Domain of agent-based modeling systems

Agent-based modeling is an intriguing new set of tools for computational social science. The techniques permit us to project forward the system-level effects of a set of assumptions about agent behavior and a given environment. What kinds of real social phenomena are amenable to treatment by the techniques of agent-based modeling? David O'Sullivan and his co-authors offer an assessment of this question in their contribution to a valuable recent handbook, Heppenstall et al, Agent-Based Models of Geographical Systems. (Andrew Crooks and Alison Heppenstall provide a valuable and clear introduction to ABM methodology in their contribution to the volume.)

O'Sullivan and colleagues offer a basic taxonomy of different applications of ABM research.

• simple abstract models where the focus is on exploring the collective implications of individual-level decision making. 
• more detailed [accounts that] locate virtual model agents in a representation of the real world setting of interest. Typically, such models operate at a regional or landscape scale
• some of the most ambitious models aim at detailed ... representations of both the geographical setting and the processes unfolding in that setting (111-112)

This taxonomy depends on the degree of abstraction and realism that the model aspires to.

Here are a handful of research projects that are amenable to these techniques, most of which are illustrated in the Heppenstall volume.

• Land use patterns in peasant agriculture 
• Residential patterns -- urban and rural
• Patterns of burglaries
• Occurrence of interpersonal violence in civil war
• Traffic patterns -- pedestrian and vehicular

What do the clear examples have in common? They are situations where a number of independent individuals react to a social and natural environment with a set of goals; and they are usually situations where individuals influence each other through their actions. These are situations of dynamic interactive choices. O'Sullivan and colleagues put these points this way:

We consider the most fundamental characteristics of agents in spatial models to be goal-direction and autonomy.... However, more specific definitions of the concept may add any of flexibility, ‘intelligence', communication, learning, adaptation or a host of other features to these two. (115)

(Crooks and Heppenstall provide a similar list: autonomy, heterogeneity, and activity; 87.)

O'Sullivan et al also pose an important question about what the circumstances are where the features of agents makes a difference in the social outcome:

This argument focuses attention on three model features: heterogeneity of the decision-making context of agents, the importance of interaction effects , and the overall size and organization of the system. If agents are the same throughout the system, then, other things being equal, an aggregate approach is likely to capture the same signifi cant features of the system as an agent-based approach.

Essentially the point here is a simple one: if an aggregate outcome results from homogenous individuals making a decision about something on the same basis as everyone else, then we don't need an agent-based model. ABM techniques become valuable when heterogeneous agents interact with each other to bring about novel outcomes.

There are quite a few social situations that do not fit the terms of these models well. Some social processes are not simply the aggregate outcome of choices by a set of independent autonomous agents. For example, the flow of work through an architectural design studio is determined by the rules of the firm, not the independent choices of the employees, and the behavior of an army is largely determined by its general staff and command structure. O'Sullivan et al put the point this way:

A more important question may be, “what should the agents in an ABM of this system represent?” If the interactions among individual actors in the real world are substantially channelled via institutions or other social or spatial structures, perhaps it is those institutions or social or spatial structures that should be represented as agents in an ABM rather than the individuals of which they are formed. (120)

So a general question for ABM methodology is this: where do structural social factors come into ABM models? Here I am thinking of things like a system of regulation and law; a pattern of racialized behavior; the architecture of the transport system; a tax system; .... We might treat these as parameters in the environment of choice for the agents. They are beyond the control of the agents and are regarded as constraints and opportunities. (This is one place where the framework of "strategic interactive fields" disagrees, since the SIF approach looks at institutions themselves as part of the field of strategic interaction in that individuals strive to modify the rules to their own benefit.)

It seems reasonable to judge that ABM techniques are very useful when we are concerned with phenomena that are aggregates of strategic behavior by individual actors; but they are not pertinent to many of the questions sociologists pose. In particular, they do not seem useful for sociological inquiries that are primarily concerned with the dynamics and effects of large social structures where the behavior of individuals is routine, homogeneous, or largely determined exogenously. These are the circumstances where the premises of the ABM approach -- autonomy, heterogeneity, and activity -- are not satisfied.

