Modeling Dynamic Social Systems. Sage, Newbury Park, 1988.
Juan Carlos Martinez Coll. A bioeconomic model of Hobbes’ “state of nature”. Social Science Information, 25(2):493–505, 1986.
John Maynard Smith. Evolution and the Theory of Games. Cambridge University Press, Cambridge, 1982.
Dennis L. Meadows et al. Dynamics of Growth in a Finite World. MIT Press, Cambridge, Mass., London, 1974.
Dennis Meadows, Donella Meadows, Erich Jahn, and Peter Milling. Die Grenzen des Wachstums. Bericht des Club of Rome zur
Lage der Menschheit. Deutsche Verlagsanstalt, Stuttgart, 1972.
• Meadows, Donnella H., Dennis L. Meadows, and Jørgen Randers(2004). The Limits to Growth: The 30-Year Update . Post Mills:
Chelsea Green.
• Donella H. Meadows, Dennis L. Meadows, and Jørgen Randers. Beyond the Limits. Chelsea Green, Post Mills, Vermont, 1992.
• Donella H. Meadows, Dennis L. Meadows, and Jørgen Randers. Die neuen Grenzen des Wachstums. Die Lage der Menschheit:
•
•
•
•
•
•
•
Bericht und Zukunftschancen. Deutsche Verlagsanstalt, Stuttgart, 1992.
• Alexander L. Pugh III. DYNAMO User’s Manual. MIT Press, Cambridge, Mass., 1976.
09/10/2015
50 Years of Social Simulation
145/152
Informatik
Further reading : Microsimulation
•
•
•
•
•
•
•
•
•
•
•
Hauser, Richard, Uwe Hochmuth, and Johannes Schwarze. Mikroanalytische Grundlagen der Gesellschaftspolitik. Ausgewählte Probleme
und Lösungsansätze. Ergebnisse aus dem gleichnamigen Sonderforschungsbereich an den Universitäten Frankfurt und Mannheim, Band
1. Akademie-Verlag, Berlin, 1994.
Hauser, Richard, Notburga Ott, and Gert Wagner. Mikroanalytische Grundlagen der Gesellschaftspolitik. Erhebungsverfahren,
Analysemethoden und Mikrosimulation. Ergebnisse aus dem gleichnamigen Sonderforschungsbereich an den Universitäten Frankfurt
und Mannheim, Band 2. Akademie-Verlag, Berlin, 1994.
Habib, Jack. Microanalytic simulation models for the evaluation of integrated changes in tax and transfer reform in Israel. In Guy H.
Orcutt, Joachim Merz, and Hermann Quinke, editors, Microanalytic simulation models to support social and financial policy, Information
Research and Resource Reports, vol. 7, pages 117–134. North Holland, Amsterdam, New York, Oxford, 1986.
Harding, Ann. Microsimulation and Public Policy. Contributions to Economic Analysis. North Holland, Amsterdam, Lausanne etc. 1996
Heike, Hans-Dieter, Kai Beckmann, Achim Kaufmann, Harald Ritz, and Thomas Sauerbier. A comparison of a 4GL and an objectoriented approach in micro macro simulation. In Klaus G. Troitzsch, Ulrich Mueller, Nigel Gilbert, and Jim E. Doran, editors, Social Science
Microsimulation, chapter 1, pages 3–32. Springer, Berlin. Heidelberg, New York, 1996.
Henize, John. Critical issues in evaluating socio-economic models. In Tuncer I. O¨ ren, Bernard P. Zeigler, and Maurice S. Elzas, editors,
Simulation and Model-Based Methodologies: An Integrative View, NATO Advanced Science Institutes Series, Series F: Computer and
Systems Science, vol. 10, pages 557–590. Springer, Berlin, Heidelberg, New York, Tokyo, 1984.
Lietmeyer, Volker . Microanalytic tax simulation models in Europe: Developmentand experience in the German Federal Ministry of
Finance. In Guy H. Orcutt, Joachim Merz, and Hermann Quinke, editors, Microanalytic simulation models to support social and financial
policy, Information Research and Resource Reports, vol. 7, pages 139–152. North Holland, Amsterdam, New York, Oxford, 1986.
