488873

research-article2013
POS43310.1177/0048393113488873Philosophy of the Social SciencesMarchionni and Ylikoski

Article
Philosophy of the Social Sciences
43(3) 323340
Generative Explanation The Author(s) 2013
Reprints and permissions:
and Individualism in sagepub.com/journalsPermissions.nav
DOI: 10.1177/0048393113488873
Agent-Based Simulation pos.sagepub.com

Caterina Marchionni1 and Petri Ylikoski1

Abstract
Social scientists associate agent-based simulation (ABS) models with three
ideas about explanation: they provide generative explanations, they are
models of mechanisms, and they implement methodological individualism.
In light of a philosophical account of explanation, we show that these ideas
are not necessarily related and offer an account of the explanatory import of
ABS models. We also argue that their bottom-up research strategy should
be distinguished from methodological individualism.

Keywords
agent-based simulation, explanation, mechanism, methodological individualism

1. Introduction
Over the past two decades, agent-based simulations (ABSs) have been
increasingly employed throughout the social sciences. However, the method-
ology of ABS is not yet sufficiently understood, and its legitimacy and impli-
cations are still subject of debate. For example, many economists are wary of
simulations because they regard them as inferior vis--vis analytical models

Received 10 April 2013

1University of Helsinki, Helsinki, Finland

Corresponding Author:
Caterina Marchionni, Finnish Centre of Excellence in the Philosophy of the Social Sciences,
Department of Political and Economic Studies, P.O. Box 24, 00014 University of Helsinki,
Finland.
Email: caterina.marchionni@helsinki.fi
324 Philosophy of the Social Sciences 43(3)

(Lehtinen and Kuorikoski 2007). In contrast, advocates of ABS methodology

praise its ability to overcome the limitations of traditional formal modeling
tools (Macy and Flache 2009; Miller and Page 2007).
In this article, our focus is on the explanatory uses of ABS (though of course
ABS is used for other purposes as well). In particular, we examine three ideas
about explanation that are often associated with the methodology of ABS: it
provides explanations that are generative (J. M. Epstein 2006) and mechanistic
(e.g., Gilbert and Ahrweiler 2009; Hedstrm 2005; Macy et al. 2011) and it is
an implementation of methodological individualism (e.g. Macy and Flache
2009; Neumann 2008). We argue that, contrary to what is sometimes believed,
these ideas are not necessarily related. With the help of a theory of explanation,
we offer an account of how ABS models explain. We show that the ideas of
generation and mechanism are sufficient to define the bottom-up research strat-
egy of ABS, but that the latter should not be conflated with the doctrine of
methodological individualism.
The structure of the article is the following. Section 2 briefly introduces
the main features of ABS in the social sciences. Section 3 presents our
account of how model-based simulations provide explanations. In Section 4,
we show how the idea of bottom-up explanation is related to ideas of genera-
tive and mechanism-based explanation. Section 5 brings up the idea of meth-
odological individualism. Section 6 presents an example of ABS study and
shows that some of its key explanatory variables are structural rather than
individual properties. Finally, Section 7 concludes our argument that it is
misleading to regard ABS models as implementations of methodological
individualism.

2. ABS in Social Science

In this section, we briefly summarize the main features of ABS as they are
typically employed in social science. ABS includes a set of agents and a set
of rules describing the behavior of those agents. Agents interact with each
other and with the environment, and these local interactions often bring
about surprising phenomena at the social level. As summarized by Macy
and Flache (2009, 247), the agents in ABS models are typically endowed
with interesting properties at the cognitive and social levels. Agents are
heuristic in that they follow simple behavioral rules that can be interpreted
as habits, rituals, routines, norms, and the like. They are also adaptive in
that they respond to feedback from their environment through learning and
evolution. Agents are autonomous and able to change aspects of their envi-
ronment to attain their goals. Their autonomy, however, is constrained:
agents are assumed to be interdependent insofar as the local environment to
Marchionni and Ylikoski 325

which they react is constituted by the beliefs, goals, and behaviors of other
agents. ABS methodology also permits the modeling of heterogeneous
agents, namely, agents who differ in their beliefs, goals, and rules of behav-
ior. Finally, agents can be embedded in networks. All this means that popu-
lation dynamics are emergent outcomes of local interactions. ABS also
allows agents to change their structural locations or break off relations with
their neighbors and seek out new relations (Macy and Flache 2009; Miller
and Page 2007).
Compared with the traditional modeling tools employed in the social sci-
ences, ABS is claimed to have a number of advantages. First, as we have
seen, simulations allow one to be flexible about the characteristics of agents
and hence break away from the strictures of the assumption of optimizing
behavior characteristic of rational choice models. Furthermore, ABS allows
modelers to study equilibrium outcomes as well as the dynamics of systems.
Finally, unlike standard economic models, where for reasons of tractability
the modeler has to work with cases with either one, two, or an infinite number
of agents, ABS can work with any number of agents and hence can provide
more realistic models of social processes (Macy and Flache 2009; Miller and
Page 2007; Ylikoski, forthcoming.).