 

Strategic action fields

Sometimes a rethinking of ontology and social categories results in an important step forward in social theory. This appears to be the case in some recent reflections on the relationships that exist between social movements theory and the sociology of organizations.  The presumption of existing writings on these fields is that they refer to separate but related phenomena.  One is more about social actors and the other is more about stable social structures.  What happens when we consider the possibility that they actually refer to the same kinds of social phenomena?

This is the perspective taken by Neil Fligstein and Doug McAdam in a recent contribution to Sociological Theory, "Toward a General Theory of Strategic Action Fields"(link). (They develop these ideas more fully in A Theory of Fields.) In the Sociological Theory article they write:

We assert that scholars of organizations and social movements -- and for that matter, students of any institutional actor in modern society -- are interested in the same underlying phenomenon: collective strategic action. (2)

Fligstein and McAdam formulate their novel approach in terms of the idea of "strategic action fields." They put it forward that "strategic action fields … are the fundamental units of collective action in society" (3). Power and advantage play key roles in their construction: "We too see SAFs as socially constructed arenas within which actors with varying resource endowments vie for advantage. Membership in these fields is based far more on subjective 'standing' than objective criteria" (3).
Here are types of social items they include in this theory:

1. strategic action fields 
2. incumbents, challengers, and governance units 
3. social skill 
4. the broader field environment 
5. exogenous shocks, field ruptures, and the onset of contention 
6. episodes of contention 
7. settlement (2) 

This approach is importantly couched at the level of social ontology: what sorts of things should we identify and analyze as explanatory factors in our theories? The move to SAFs is a move against the idea of the fixity of social "structures," institutions, and organizations. For example, they write against the ontology of new institutionalism: "The general image for most new institutionalists is one of routine social order and reproduction" -- or in other words, a static set of rules and constraints within which action takes place. Their ontology, on the other hand, emphasizes the fluidity of the constraints and circumstances of action from the actors' points of view; so the field shifts as actors undertake one set of strategies or another. "This leaves great latitude for the possibility of piecemeal change in the positions that actors occupy" (5).

So both stability and change are incorporated into a single framework of analysis: actors react strategically to the field of constraints and positions within which they act, with results that sometimes reinforce current positions and other times disrupt those positions.

They account for what looks like institutional rigidity by calling out the power of some actors to maintain their positions in the social order: "Most incumbents are generally well positioned and fortified to withstand these change pressures. For starters they typically enjoy significant resource advantages over field challengers" (9). But institutions should not be expected to maintain their structures indefinitely: "The expectation is that when even a single member of the field begins to act in innovative ways in violation of field rules, others will respond in kind, precipitating an episode of contention" (9).

So what is intended by the idea of "strategic action" in this theory? Here is what they have to say on that subject:

We define strategic action as the attempt by social actors to create and maintain stable social worlds by securing the cooperation of others. Strategic action is about control in a given context. The creation of identities, political coalitions, and interests serves to promote the control of actors vis-a-vis other actors. (7)

Here is one other interesting ontological feature of this approach. Their language suggests some parallels with assemblage theory (link), in the sense that social constructs fit upwards and downwards into strategic action fields at a range of fields. "We conceive of all fields as embedded in complex webs of other fields" (8). This set of ideas seems to suggest an unexpected affinity to "actor-network theory" and the sociological ideas of Bruno Latour (ANT) (link). But at the other end of some obscure spectrum of theory differentiation, their account also seems to rub shoulders with rational-choice theory, where both actions and rules are subject to deliberation and change by prudential actors.

