Thinking about pandemic models

One thing that is clear from the pandemic crisis that is shaking the world is the crucial need we have for models that allow us to estimate the future behavior of the epidemic. The dynamics of the spread of an epidemic are simply not amenable to intuitive estimation. So it is critical to have computational models that permit us to project the near- and middle-term behavior of the disease, based on available data and assumptions.

Scott Page is a complexity scientist at the University of Michigan who has written extensively on the uses and interpretation of computational models in the social sciences. His book, The Model Thinker: What You Need to Know to Make Data Work for You, does a superlative job of introducing the reader to a wide range of models. One of his key recommendations is that we should consider many models when we are trying to understand a particular kind of phenomenon. (Here is an earlier discussion of the book; link.) Page contributed a very useful article to the Washington Post this week that sheds light on the several kinds of pandemic models that are currently being used to understand and predict the course of the pandemic at global, national, and regional levels ("Which pandemic model should you trust?"; (link). Page describes the logic of "curve-fitting" models like the Institute for Health Metrics and Evaluation (IHME) model as well as epidemiological models that proceed on the basis of assumptions about the causal and social processes through which disease spreads. The latter attempt to represent the process of infection from infected person to susceptible person to recovered person. (Page refers to these as "microfoundational" models.) Page points out that all models involve a range of probable error and missing data, and it is crucial to make use of a range of different models in order to lay a foundation for sound public health policies. Here are his summary thoughts:

All this doesn’t mean that we should stop using models, but that we should use many of them. We can continue to improve curve-fitting and microfoundation models and combine them into hybrids, which will improve not just predictions, but also our understanding of how the virus spreads, hopefully informing policy. 

Even better, we should bring different kinds of models together into an “ensemble.” Different models have different strengths. Curve-fitting models reveal patterns; “parameter estimation” models reveal aggregate changes in key indicators such as the average number of people infected by a contagious individual; mathematical models uncover processes; and agent-based models can capture differences in peoples’ networks and behaviors that affect the spread of diseases. Policies should not be based on any single model — even the one that’s been most accurate to date. As I argue in my recent book, they should instead be guided by many-model thinking — a deep engagement with a variety of models to capture the different aspects of a complex reality. (link)

Page's description of the workings of these models is very helpful for anyone who wants to have a better understanding of the way a pandemic evolves. Page has also developed a valuable series of videos that go into greater detail about the computational architecture of these various types of models (link). These videos are very clear and eminently worth viewing if you want to understand epidemiological modeling better.

Social network analysis is crucial to addressing the challenge of how to restart businesses and other social organizations. Page has created "A Leader's Toolkit For Reopening: Twenty Strategies to Reopen and Reimagine", a valuable set of network tools and strategies offering concrete advice about steps to take in restarting businesses safely and productively. Visit this site to see how tools of network analysis can help make us safer and healthier in the workplace (link). 

Another useful recent resource on the logic of pandemic models is Jonathan Fuller's recent article "Models vs. evidence" in Boston Review (link). Fuller is a philosopher of science who undertakes two tasks in this piece: first, how can we use evidence to evaluate alternative models? And second, what accounts for the disagreements that exist in the academic literature over the validity of several classes of models? Fuller has in mind essentially the same distinction as Page does, between curve-fitting and microfoundational models. Fuller characterizes the former as "clinical epidemiological models" and the latter as "infectious disease epidemiological models", and he argues that the two research communities have very different ideas about what constitutes appropriate use of empirical evidence in evaluating a model. Essentially Fuller believes that the two approaches embody two different philosophies of science with regard to computational models of epidemics, one more strictly empirical and the other more amenable to a combination of theory and evidence in developing and evaluating the model. The article provides a level of detail that would make it ideal for a case study in a course on the philosophy of social science.

Joshua Epstein, author of Generative Social Science: Studies in Agent-Based Computational Modeling, gave a brief description in 2009 of the application of agent-based models to pandemics in "Modelling to Contain Pandemics" (link). Epstein describes a massive ABM model of a global pandemic, the Global-Scale Agent Model (GSAM), that attempted to model the spread of the H1N1 virus in 1996. Here is a video in which Miles Parker explains and demonstrates the model (link). 

Another useful resource is this video on "Network Theory: Network Diffusion & Contagion" (link), which provides greater detail about how the structure of social networks influences the spread of an infectious disease (or ideas, attitudes, or rumors).

My own predilections in the philosophy of science lean towards scientific realism and the importance of identifying underlying causal mechanisms. This leaves me more persuaded by the microfoundational / infectious disease models than the curve-fitting models. The criticisms that Nancy Cartwright and Jeremy Hardie offer in Evidence-Based Policy: A Practical Guide to Doing It Better of the uncritical methodology of randomized controlled trials (link) seem relevant here as well. The IHME model is calibrated against data from Wuhan and more recently northern Italy; but circumstances were very different in each of those locales, making it questionable that the same inflection points will show up in New York or California. As Cartwright and Hardie put the point, "The fact that causal principles can differ from locale to locale means that you cannot read off that a policy will work here from even very solid evidence that it worked somewhere else" (23). But, as Page emphasizes, it is valuable to have multiple models working from different assumptions when we are attempting to understand a phenomenon as complex as epidemic spread. Fuller makes much the same point in his article:

Just as we should embrace both models and evidence, we should welcome both of epidemiology’s competing philosophies. This may sound like a boring conclusion, but in the coronavirus pandemic there is no glory, and there are no winners. Cooperation in society should be matched by cooperation across disciplinary divides. The normal process of scientific scrutiny and peer review has given way to a fast track from research offices to media headlines and policy panels. Yet the need for criticism from diverse minds remains.