3. Explaining with Simulation Models

Our account of explanation is realistic in spirit, for it is about tracking the
network of dependencies related to the phenomenon of interest. In other
words, scientific understanding is constituted by knowledge of dependencies,
knowledge that gives us the ability to make correct what-if inferences about
the phenomenon (Woodward 2003; Ylikoski and Kuorikoski 2010). The
point of (explanatory) models is to represent those dependencies in a cogni-
tively salient way, so that they can be used to make what-if inferences about
the phenomena under investigation. In ABS models, we can call assumptions
that pick out explanatorily relevant factors explanans variables and the out-
come to be explained the explanandum variable.
In this account, mechanisms can be understood as descriptions of the
networks of counterfactual dependencies that characterize the system in
question. Knowledge about the components of the system, their properties,
and their organization makes it possible to understand the counterfactual
dependencies that characterize the system and their background conditions
(Craver 2007; Woodward 2003; Ylikoski and Kuorikoski 2010). One of
the advantages of ABS methodology is that it allows the systematic study
of how the organization of the components affects the overall behavior of
the system.
326 Philosophy of the Social Sciences 43(3)

How can ABS models be used to provide explanations? To answer this

question, we distinguish two elements in the process of using ABS for
explanation. First, there is the generative component of the simulation
practice. Here, the simulator sets the rules of interaction for the agents and
runs the simulation to produce some macro-outcome. If the macro-out-
come has the characteristics of interestsuch as ending with a segregated
equilibrium (Schelling 1978) or creating a pattern found in empirical
material (J. M. Epstein 2006)then we have a proof of possibility. In such
cases, the simulation demonstrates that the mechanism implemented
therein can, in principle, generate the pattern of interest.
While such a proof of possibility is valuable, its contribution to explana-
tion is limited. The simulation does not allow us to identify the conditions
under which the mechanism produces the outcome nor which assumptions of
the model are responsible for the outcome of interest. In other words, the
simulation does not provide us an understanding of how the outcomes sys-
tematically depend on the assumptions of the model. Such an understanding
is obtained in the experimental component of simulation practice.
The experimental component consists of the systematic variation of the
simulation assumptions.1 The point of such experimentation is to see how
the assumptions make a difference to the outcome. It provides an understand-
ing of how the outcome of the simulation depends on various assumptions,
thus providing knowledge about the relevant networks of dependence. On the
basis of this knowledge, the simulator can make what-if inferences about the
simulation. As his or her ability to make such inferences increases, his or her
understanding of the simulation model also increases.
All ABS models include a generative component, but some almost com-
pletely lack an experimental component. An example of such a simulation is
Kirman and Vriends (2001) simulation of the Marseille fish market. The
stated goal of this model is to explain two stylized facts that characterize the
fish market in Marseille, namely, the high loyalty of buyers to sellers and the
persistence of price dispersion (Kirman and Vriend 2001). These characteris-
tics are puzzling, as the same buyers and sellers meet every day in the same
place. In their ABS model, there are two kinds of agents, namely, buyers and
sellers. Sellers decide on the quantities to supply, the prices, and whether to
treat loyal buyers differently. Buyers only decide which sellers to visit and
which price to accept. Here, it is assumed that the agents learn through rein-
forcement. In the words of Kirman and Vriend (2001, 459),

1Mki (2005) and Morgan (2003) compare the manipulation of modeling assumptions
to experimental manipulation.
Marchionni and Ylikoski 327

The model explains both stylized facts, price dispersion and high loyalty. In a
coevolutionary process, buyers learn to become loyal as sellers learn to offer
higher utility to loyal buyers, while these sellers, in turn, learn to offer higher
utility to loyal buyers as they happen to realize higher gross revenues from loyal
buyers.

In our view, the authors may be overstating the extent to which the model
explains the stylized facts. This is not to understate the importance of this
model and its result. As their simulation can generate high loyalty and price
dispersion, it achieves an abstract proof of possibility. Such a proof shows
what kind of assumptions could produce the outcome, but not how that occurs
or whether those assumptions are the only way to generate the outcome of
interest. The model does not provide much insight into the crucial how-ques-
tions, as the experimental component is almost completely absent (the authors
only vary one condition, namely, the heterogeneity of buyers). Thus, insofar
as the authors do not systematically explore how the outcome of interest
depends on the details of the model, the full explanatory import of the model
remains an open question.
On our account of explanation, to provide a proper explanation of the
phenomenon in question, the simulator should show not only that the assump-
tions made about the agents bring about the observed macro-outcome but
also how they do so. This is done by spelling out the mechanism implemented
in the simulation and by showing how the differences in outcomes systemati-
cally depend on changes in the assumptions of the simulation. When simula-
tion practice involves an experimental component, the assumptions of the
simulation are systematically varied to learn the effects of these changes on
the outcome of the simulation. This makes it possible to learn more about the
network of counterfactual dependencies that characterizes the simulated sys-
tem. This network of dependencies is precisely what the description of the
explanatory mechanism implemented in the simulation amounts to.
Note that there are two reasons for the systematic variation of the assump-
tions of a simulation. The first is to learn about the systematic dependencies
between the explanans and the explanandum within the specified mechanical
configuration. The second is to learn which unrealistic assumptions matter for
the models results, a purpose that is typically achieved by robustness analysis
(Kuorikoski, Lehtinen, and Marchionni 2010; Levins 1966; Weisberg 2006;
Wimsatt 1981). Both purposes are equally important for the explanatory use of
simulations, but the contribution of robustness analysis to explanation is only
indirect. Usually robustness analysis is aimed at investigating the role of those
assumptions made to facilitate the tractability of the underlying model or of
the simulation itself (as, for instance, when a city is represented as a
328 Philosophy of the Social Sciences 43(3)