There are several features of this approach that seem promising to me. One is the fact that it directly challenges the tendency towards reification that sometimes blocks sociological thinking -- the idea that social "things" like states persist largely independently from the individuals who make them up. This new approach leads to a way of thinking about the social world that emphasizes contingency and plasticity (link, link) rather than rigid and homogeneous social structures. It also seems consistent with the thinking that leads to the idea of "methodological localism" -- the idea that social phenomena rest upon "molecules" of socially constructed, socially situated individuals (link). I also like the fact that their analysis is explicitly couched at the meso level -- neither macro nor micro.

One concern this approach raises, however, is suggested by the point mentioned above about its apparent proximity to some versions of rational choice theory -- the view that all social outcomes and processes are ultimately the consequence of prudential actors pursuing their interests. But this assumption -- which McAdam certainly does not share elsewhere in his writing (e.g. Dynamics of Contention) -- threatens to push out of consideration social realities like normative systems, social identities, and distributed systems of power that somehow or other seem to demand inclusion in our understanding of social processes.

Finally, we can ask whether this innovation provides a basis for more fruitful empirical research into concrete phenomena like how corporations and revolutionary parties function, how demonstrations against Islamophobia take shape, and how resistance to racial discrimination emerges.  If the theoretical innovation doesn't lead to richer empirical research, then it is reasonable to be skeptical about why we should adopt the new theoretical tools.

Actor-centered sociology and agent-based models

Actor-centered sociology (ACS) begins in the intuition that social processes begin in the interactions of socially constructed individuals, and it takes seriously the idea that actors have complex and socially inflected mental schemes of action and representation. So actor-centered sociologists are keen not to over-simplify the persons who constitute the social domain of interest. And this means that they are generally not content with sparse abstract schemata of actors like those proposed by most versions of rational choice theory.

Agent-based modeling (ABM) is a collection of aggregative techniques aimed at working out the aggregate consequences of the hypothetical choices of a number of individuals interacting in a series of social environments. ABM models generally represent the actors' motivations and decision rules very abstractly -- sometimes as economic actors, sometimes as local optimizers, sometimes as heuristically driven decision makers. An ABM model may postulate several groups of actors whose decision rules are different -- predators and prey, landlords and tenants, bandits and generals. The goal is to embody a set of behavioral assumptions at the actor level and then to aggregate the results of the actions and interactions of these actors at a macro level. (Stephen Railsback's Agent-Based and Individual-Based Modeling: A Practical Introduction provides an accessible introduction.)

My question here is a focused one: do these apparently similar approaches to explaining social outcomes actually have as much in common as they appear to at first glance? And the answer I'll suggest is -- not yet, and not enough. (Here are earlier discussions of the two frameworks; link, link.)

The sticking point between them is the issue of abstraction and granularity concerning the nature of the actors. ACS researchers are critical of the methodological move towards abstraction in the description of the actor. They believe that the socially embedded and rather specific features of deliberation and action that they investigate in various historical and cultural settings are crucial and are lost when we move to a more abstract desire-belief-opportunity (DBO) approach. ABM theorists argue that abstraction about the agent is necessary if a social situation is supposed to be tractable -- to model the behavior of a population of agents we need to be able to represent their decision rules in a reasonably compact and mathematically representable way. So if we take the view that each individual is a unique bundle of mental frameworks and action-practices, we will have to give up the enterprise of modeling their collective behavior.

However, some efforts to apply ABM techniques to real contemporary and historical problems -- for example, land use patterns in contemporary African agriculture -- have had disappointing results. The patterns predicted by the simulation diverge fairly significantly from the observed patterns on the ground. And some of these researchers believe that the problem lies not with the model but with the assumptions made about the actor. Those assumptions are basically Chicago-style rational choice assumptions, and researchers are coming to see that the actors -- farmers, herders, traders -- are operating on the basis of rules that are more nuanced. The actors are prudential and they make deliberative choices; but their reasoning doesn't reduce to an application of expected-utility calculation. So the researchers themselves are asking whether it would be better to incorporate more realistic assumptions about actors' motivations and reasoning frameworks.