Merz, Joachim . MICSIM: Concept, developments, and applications of a PC microsimulation model for research and teaching. In Klaus G.
Troitzsch, Ulrich Mueller, Nigel Gilbert, and Jim E. Doran, editors, Social Science Microsimulation, chapter 2, pages 33–65. Springer, Berlin.
Heidelberg, New York, 1996.
Lavinia Mitton, Holly Sutherland and Melvyn Weeks (eds) (2000): Microsimulation Modelling for Policy Analysis: Challenges and
Innovations. Cambridge University Press, Cambridge.
Orcutt, Guy H., Joachim Merz, and Hermann Quinke, editors, Microanalytic simulation models to support social and financial policy,
Information Research and Resource Reports, vol. 7. North Holland, Amsterdam, New York, Oxford, 1986.
Sauerbier, Thomas. UMDBS — a new tool for dynamic microsimulation. Journal of Artificial Societies and Social Simulation, 5/2/5.
http://jasss.soc.surrey.ac.uk/5/2/5.html.
09/10/2015
50 Years of Social Simulation
146/152
Informatik
Further reading: Cellular automata
•
•
•
•
•
•
Hegselmann, R. (1996) ‘Cellular automata in the social sciences: Perspectives,
Restrictions and Artefacts’, in R. Hegselmann, U. Mueller and K. Troitzsch
(eds.) Modelling and simulation in the social sciences from the philosophy of
science point of view. Dordrecht: Kluwer.
Wolfram, S. (1986) Theory and applications of cellular automata. Singapore:
World Scientific.
Wolfram, S. (2002) A new kind of science. Wolfram Media.
Toffoli, T and Margolus, N. (1987) Cellular Automata Machines Cambridge,
Mass: MIT Press.
Nowak, A. and Latané, B. (1994) ‘Simulating the emergence of social order
from individual behaviour’, in N. Gilbert and J. Doran Simulating Societies,
London: UCL.
Lomborg, B (1996) ‘Nucleus and Shield: the evolution of social structure in
the iterated prisoner’s dilemma’, American Sociological Review Vol 61, 278307.
09/10/2015
50 Years of Social Simulation
147/152
Informatik
Further reading: Cellular automata
•
•
•
•
Thomas C. Schelling (1971) ‘Dynamic models of segregation’ J. Mathematical
Sociology, Vol.1, 143–186.
Thomas C. Schelling (1978) Micromotives and macrobehaviour. New York:
Norton
K. M. Kontopoulos (1993) The logics of social structure. Cambridge University
Press.
Stephen Wolfram (2002) A new kind of science. Wolfram Media.
09/10/2015
50 Years of Social Simulation
148/152
Informatik
Further reading: Agent-Based Models
•
•
•
•
•
Ahrweiler, P., Pyka, A., & Gilbert, N. (2004). Simulating knowledge dynamics in
innovation networks (skin). In R. Leombruni & M. Richiardi (Eds.), Industry and labor
dynamics: The agent-based computational economics approach. Singapore: World
Scientific Press.
Axelrod, R. (1997). Advancing the art of simulation in the social sciences. In R. Conte, R.
Hegselmann & P. Terna (Eds.), Simulating social phenomena (pp. 21-40). Berlin:
Springer.
Batten, D., & Grozev, G. (2006). Nemsim: Finding ways to reduce greenhouse gas
emissions using multi-agent electricity modelling. In P. Perez & D. Batten (Eds.),
Complex science for a complex world (pp. 227-252). Canberra: Australian National
University.
Deffuant, G., Amblard, F., & Weisbuch, G. (2002). How can extremism prevail? A
study based on the relative agreement interaction model. Journal of Artificial Societies
and Social Simulation, 5(4).
Dray, A., Perez, P., Jones, N., Le Page, C., D'Aquino, P., White, I., et al. (2006). The
atollgame experience: From knowledge engineering to a computer-assisted role
playing game. Journal of Artificial Societies and Social Simulation, 9(1).