checkerboard). Such assumptions are known to be unrealistic but are made

because replacing them with realistic assumptions is either difficult or impos-
sible in the first place. In robustness analysis, such assumptions are usually
replaced with other unrealistic assumptions (as, for instance, when an environ-
ment is modeled as a grid and then as a torus). The point is to learn whether
the change makes a difference to the simulation outcome. If the result is robust
with respect to the change, the researcher can infer that the given unrealistic
assumption does not play a crucial role in the model and thus it is not an
important explanatory variable in the model.
As the purpose of robustness analysis is to identify those assumptions that
do not play a central role in the model, it does not matter whether they are
realistic. In contrast, explanatory analysis targets those assumptions that do
affect the simulation outcomes. Here, it is important that the assumptions
make a difference within a realistic range of variation. Only these assump-
tions enable inferences to real-world systemseven though whether a given
modification has a reasonable empirical counterpart is an empirical question
through and through.
Sometimes, modelers do explicitly distinguish between the two kinds of
manipulations, but this is not the main criterion for drawing the distinction.
In many cases, robustness analysis can direct attention to factors that were
previously thought to be irrelevant (or vice versa), thus defying the modelers
expectations. In this sense, robustness analysis and the variation of assump-
tions for explanatory purposes are complementary and often concurrent prac-
tices. At an abstract level, the purpose of both is to explore the properties of
the model and its components to learn what depends on what.
More importantly for our purposes, the distinction between the two differ-
ent aims of variation of a simulations assumptions provides the criteria for
identifying explanatory variables in ABS models: we should look for assump-
tions that are varied by the simulators (in a more or less systematic way),
make a difference to the explanandum variables of interest, and are assumed
to have a realistic sociological interpretation. We will illustrate these points
in Section 6 with the help of a simulation model of the spread of unpopular
norms, which includes manipulations of various of its assumptions.

4. The Bottom-Up Strategy of Explanation in ABS

The typical aim of ABS is to provide explanation for macrophenomena such
as the emergence and dynamics of system-level properties (e.g., distributions
of beliefs and behaviors within the population, levels of segregation, net-
works between the agents). In contrast to more traditional equation-based
simulation approaches, ABS does not directly relate macrovariables to each
Marchionni and Ylikoski 329

other; rather, it represents a bottom-up research strategy that investigates

macrophenomena by focusing on agents and their properties, behaviors, and
interactions (bottom). It shows how what occurs at this level produces, or can
produce, the phenomena of interest (up). When building the model, the simu-
lator specifies the rules of behavior for the parts of the system (the agents)
and studies how their local interactions generate the relevant macro-out-
comes. Thus, the simulation requires an explicit consideration of the relevant
causal mechanisms. This is in sharp contrast to variable-based simulations,
which often operate with macrovariables and do not include explicit thinking
about mechanisms. Furthermore, the macro-outcomes of ABS models are
often surprising, which supports the idea of a bottom-up research strategy.
The concept of bottom-up research strategy suffices to make sense of two
ideas about explanation in ABS: generation and mechanism. We briefly
examine each in turn and then argue that methodological individualism does
not capture anything useful that is not captured by the idea of bottom-up
explanation.

4.1. Generative Explanation

As the purpose of ABS is to obtain macro-outcomes from assumptions about
agents, their interdependent behaviors, and the features of the environment in
which they act, it is quite natural then to interpret ABS models in terms of
generative explanation. Joshua M. Epstein (2006) gives the strongest formu-
lation of the idea of generative explanation. According to J. M. Epstein,
growing the macro-level outcome explanandum with ABS is a necessary con-
dition for its explanation. This is the origin of the well-known slogan: If you
didnt grow it, you didnt explain it (J. M. Epstein 2006, 8-10). J. M. Epsteins
claim points toward an important feature of ABS: if we are interested in the
bottom-up explanations of social phenomena, ABS provides a much-needed
tool. However, J. M. Epsteins claim is overstated (Ylikoski, forthcoming):
First, it is reasonable to assume that explanations of social phenomena have
been given before the use of ABS. Second, it is often the case that the result
obtained in a given simulation could be generated in a number of alternative
ways; thus, the simulation only provides a how-possibly explanation. Hence,
generative sufficiency is an essential explanatory step, but it does not imply
that the empirical outcome to be explained was actually produced in that
fashion. Third, simply growing the phenomenon of interest is not sufficient
for a proper understanding of it. Merely having a simulation thatafter a lot
of tinkeringproduces the right kind of results does not yet yield under-
standing as it does not enable the making of what-if inferences beyond that
particular simulation configuration. The further challenge is to understand
330 Philosophy of the Social Sciences 43(3)

how the specified microconfiguration produces the phenomenon and what

the background conditions of its production are. Thus, Macy and Flache
(2009, 263) are right when they challenge J. M. Epsteins slogan by claim-
ing, If you dont know how you grew it, you didnt explain it. The idea of
generation is necessary for explanatory understanding, but it is not sufficient.
We also need to understand the working of the simulation and the systematic
dependencies that the experimental component aims to uncover. The idea of
mechanism aims to capture precisely these elements of ABS.