This suggests that there is perhaps more fertile ground between the ACS and ABM frameworks than has yet been exploited. ACS focuses its attention on the question of refining our understanding of how actors are constituted, and ABM provides a rich set of techniques for transporting from assumptions about individual actors to the simulated result of aggregating these actors' behaviors onto a collective pattern.

The hybrid approach still requires abstraction about actors. But perhaps it is worth considering adjusting the focus, from "farmers in an environment" to "Kenyan farmers with X, Y, Z features of goals and reasoning schemes". Perhaps the disaggregation of types of actors needs to go even further. And perhaps the question of "what kinds of actors are involved in land use in Kenyan agriculture?" needs to be driven by empirical investigation rather than methodological fiat or computational convenience.

So maybe the great centers for complexity studies around the country would be well advised to begin including anthropologists and cultural sociologists within their research teams. And maybe the result will be a fertile marriage of modeling with greater cultural specificity. 

Elder-Vass on social realism

Dave Elder-Vass's arguments for the real causal powers of social structures have been considered here several times (link, link).  Elder-Vass's recent book, The Reality of Social Construction, addresses this subject from a different point of view.  Here he is interested in the question of the collision of social realism and social constructivism, generally thought of as being incompatible perspectives on the nature of the social world.  E-V does not believe they are in fact incompatible, and the thrust of the current book is to make the case for this point of view.

This book, however, develops and substantiates the critical realist argument that social scientists should be both realists and social constructionists. (3)

Here are some of the ideas he offers as aspects of social constructionism:

If there is one claim that is definitive of social constructionism, it is the argument that the ways in which we collectively think and communicate about the world affect the way that the world is. (4) 

Social constructionisms derive their force from a further claim: that changing the ways in which people collectively think and/or communicate about the world in itself constitutes a change with significance for the social world. (5) 

Radical constructionists tend to deny any such distinction [between what depends upon how we think about it and what does not], on the grounds that everything depends on the ways in which we think about it, or at least to include in the socially constructed category things that realists would not. (6)

And here is social realism:

Realism … may be taken as the belief that there are features of the world that are the way they are independently of how we think about them. (6)

E-V rejects the exclusionary position, which holds that realism and constructivism are incompatible. Instead, he thinks there is a way of interpreting social ontology that makes the position of realist social constructionism a coherent one.

This book argues for a realist social constructionism -- or, if you prefer, a socially constructionist realism…. I hope this book will encourage more realists to embrace a moderate social constructionism and indeed to recognize that many of them already do so implicitly; that it will encourage social constructionists to recognize the value of realism and their own need for it; and that it will show those with no previous commitment to either tradition that they can be combined fruitfully. (7)

So what are the social items that need to be interpreted both as real features of the social world and as socially constructed? E-V highlights several fundamental kinds of things -- norms, language, meaning, cultural practices, and institutions, for example.

E-V's position requires that we answer two symmetrical questions: How is it that things like these can be thought to be socially constructed? And in what sense are they "real"?  Elder-Vass's most basic answer to the first question is to say that they are socially constructed because they depend unavoidably on intensional representations embodied in what he calls norm circles. Basically, a norm circle is a group of people who interact with each other and who reinforce each other's behavior with respect to one or more social rules of conduct (22). Demonstration of behavior that conforms to a certain norm, and positive or negative feedback to others depending on their conformance or deviance from the norm, creates a situation in which individuals come to internalize these rules of behavior into their own practical rationality.

I argue that a norm circle is an entity with the emergent causal power to increase the dispositions of individuals to conform to the norm endorsed and enforced by the norm circle concerned…. What norm circles produce in individuals is a set of beliefs or dispositions regarding appropriate behavior; the influence of the norm circle, we may say, is mediated through these beliefs or dispositions. (26, 27)

E-V believes that this construct helps to formulate the description of a wide range of social phenomena, including linguistic, cultural, and epistemic social behavior.