09/10/2015
50 Years of Social Simulation
149/152
Informatik
Further reading:Agent-Based Models
•
•
•
•
•
•
•
Epstein, J. M. (1999). Agent-based computational models and generative social science.
Complexity, 4(5), 41-60.
Epstein, J. M., Axtell, R., & Project. (1996). Growing artificial societies : Social science from the
bottom up. Washington, D.C. ; Cambridge, Mass. ; London: Brookings Institution Press : MIT
Press.
Friedman-Hill, E. (2003). Jess in action : Rule-based systems in java. Greenwich, Conn.:
Manning.
Gilbert, N. (2006). A generic model of collectivities, ABModSim 2006, International
Symposium on Agent Based Modeling and Simulation University of Vienna: European
Meeting on Cybernetic Science and Systems Research.
Gilbert, N., & Abbott, A. (Eds.). (2005). Special issue: Social science computation (Vol. 110
(4)). Chicago: The University of Chicago Press.
Gilbert, N., & Terna, P. (2000). How to build and use agent-based models in social science.
Mind and Society, 1(1), 57 - 72.
Luke, S., Cioffi-Revilla, C., Panait, L., Sullivan, K., & Balan, G. (2005). Mason: A java multiagent simulation environment, . Simulation: Transactions of the Society for Modeling and
Simulation International, 81(7), 517–527.
09/10/2015
50 Years of Social Simulation
150/152
Informatik
Further reading: Agent-Based Models
•
•
•
•
•
•
•
Macy, M., & Willer, R. (2002). From factors to actors: Computational sociology and agentbased modeling. Annual Review of Sociology, 28, 143-166.
Miles, R. (2006). Learning UML 2.0. Sebastopol, CA: O'Reilly.
Ramanath, A. M., & Gilbert, N. (2004). The design of participatory agent-based social
simulations. Journal of Artificial Societies and Social Simulation, 7(4).
Schelling, T. C. (1971). Dynamic models of segregation. Journal of Mathematical Sociology, 1,
143-186.
Strader, T. J., Lin, F.-r., & Shaw, M. J. (1998). Simulation of order fulfillment in divergent
assembly supply chains. Journal of Artificial Societies and Social Simulation, 1(2).
Tesfatsion, l., & Judd, K. (2006). Handbook of computational economics (Vol. 2): NorthHolland.
Wilensky, U. (1999). NetLogo. Evanston, IL: Center for Connected Learning and ComputerBased Modeling, Northwestern University.
09/10/2015
50 Years of Social Simulation
151/152
Informatik
Further reading: EMIL and EMIL-S
•
•
•
•
•
•
•
Andrighetto, Giulia, Marco Campennì, Rosaria Conte, and Marco Paolucci. On the
immergence of norms: a normative agent architecture. In Proceedings of AAAI Symposium,
Social and Organizational Aspects of Intelligence, Washington DC, 2007.
Andrighetto, Giulia, Rosaria Conte, and Paolo Turrini. Emergence in the loop: Simulating
the two way dynamics of norm innovation. In Guido Boella, Leendert van der Torre, and
Harko Verhagen, editors, Dagstuhl Seminar Proceedings 07122, Normative Multi-agent
Systems, Vol. I, 2007.
Andrighetto, Giulia, Marco Campennì, Federico Cecconi, and Rosaria Conte. Conformity in
Multiple Contexts: Imitation vs. Norm Recognition. Paper submitted to WCSS 08.
Campennì, Marco. The norm recogniser at work. Presentation at AAAI'2007, Washington.
Lotzmann, Ulf, and Michael Möhring. A TRASS-based agent model for traffic simulation.
Paper presented at the 22nd European Conference on Modelling and Simulation ECMS
2008.
Lotzmann, Ulf , Michael Möhring, Klaus G. Troitzsch. Simulating Norm Formation in a Traffic
Scenario. Paper accepted for ESSA 2008.
Troitzsch, Klaus G. Collaborative Writing: Software Agents Produce a Wikipedia. Paper
accepted for ESSA 2008.
09/10/2015
50 Years of Social Simulation
152/152
Descargar

Document