4.2. Mechanistic Explanation

In the ABS literature, the idea of generative explanation is often associated
with the idea of mechanism-based explanation (Hedstrm and Ylikoski 2010)
as suggested, for example, by the following quotation:

When we write a set of computational algorithms (the program), formalizing the

generative hypotheses the consequences of which are to be studied, what we are
doing is hypothesizing a series of generative mechanisms. When we execute the
program . . . we engender the process deriving from the set of posited generative
mechanisms. With the technical distinction between program writing,
compilation, and execution it becomes clear that a process is nothing more
than the dynamic aspect of one (or several) mechanism(s): it is what the mechanism
can trigger. (Manzo 2007, 5-6)

One of the premises of the mechanistic approach in social science is that a

proper explanation of social phenomena requires an understanding of the
causal mechanisms that bring them about. This implies paying attention to the
entities of which mechanisms are made (the agents, their properties, actions,
and relations) rather than treating them as black boxes. That is, for the expla-
nation of social phenomena, it is not sufficient to identify the macro-level
changes that produce them. It is crucial to show how macrostates affect indi-
viduals at a certain point in time and how the actions of those individuals
produce new macrostates at a later time (Hedstrm and Ylikoski 2010). From
this perspective, ABS represents an especially useful tool for theoretically ori-
ented social scientists: it allows testing whether certain theoretical ideas are,
even in principle, sufficient to generate the phenomenon in question (J. M.
Epstein 2006). Furthermore, ABS models can easily be interpreted as repre-
senting social interactions: the agents can be interpreted as having goals and
beliefs, standing in various social relations to each other, having the resources
needed to attain their goals, and also affecting each others possibilities of
action via their behavior. As it should be clear by now, the methodology of
Marchionni and Ylikoski 331

ABS, and in particular its experimental component, facilitates the provision of

mechanism-based explanations as it enables the systematic study of how the
macrophenomenon is generated from assumptions about agents as well as
from various nonindividualistic assumptions.

5. Is This Methodological Individualism?

The bottom-up research strategy suffices to capture the potential of ABS as
tools for yielding generative and mechanistic explanations. We have seen that
the generative and experimental components of simulation practice enable
the simulator to explain macrophenomena by focusing on agents and their
properties, behaviors, and interactions and by showing how what occurs at
this level generates or can generate the phenomena of interest. Many authors,
however, have suggested that ABS is an implementation of methodological
individualism, or that there is an inherent connection between ABS and meth-
odological individualism. For example, Macy et al. (2011, 252) write that
ABS modeling is a formal implementation of methodological individual-
ism (see also Neumann 2008). The ideas of generative mechanisms and
methodological individualism are sometimes packaged together. For exam-
ple, Keith Sawyer (2004, 263) states that [a]rtificial societies are mechanis-
tic in the contemporary sense. They are also firmly methodologically
individualist (see also Sawyer 2003, 340). Our goal is to show that there is
no inherent (let alone necessary) connection between ABS, the generative
mechanism-based explanations they can deliver, and methodological indi-
vidualism. In fact, as we will argue, there is much to gain by breaking down
this connection.
What leads some practitioners and commentators to associate ABS with
methodological individualism is quite clear. In any sensible interpretation,
methodological individualism would imply something like a bottom-up
research strategy (although, of course, not all methodological individualists
are likely to be excited about the ABS methodology). However, the crucial
question is whether the opposite implication holds true. Here, the problem is
in discerning what the doctrine of methodological individualism actually
claims. As is well known, methodological individualism has had a variety of
meanings and interpretations over the years (Hodgson 2007; Kincaid 1997;
Pettit 1993; Udehn 2001). Consequently, the term lacks a clear and generally
accepted definition. This is a problem for advocates and critics of method-
ological individualism. As we believe that the debate over the proper defini-
tion of methodological individualism is a distraction from the real
methodological issues, we will not review the many interpretations of meth-
odological individualism available in the literature. Instead, we will work
332 Philosophy of the Social Sciences 43(3)

with a definition of methodological individualism that is substantive enough

to be nontrivial but that retains the typical connotation of the concept.
Therefore, we take the doctrine of methodological individualism to
roughly pertain to the following:

[MI] Social phenomena can only be explained (or they are best explained) by
accounts that only refer to individuals, their properties and their interactions.