So what is "socially constructed" about a norm? And in what sense is there a person-independent social reality to a norm? The social construction part of the story seems straightforward. In order to have a normative expectation about a certain kind of behavior in a certain social context, it is necessary first to have a cognitive frame or representation for the behavior.  This is an intensional attitude on the part of the actor. To know how to behave when one is introduced to the Queen of England, we need to have a set of beliefs about royalty, monarchy, social roles, and particular persons.  Without mental frameworks involving these sorts of things, we cannot entertain the notion of a norm governing behavior in such a circumstance.  The situation of "being introduced to the Queen of England" is dependent on our conceptual system.

Having said this, it is also open to us to notice the relative stability and permanence of the patterns of behavior that surround this situation.  Most people observe the correct protocol, and those who do not are admonished by others for their breaches.  So "protocols of behavior surrounding introduction to the Queen" functions as a social reality that is independent from the individual.  The radical egalitarian who regards the concept of royalty as delusion and self-deception, is no less governed by the norm. So there is a crucial component of actor-independence that is possessed by the normative system as embodied in the norm circle. Here is how E-V summarizes this point:

Rules and norms, therefore, may still feature in our causal accounts of culture, but not as entities with causal powers, not as ideas that exist externally to the individual actors concerned, and not as beliefs that are completely and precisely homogenized across the norm circles concerned. Since it is not norms themselves but the norm circles that endorse and enforce them that are the bearers of the causal powers concerned, none of these constraints undermines the causal account of normatively outlined in this chapter and the previous one…. Culture, it has argued, is produced by norm circles, and indeed culture and normatively are one and the same. (53, 54)

So norm circles play a crucial role in E-V's social ontology.  If we were to distill the idea down to its simplest form, it seems to go along these lines: individuals have the capacity to form ideas, rules, and representations of various kinds. They reinforce their ideas and beliefs through interactions with other individuals who (approximately) share those mental representations. This is what makes a given norm system or conceptual framework a social feature rather than simply an individual feature. The representations are constantly tuned through interactions with other members of this representation-sharing group. These representations include ideas, conceptual frameworks, beliefs, and norms.  The groups of people who share these and interact on the basis of them constitute a norm circle.

This formulation brings Elder-Vass's view into parallel with those of Margaret Archer and her concept of "morphogenesis" (link). In thinking about the reality of social structures, E-V writes the following:

Over a period of time, individuals may act in ways that tend to reproduce the structure more or less unchanged, or they may act in ways that tend to transform it. In analysing how these structures work and how they develop, we must take account not only of the collective power that they have but also of the ways in which individual participation in them jointly produces and influences the collective outcome. (254)

And this in turn means that E-V is in a position to deny two distinct polarities: between social construction and realism, and between agent and structure.

Social embeddedness

To what extent do individuals choose their courses of action largely on the basis of a calculation of costs and benefits? And to what extent, on the contrary, are their actions importantly driven by the normative assumptions they share with other individuals with whom they interact? Mark Granovetter formulated this foundational question for the social sciences in his important 1985 contribution to the American Journal of Sociology, "Economic Action and Social Structure: The Problem of Embeddedness" (link). He used the concept of embeddedness as a way of capturing the idea that the actions individuals choose are importantly refracted by the social relations within which they function. This is a topic we've addressed frequently in prior posts under the topic of the social actor, and Granovetter's contribution is an important one to consider as we try to further clarify the issues involved.

The large distinction at issue here is the contrast between rational actor models of the social world, in which the actor makes choices within a thin set of context-independent decision rules, and social actor models, in which the actor is largely driven by a context-defined set of scripts as he/she makes choices. The contrast is sometimes illustrated by contrasting neoclassical economic models of the market with substantivist models along the lines of Karl Polanyi's The Great Transformation: The Political and Economic Origins of Our Time, and it links to the debate in economic anthropology between formalists and substantivists. Here is how Granovetter puts the fundamental question:

How behavior and institutions are affected by social relations is one of the classic questions of social theory. (481)

He argues that neither of the polar positions are tenable.  The formalist approach errs in taking too a-social view of the actor:

Classical and neoclassical economics operates, in contrast, with an atomized, undersocialized conception of human action, continuing in the utilitarian tradition. ... In classical and neoclassical economics, therefore, the fact that actors may have social relations with one another has been treated, if at all, as a frictional drag that impedes competitive markets. (483, 484)

But the extreme alternative isn't appealing either:

More recent comments by economists on "social influences" construe these as processes in which actors acquire customs, habits, or norms that are followed mechanically and automatically, irrespective of their bearing on rational choice. (485)

So action doesn't reduce to abstract optimizing rationality, and it doesn't reduce to inflexible cultural or normative scripts either. Instead, Granovetter proposes an approach to this topic that reframes the issue around a more fluid and relational conception of the actor. Like the pragmatist theories of the actor discussed in earlier posts (Abbott, Gross, Joas), he explores the idea that the actor's choices emerge from a flow of interactions and shifting relations with others. The actor is not an atomized agent, but rather a participant in a flow of actions and interactions.

At the same time, Granovetter insists that this approach does not deny purposiveness and agency to the actor. The actor reacts and responds to the social relations surrounding him or her; but actions are constructed and refracted through the consciousness, beliefs, and purposes of the individual.

The idea of embeddedness is crucial for Granovetter's argument; but it isn't explicitly defined in this piece.  The idea of an "embedded" individual is contrasted to the idea of an atomized actor; this implies that the individual's choices and actions are generated, in part anyway, by the actions and expected behavior of other actors.  It is a relational concept; the embedded actor exists in a set of relationships with other actors whose choices affect his or her own choices as well.  And this in turn implies that the choices actors make are not wholly determined by facts internal to their spheres of individual deliberation and beliefs; instead, actions are importantly influenced by the observed and expected behavior of others.

Their attempts at purposive action are instead embedded in concrete, ongoing systems of social relations. (487)

Some of Granovetter's discussion crystallizes around the social reality of trust within a system of economic actors. Trust is an inherently relational social category; it depends upon the past and present actions and interactions within a group of actors, on the basis of which the actors choose courses of action that depend on expectations about the future cooperative actions of the other actors. Trust for Granovetter is therefore a feature of social relations and social networks:

The embeddedness argument stresses instead the role of concrete personal relations and structures (or "networks") of such relations in generating trust and discouraging malfeasance. (490)

And trust is relevant to cooperation in all its variants -- benevolent and malicious as well. As Granovetter points out, a conspiracy to defraud a business requires a group of trusting confederates. So it is an important sociological question to investigate how those bonds of trust among thieves are created and sustained.

This line of thought, and the theory of the actor that it suggests, is an important contribution to how we can understand social behavior in a wide range of contexts. The key premise is that individuals choose their actions in consideration of the likely choices of others, and this means that their concrete social relations are critical to their actions. How frequently do a set of actors interact? Has there been a history of successful cooperation among these actors in the past? Are there rivalries among the actors that might work to reduce trust? These are all situational and historical facts about the location and social relations of the individual. And they imply that very similar individuals, confronting very similar circumstances of choice, may arrive at very different patterns of social action dependent on their histories of interaction with each other.

It seems that this theory of the actor would be amenable to empirical investigation.  The methodologies of experimental economics could be adapted to study of the relational intelligence that Granovetter describes here. Recent works by Ernst Fehr and Klaus Schmidt explore related empirical questions about decision making in the context of problems involving fairness and reciprocity (Foundations of Human Sociality: Economic Experiments and Ethnographic Evidence from Fifteen Small-Scale Societies and "The Economics of Fairness, Reciprocity and Altruism - Experimental Evidence and New Theories"; link).

(These topics have come up in earlier discussions here. Here is a post on Chuck Tilly's treatment of trust networks; link. Amartya Sen's discussion of "rational fools" is relevant as well, as is his account of the role that commitments play in action (link). It seems likely that Granovetter would argue that Sen's solution is still too formalist, in that it attempts to internalize he social relations component into the actor's calculations. This is true of the "identity economics" approach as well; (link).)