There are two aspects to note about this reading of methodological indi-
vidualism. First, [MI] is a thesis about explanation, not about ontology.
Therefore, arguments about the existence of social wholes, structures, and
such entities vis--vis individuals do not directly bear on arguments about
explanation of social phenomena. Second, [MI] qualifies as a strong version
of methodological individualism in that it holds that explanation of social
phenomena should appeal only to individuals, their properties, and interac-
tions. The corollary of such a view is that nonindividual properties are denied
nonderivative explanatory status.
Some individualists endorse weaker versions of the thesis, according to
which nonindividual properties can play an (nonderivative) explanatory role
via their effects on individuals. It is legitimate to ask, however, in what sense
these more liberal positions are individualistic and whether they represent the
same position that many anti-individualists are trying to defend (see Udehn
2001). As said above, we do not want to get entangled in debates about the
proper definition of individualism. For our argument, it is sufficient that [MI]
is sufficiently close to what is commonly understood as the content of the
methodological individualist doctrine.

6. An Example: A Model of the Spread of Unpopular

Norms
Our strategy against the association of ABS with methodological individual-
ism is to argue that if some of the key explanatory variables in ABS models
cannot be meaningfully interpreted as individualistic, then ABS explana-
tions should not be regarded as implementations of methodological individu-
alism. To make such an argument, we need a real example of an explanatory
ABS model. We have chosen a simulation model designed by Damon Centola,
Robb Willer, and Michael Macy (2005) as we think it is one of the best exam-
ples of the use of an ABS model in theoretical sociology, and the authors have
elsewhere suggested that there is a close relationship between ABS and meth-
odological individualism (Macy et al. 2011).
Marchionni and Ylikoski 333

The model implements the emperor dilemma familiar from the Hans
Christian Andersens fable. In the model, agents must decide whether to com-
ply with and enforce a norm that is supported by a few fanatics and opposed
by the vast majority. The model is used to examine the population-level
implications of the use of norm enforcement to falsely signal genuine convic-
tion. The idea is to study whether a very small fraction of true believers can
spark a cascade of conformity and false enforcement that quickly engulfs a
vulnerable population. Thus, the norm does not become enforced because
people are converted to new beliefs; rather, it is because they feel the need to
affirm the sincerity of their (false) conformity.
In the simulation, the population consists of agents who differ in their
beliefs and convictions. A small group of true believers is assumed to have
such strong convictions that they always comply with the norm. When dis-
satisfied with the level of compliance of others, they may enforce the norm.
The remainder of the population consists of disbelievers who privately
oppose the norm, but with less conviction than that of the true believers. The
disbelievers may deviate from the norm or even pressure others to deviate as
well. However, the disbelievers can also be pressured to support the norm and
even to enforce it. At every iteration of the simulation, each agent observes
how many of his or her neighbors are complying with the norm and how
many are deviating. They also observe how many neighbors are enforcing the
compliance and how many are enforcing deviations from the norm. Based on
this information, the agents decide whether to comply or deviate and whether
to force others to behave similarly in the next round.
In the article, Centola, Willer, and Macy (2005) report the manipulation of
three kinds of variable: (1) access to information about the behavior of other
agents, (2) the frequency distribution and clustering of true believers, and
(3) the network topology. The results of these simulations are surprising: cas-
cades are much easier to achieve than expected. A small group of true believ-
ers can bring about a cascade in population where neighborhoods are local;
however, they are unable to do so in fully connected populations. Moreover,
the clustering of true believers turns out to be relevant: a very small cluster of
believers can trigger a cascade, while a great number of randomly distributed
believers cannot. Finally, when a small number of random ties reduce the
overlap between local neighborhoods, cascades are prevented. On the basis
of these observations, the authors conclude that unpopular norms thrive on
local misrepresentations of the underlying population distribution (Centola,
Willer, and Macy 2005, 1034). However, the most interesting result is that
disbelievers are crucial for the emergence of cascades. Without them, cas-
cades do not begin, and if the agents start to convert into true believers, the
following of the norm might paradoxically collapse.
334 Philosophy of the Social Sciences 43(3)

Now let us take a closer look at the manipulated variables. From the point
of view of our argument, the crucial question is whether they are individual
or structural properties. By structural properties, we mean properties that are
attributed to larger scale entities than individuals or if they are attributed to
individuals, they presuppose some larger scale entities. These nonindividual
properties constitute a rather heterogeneous class; what they share in com-
mon is the property of being nonindividual properties (Ylikoski 2012). The
network topology is clearly a structural assumption about the macrostructure
of the population. It is a structural property in the sense that there is no mean-
ingful way to attribute it to an individual; it is always attributed to a larger
scale entity. Similarly, the frequency and degree of clustering of true believ-
ers (and other agents) is a population-level attribute that cannot be applied to
individuals. Finally, while the access to information about other agents is
attributed to an individual agent, it is more properly understood as a structural
assumption about relations between agents. The relations between individual
agents and the overall configuration of these relations in the population are
population-level attributes that do not apply to individuals.
Thus, all three key variables are rather prototypical nonindividual struc-
tural macroproperties (Ylikoski 2012). Furthermore, all three satisfy our sug-
gested criteria for explanatory variables. First, they make a difference to the
outcome, as whether or not a cascade is triggered depends on them. Second,
they also have a realistic sociological interpretation, as they capture the
degree to which agents can obtain an accurate picture of how widespread
genuine belief in a given norm is.
Explanatory structural variables like these are not unique to this particular
case. It is quite common to find ABS models that focus on variables such as
the composition of the population, contacts between agents, and agents free-
dom of movement (e.g., Centola and Macy 2007; Flache and Macy 2011).
This is the case even for the relatively simple segregation models inspired by
Thomas Schellings work (e.g., Benard and Willer 2007; Bruch and Mare
2006; Clark and Fossett 2008; Fossett 2006; also see Ylikoski, forthcoming).
One of the attractions of ABS modeling is precisely that such assumptions
can be systematically manipulated together with assumptions about individu-
als. For example, the effects of changes of the size of the neighborhood with
which agents interact or of the connections that agents have beyond their
immediate neighborhood can be studied, for example, by a random rewiring
of the links between agents (e.g., Centola and Macy 2007). The agents free-
dom of movement brings in another set of structural assumptions, such as the
relative size of the available empty spaces and the rules for movement across
them. The experimental manipulation of structural and individual assump-
tions makes it possible for the simulator to investigate the network of
Marchionni and Ylikoski 335

dependencies (i.e., the mechanism) that characterizes the phenomenon to be

explained. What counts as part of a mechanism depends on what makes a
difference to the macrophenomenon, regardless of whether the variables are
structural or individualistic.

7. Breaking the Association between Methodological

Individualism and ABS
As the model of Centola, Willer, and Macy (2005) demonstrates, some of the
crucial explanatory variables in ABS are structural. This constitutes a prima
facie case against the association of ABS with methodological individualism.
The burden of proof then shifts to those who wish to make the association:
they should show that those explanatory variables are either not explanatory,
or that they can be credibly interpreted as individualistic. The discussion
above makes clear that denying their explanatory relevance is not a viable
strategy; reinterpreting them as individualistic properties does not look a very
promising strategy either. In this case, the challenge is to find a credible for-
mulation of individualistic properties that could still accommodate the rele-
vant explanatory variables. The definition of individualistically acceptable
properties should not be allowed to trivialize the debate between traditional
individualists and their critics. To make this argument more compelling, we
offer some substantial arguments against the strong version of methodological
individualism and some points of caution against associating ABS with meth-
odological individualism that apply to its strong and weak formulations.
Our first substantial argument is based on the observation that ABS meth-
odology does not dictate how the agents are to be interpreted: while in many
applications it is natural to interpret agents as individuals, they can also be
households, groups, organizations, or even intrapersonal cognitive processes.
The interpretation of the simulation only requires that it makes sense to
ascribe to the agents the behavior defined by the rules. Consequently, even
within the same simulation, agents can be of various kinds. Thus, in this
respect, ABS is not by itself individualistic, and restricting its use to individu-
alistically acceptable applications (i.e., applications in which agents are to be
interpreted as individuals) would require strong arguments. Whether such
arguments can be provided is an open question in which advocates of meth-
odological individualism carry the burden of proof.
Another strategy for the individualist is to claim that the structural assump-
tions we have identified as doing explanatory work do not pose a problem
because structural properties can be shown to result from purely individual-
istic processes. For example, one might claim that network structures are
336 Philosophy of the Social Sciences 43(3)

generated by certain kinds of individual preferences in interaction, thus mak-

ing them individualistically acceptable. Our replies to this objection consti-
tute the second argument against associating ABS with methodological
individualism. First, within a given ABS, variables such as network struc-
tures are structural and explanatory. That these variables could be themselves
explained in individualist terms does not imply that their explanatory rele-
vance vanishes. This kind of regress argument does not work in the case of
explanation: the explanatory status of a variable does not depend on whether
it can itself be explained (Ylikoski 2012). Second, the individualist would
still have to demonstrate that it is, in fact, possible to give a purely individu-
alistic explanation for these kinds of variables. It is difficult to see how this
might happen (Kincaid 1996); nevertheless, the burden of proof is again with
the individualist.
Even if the individualist were to stretch his or her idea of methodological
individualism to cover both the cases presented above, what would be the
benefit of such a dialectical victory? Probably none.
Our first point of caution is the observation that the track record of the
debate around methodological individualism in resolving real micromacro
problems in the social sciences is controversial. It can be argued that such prob-
lems can be discussed more fruitfully without getting bogged down by issues
related to the proper definition of individual properties and individual
level, or by attempts to clarify different meanings of methodological individu-
alism.2 The notion of bottom-up strategy of explanation we have articulated
above suffices to capture the idea of generation and the virtue of understanding
mechanisms, without implying any commitments to principles like [MI].
Moreover, the association of ABS with methodological individualism can
make some social scientists less receptive to the productive possibilities of
ABS methodology. Considering that the narrow and negative understanding of
methodological individualism is common within the social scientific com-
munity, an overly close association between ABS and methodological individ-
ualism will only turn some potentially interested people away from ABS.
Finally, there is the danger that a conceptualization of micromacro prob-
lems in terms of methodological individualism may lead to biased strategies
in ABS research (Wimsatt 1983). For example, understanding the agents in
ABS exclusively as individuals might make researchers blind to large-scale
structural factors that have significant explanatory import and that can be
effectively modeled with ABS methodology. Similarly, thinking of properties