 

Path dependency and contingency

Historical outcomes are rarely determined by some set of antecedent conditions. Instead, the outcome that occurs is the result of an extended series of events and conditions, each stage of which is subject to substantial contingency and conjunctures. This is most visible in the occurrence of large, dramatic events like stock market crashes, peasant uprisings, and wars. But it is also true in more mundane occurrences -- scandals, resignations, and insolvencies, for example.

The idea of path dependency is chiefly relevant in application to particular events and happenings. But there are large social structures that display path dependency as well. These examples amount to structuring conditions that persist for a long time and further condition other historical developments. Their emergence is contingent, and other directions were feasible in the early stage. But once the structure is in place it creates its own conditions for persistence. Because the Interstate highway program privileged automobile transport over rail in the 1950s, it is now difficult to create a financially viable passenger rail system in the United States.

Examples of historical developments that display significant path dependency include the choice of major technologies (internal combustion engines versus electric vehicles, alternating current rather than direct current as the long-distance power transmission standard, the QWERTY keyboard versus other more efficient arrangements), major infrastructures (rail rather than canal in the late nineteenth century, personal automobile rather than passenger rail in the mid-twentieth century), and the division of tasks and specialization across adjacent professions (social workers, psychiatrists). In each of these cases there is a contingent first step that could have led in another direction, and then an accumulated resistance to reverting to the other choice (substantial social investment already committed to the first option, substantial development of the knowledge systems required for the first option but not the second, substantial social power invested in individuals and organizations that have a vested interest in maintaining the first option, etc.). (Tom Hughes refers to these factors as "technological momentum" in Networks of Power: Electrification in Western Society, 1880-1930.) It isn't impossible to retrace steps and adopt the alternative solution -- but it is very difficult and uncommon.

James Mahoney has done useful work in explicating this concept in "Path dependence in historical sociology" (link). Here is how he describes his own position:

In this article, I argue that path dependence characterizes specifically those historical sequences in which contingent events set into motion institutional patterns or event chains that have deterministic properties. The identification of path dependence therefore involves both tracing a given outcome back to a particular set of historical events, and showing how these events are themselves contingent occurrences that cannot be explained on the basis of prior historical conditions. (507-508)

Mahoney identifies two basic mechanisms that are associated with path dependency:

First, some path-dependent investigators analyze self-reinforcing sequences characterized by the formation and long-term reproduction of a given institutional pattern. (508)

A second basic type of path-dependent analysis involves the study of reactive sequences. Reactive sequences are chains of temporally ordered and causally connected event. (509)

Here are the three features Mahoney finds to be critical.

I suggest that all path-dependent analyses minimally have three defining features. First, path-dependent analysis involves the study of causal processes that are highly sensitive to events that take place in the early stages of an overall historical sequence. (510)

Second, in a path-dependent sequence, early historical events are contingent occurrences that cannot be explained on the basis of prior events or "initial conditions." (511)

Third, once contingent historical events take place, path-dependent sequences are marked by relatively deterministic causal patterns or what can be thought of as "inertia" - i.e., once processes are set into motion and begin tracking a particular outcome, these processes tend to stay in motion and continue to track this outcome. (511)

Mahoney's work here represents a great example of careful, detailed analysis that crosses the categories of methodology, theory, and philosophy. This is how we make progress in the social sciences -- by taking theories and concepts seriously and providing careful, logical analysis of the ways these concepts express important features of the historical process.

The idea of path dependency invokes the intersection of two important intuitions which are in tension with each other. The first is the sense of contingency that study of important historical transitions instills. The second is the idea that explanation involves generalizing across cases. If a case is wholly particular, then it is hard to find anything explanatory in studying it. Careful analysis of the concept of path dependency has the potential of offering a higher-level resolution of these intuitions about the historical process.

Problems with causal mechanisms

There are a couple of problems with the theory of causal mechanisms that will be difficult to address. Jim Mahoney raises a general concern in "Beyond Correlational Analysis" -- there is no consensus about how to define a mechanism. But there are more specific problems as well.