2Ylikoski (2012) offers an account of macromicro relations that dispenses with many
of the problems that afflict the individualismholism debate.
Marchionni and Ylikoski 337

like network topologies as explanatorily inert may lead to researchers sys-

tematically choosing to make simplifications and idealizations at this level
rather than at the level of the individuals. In turn, this could lead to a situation
in which more realistic accounts of the networks in which individuals are
embedded are seldom tried out. Both kinds of bias would be unfortunate for
the development of ABS research because one of its advantages is precisely
that it allows social scientists to overcome (some of) the traditional limita-
tions of social scientific model-building and theorizing.
It is important to note that our argument against methodological individu-
alism does not in any way imply that the traditional opponents of individual-
ism are right. Our point is that the association of ABS with methodological
individualism does nothing to advance our understanding of how ABS mod-
els explain, while it may turn out to be an obstacle to their fruitful use. It may
be true that ABSs exemplify many epistemic virtues, such as an explicit con-
sideration of mechanisms and a focus on agents, which traditional method-
ological individualists have found important. However, these virtues are
independent of methodological individualism. Thus, our suggestion is to
ignore the individualismholism debate and focus instead on how the bot-
tom-up research strategy that characterizes ABS can offer new ways of study-
ing micromacro relations.

8. Conclusion
Philosophical reflections on social simulations have been scarce compared
with general accounts of the epistemology of simulations and specific analy-
ses of simulations in the natural sciences (see, however, B. Epstein 2012;
Grne-Yanoff and Weirich 2010). In this article, we have offered an account
of the way in which ABS is used for explanatory purposes in social science.
The bottom-up research strategy embodied in typical ABS delivers genera-
tive explanations: it generates the macrophenomenon to be explained by
appeal to the actions and interactions of the agents. The bottom-up research
strategy of ABS can also yield mechanism-based explanations: it tracks the
(possible) network of dependencies behind the phenomenon to be explained.
However, we argue that the idea that ABS is an implementation of method-
ological individualism is misleading because structural assumptions often
play an irreducible explanatory role. Finally, we have offered practical (rather
than conceptual) arguments for resisting the association of ABS with meth-
odological individualism, even when the latter is interpreted to accommodate
the explanatory role of nonindividualistic assumptions.
338 Philosophy of the Social Sciences 43(3)

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research,
authorship, and/or publication of this article.

Funding
The author(s) disclosed receipt of the following financial support for the research,
authorship, and/or publication of this article: Caterina Marchionnis work was con-
ducted with funding from the Academy of Finland.

References
Benard, Stephen, and Robb Willer. 2007. A Wealth and Status-Based Model of
Residential Segregation. Journal of Mathematical Sociology 31 (2): 149-74.
Bruch, Elizabeth, and Robert Mare. 2006. Neighborhood Choice and Neighborhood
Change. American Journal of Sociology 112 (3): 667-709.
Centola, Damon, and Michael Macy. 2007. Complex Contagions and the Weakness
of Long Ties. American Journal of Sociology 113 (3): 702-34.
Centola, Damon, Robb Willer, and Michael Macy. 2005. The Emperors Dilemma: A
Computational Model of Self-Enforcing Norms. American Journal of Sociology
110 (4): 1009-40.
Clark, William A.V., and Mark Fossett. 2008. Understanding the Social Context
of the Schelling Segregation Model. Proceedings of the National Academy of
Sciences of the United States of America 105 (11): 4109-14.
Craver, Carl F. 2007. Explaining the Brain: Mechanisms and the Mosaic Unity of
Neuroscience. Oxford: Clarendon Press.
Epstein, Brian. 2012. Agent-Based Modeling and the Fallacies of Individualism. In
Models, Simulations and Representations, edited by P. Humphreys and C. Imbert,
115-44. New York: Routledge.
Epstein, Joshua M. 2006. Generative Social Science. Studies in Agent-Based
Computational Modeling. Princeton: Princeton University Press.
Flache, Andreas, and Michael W. Macy. 2011. Small Worlds and Cultural
Polarization. Journal of Mathematical Sociology 35 (1): 146-76.
Fossett, Mark. 2006. Ethnic Preferences, Social Distance Dynamics, and Residential
Segregation: Theoretical Explorations Using Simulation Analysis. Journal of
Mathematical Sociology 30 (3-4): 185-273.
Gilbert, Nigel, and Petra Ahrweiler. 2009. The Epistemologies of Social Simulation
Research. In EPOS 2006, LNAI 5466, edited by F. Squazzoni, 12-28. Berlin:
Springer-Verlag.
Grne-Yanoff, Till, and Paul Weirich. 2010. The Philosophy and Epistemology of
Simulation: A Review. Simulation & Gaming 41 (1): 20-50.
Hedstrm, Peter. 2005. Dissecting the Social: On the Principles of Analytical
Sociology. Cambridge: Cambridge University Press.
Hedstrm, Peter, and Petri Ylikoski. 2010. Causal Mechanisms in the Social
Sciences. Annual Review of Sociology 36:49-67.
Marchionni and Ylikoski 339

Hodgson, Geoffrey M. 2007. Meanings of Methodological Individualism. Journal

of Economic Methodology 14 (2): 211-26.
Kincaid, Harold. 1996. Philosophical Foundations of the Social Sciences. Cambridge:
Cambridge University Press.
Kincaid, Harold. 1997. Individualism and the Unity of Science: Essays on Reduction,
Explanation, and the Special Sciences. Lanham: Rowman & Littlefield.
Kirman, Alan P., and Nicolaas J. Vriend. 2001. Evolving Market Structure: An ACE
Model of Price Dispersion and Loyalty. Journal of Economic Dynamics and
Control 25:459-502.
Kuorikoski, Jaakko, Aki Lehtinen, and Caterina Marchionni. 2010. Economic
Modelling As Robustness Analysis. British Journal for the Philosophy of
Science 61 (3): 541-67.
Lehtinen, Aki, and Jaakko Kuorikoski. 2007. Computing the Perfect Model: Why
Do Economists Shun Simulation? Philosophy of Science 74 (3): 304-29.
Levins, Richard. 1966. The Strategy of Model Building in Population Biology.
American Scientist 54 (4): 421-31.
Macy, Michael W., Damon Centola, Andreas Flache, Arnout ven de Rijt, and Robb
Willer. 2011. Social Mechanisms and Generative Explanations: Computational
Models with Double Agents. In Analytical Sociology and Social Mechanisms,
edited by P. Demeulenaere, 250-65. Cambridge: Cambridge University Press.
Macy, Michael W., and Andreas Flache. 2009. Social Dynamics from the Bottom
Up: Agent-Based Models of Social Interaction. In The Oxford Handbook of
Analytical Sociology, edited by P. Hedstrm and P. Bearman, 245-68. Oxford:
Oxford University Press.
Mki, Uskali. 2005. Models Are Experiments, Experiments Are Models. Journal of
Economic Methodology 12 (2): 303-15.
Manzo, Gianluca. 2007. Comment on Andrew Abbott. Sociologica 2:1-8.
Miller, John H., and Scott E. Page. 2007. Complex Adaptive Systems: An Introduction
to Computational Models of Social Life. Princeton: Princeton University Press.
Morgan, Mary S. 2003. Experiments without Material Intervention: Model
Experiments, Virtual Experiments and Virtually Experiments. In The Philosophy
of Scientific Experimentation, edited by Hans Radder, 216-35. Pittsburgh:
University of Pittsburgh Press.
Neumann, Martin. 2008. Homo Socionicus: A Case Study of Simulation Models of
Norms. Journal of Artificial Societies and Social Simulation 11 (4): Article No.
6. http://jasss.soc.surrey.ac.uk/11/4/6.html.
Pettit, Philip. 1993. The Common Mind: An Essay on Psychology, Society, and
Politics. Oxford: Oxford University Press.
Sawyer, R. Keith. 2003. Artificial Societies: Multi Agent Systems and the
Micromacro Link in Sociological Theory. Sociological Methods and Research
31 (3): 325-63. Reprinted in Computational Social Science, edited by N. Gilbert,
2010. Sage.
Sawyer, R.Keith. 2004. The Mechanisms of Emergence. Philosophy of the Social
Sciences 34 (2): 260-82.
340 Philosophy of the Social Sciences 43(3)

Schelling, Thomas C. 1978. Micromotives and Macrobehavior. London: W. W.

Norton.
Udehn, Lars. 2001. Methodological Individualism: Background, History, and
Meaning. London: Routledge.
Weisberg, Michael. 2006. Robustness Analysis. Philosophy of Science 73 (5):
730-42.
Wimsatt, William C. 1981. Robustness, Reliability and Overdetermination. In
Scientific Inquiry and the Social Sciences, edited by M. B. Brewer and B. E.
Collins, 124-63. San Francisco: Jossey-Bass.
Wimsatt, William C. 1983. Reductionistic Research Strategies and Their Biases in
the Units of Selection Controversy. In Conceptual Issues in Ecology, edited by
E. Saarinen, 155-201. Dordrecht: D. Reidel Publishing.
Woodward, James. 2003. Making Things Happen. A Theory of Causal Explanation.
Oxford: Oxford University Press.
Ylikoski, Petri. 2012. Micro, Macro, and Mechanisms. In The Oxford Handbook of
Philosophy of the Social Sciences, edited by H. Kincaid, 21-45. Oxford: Oxford
University Press.
Ylikoski, Petri. Forthcoming. Agent-Based Simulation and Sociological
Understanding. Perspectives on Science.
Ylikoski, Petri, and Jaakko Kuorikoski. 2010. Dissecting Explanatory Power.
Philosophical Studies 148:201-19.

Author Biographies
Caterina Marchionni is an academy research fellow at the Finnish Centre of
Excellence in Philosophy of the Social Sciences, University of Helsinki. Her research
interests are in the philosophy and methodology of economics and the philosophy of
the social sciences.
Petri Ylikoski is a professor of science and technology studies and the deputy direc-
tor of the Finnish Centre of Excellence in Philosophy of the Social Sciences,
University of Helsinki. His research interests range from theoretical issues in philoso-
phy of science to empirical case studies in science and technology studies.