One is the issue raised by Johannes Persson with respect to Jon Elster's formulation, and the apparent paradox that arises from Elster's epistemic definition: according to Elster's definition, if we learn more about a mechanism, it ceases to be a mechanism (link). This has to do with whether we understand mechanisms epistemically or realistically. In my reply to Persson I argue that we need to take the realist approach (link).

Second, there is the problem that Andrew Abbott raises in his critique of the social mechanisms approach (link), where he highlights the continuity rather than discreteness of social processes. This reflects Abbott's ontological preference for activity rather than actors, processes rather than discrete mechanisms. (I made a point somewhat similar to this in a post about the "discreteness" of causal mechanisms (link). I argued we should not look at mechanisms as atoms or molecules of causation. There I was inclined to push the action down to the level of the actor.)

A third problem may be even more serious. This derives from the condition of "regularity" that many definitions of a causal mechanism invoke. "A produces B" constitutes a CM if the real properties of A work to bring about B and if A is regularly associated with B. (I refer to these as pocket-sized regularities.) The basic idea here is that once the mechanism is triggered, it leads more or less inexorably to its result.

The first part concerns the idea of real causal efficacy. Here is Rom Harre, excerpted by Mahoney:

The structures, states and inner constitutions from which the phenomena of nature flow.... the permanent or enduring conditions under which a certain kind of phenomenon will occur." "The inner constitutions, structures, powers, encompassing systems, and so on, of which natural generative mechanisms are constituted, and of which the connection between cause and effect usually consists. Harre (1970, pp. 101,102, 104)

The second criterion expresses the idea that trigger and outcome are linked by a strong regularity or law. Here is Stuart Glennan:

A mechanism underlying a behavior is a complex system which produces that behavior by the interaction of a number of parts according to direct causal laws. (Glennan, "Mechanisms and the Nature of Causation," Erkenntnis 1996)

The problem is that these two criteria are likely to pull apart in many circumstances. There may be a real causal push from A to B, but it may result in a relatively low incidence of A followed by B. In this circumstance it is true that A is part of the cause of B, but it is not true that A->B represents a causal mechanism. And in fact we do refer to mechanisms where the likelihood of the consequent is low even when the antecedent occurs -- for example, "the mechanism of George's becoming ill is his close exposure to Alice who had the contagious illness." It may be that the likelihood of catching the flu is still low following exposure; but there is no other way of getting the flu.

This would entail that it is not the case that all causal linkages are transmitted by causal mechanisms, and we would be forced to refer to causal powers or processes instead. Further, the question of whether a given linkage represents a mechanism or not depends only on the relatively boring facts about the world that push up or down the probability of the linkage -- not some fundamental ontological distinction between mechanisms and other causal connections.

This seems to suggest that we have a difficult choice. Either we can drop the "regularity" criterion or we can acknowledge that "causal mechanism" does not capture all reasonably direct causal connections. If we take the second route, we are ill advised to define causation in terms of causal mechanisms per se; instead, we need a more comprehensive concept of "real causal structures and processes", of which a subset are causal mechanisms.

It may be that considerations surrounding the third issue are impossible to resolve without dropping the requirement that a mechanism corresponds to some sort of regularity.

Another possibility is to weaken the regularity criterion so that it encompasses causal relevance rather than high probability of succession. The causal relevance test offered by Wesley Salmon would handle this approach for a mechanism like "contagion through exposure" along these lines: illness | exposure > illness | ~ exposure. Would this suffice as a corrective to the current definition? Might we reframe the definition along these lines:

A => B represents a causal mechanism if

(i) the real properties of A produce B

(ii) B | A > B | ~A

(Gambetta takes something like this approach.)

But if we take this approach, we lose the ability to infer from cause to effect through the mechanism. In alternative language, we move from thinking of mechanisms as sufficient causes to necessary conditions (or rather their probabilistic equivalents).