System effects

Quite a few posts here have focused on the question of emergence in social ontology, the idea that there are causal processes and powers at work at the level of social entities that do not correspond to similar properties at the individual level. Here I want to raise a related question, the notion that an important aspect of the workings of the social world derives from "system effects" of the organizations and institutions through which social life transpires. A system accident or effect is one that derives importantly from the organization and configuration of the system itself, rather than the specific properties of the units.

What are some examples of system effects? Consider these phenomena:

• Flash crashes in stock markets as a result of automated trading
• Under-reporting of land values in agrarian fiscal regimes 
• Grade inflation in elite universities 
• Increase in product defect frequency following a reduction in inspections 
• Rising frequency of industrial errors at the end of work shifts 

Here is how Nancy Leveson describes systems causation in Engineering a Safer World: Systems Thinking Applied to Safety:

Safety approaches based on systems theory consider accidents as arising from the interactions among system components and usually do not specify single causal variables or factors. Whereas industrial (occupational) safety models and event chain models focus on unsafe acts or conditions, classic system safety models instead look at what went wrong with the system's operation or organization to allow the accident to take place. (KL 977)

Charles Perrow offers a taxonomy of systems as a hierarchy of composition in Normal Accidents: Living with High-Risk Technologies:

Consider a nuclear plant as the system. A part will be the first level -- say a valve. This is the smallest component of the system that is likely to be identified in analyzing an accident. A functionally related collection of parts, as, for example, those that make up the steam generator, will be called a unit, the second level. An array of units, such as the steam generator and the water return system that includes the condensate polishers and associated motors, pumps, and piping, will make up a subsystem, in this case the secondary cooling system. This is the third level. A nuclear plan has around two dozen subsystems under this rough scheme. They all come together in the fourth level, the nuclear plant or system. Beyond this is the environment. (65)

Large socioeconomic systems like capitalism and collectivized socialism have system effects -- chronic patterns of low productivity and corruption in the latter case, a tendency to inequality and immiseration in the former case. In each case the observed effect is the result of embedded features of property and labor in the two systems that result in specific kinds of outcomes. And an important dimension of social analysis is to uncover the ways in which ordinary actors pursuing ordinary goals within the context of the two systems, lead to quite different outcomes at the level of the "mode of production". And these effects do not depend on there being a distinctive kind of actor in each system; in fact, one could interchange the actors and still find the same macro-level outcomes.

Here is a preliminary effort at a definition for this concept in application to social organizations:

A system effect is an outcome that derives from the embedded characteristics of incentive and opportunity within a social arrangement that lead normal actors to engage in activity leading to the hypothesized aggregate effect.

Once we see what the incentive and opportunity structures are, we can readily see why some fraction of actors modify their behavior in ways that lead to the outcome. In this respect the system is the salient causal factor rather than the specific properties of the actors -- change the system properties and you will change the social outcome.

When we refer to system effects we often have unintended consequences in mind -- unintended both by the individual actors and the architects of the organization or practice. But this is not essential; we can also think of examples of organizational arrangements that were deliberately chosen or designed to bring about the given outcome. In particular, a given system effect may be intended by the designer and unintended by the individual actors. But when the outcomes in question are clearly dysfunctional or "catastrophic", it is natural to assume that they are unintended. (This, however, is one of the specific areas of insight that comes out of the new institutionalism: the dysfunctional outcome may be favorable for some sets of actors even as they are unfavorable for the workings of the system as a whole.)

 
Another common assumption about system effects is that they are remarkably stable through changes of actors and efforts to reverse the given outcome. In this sense they are thought to be somewhat beyond the control of the individuals who make up the system. The only promising way of undoing the effect is to change the incentives and opportunities that bring it about. But to the extent that a given configuration has emerged along with supporting mechanisms protecting it from deformation, changing the configuration may be frustratingly difficult.

Safety and its converse are often described as system effects. By this is often meant two things. First, there is the important insight that traditional accident analysis favors "unit failure" at the expense of more systemic factors. And second, there is the idea that accidents and failures often result from "tightly linked" features of systems, both social and technical, in which variation in one component of a system can have unexpected consequences for the operation of other components of the system. Charles Perrow describes the topic of loose and tight coupling in social systems in Normal Accidents; 89 ff,)

New thinking about causal mechanisms

Anyone interested in the topic of causal mechanisms will be interested in the appearance of Stuart Glennan and Phyllis Illari's The Routledge Handbook of Mechanisms and Mechanical Philosophy. Both Glennan and Illari have been significant contributors to the past fifteen years of discussion about the role of mechanisms in scientific explanation, and the Handbook is a highly interesting contribution to the state of the debate.

The book provides discussion of the role of mechanisms thinking in a wide range of scientific disciplines, from physics to biology to social science to engineering and cognitive science. It consists of four large sections: "Historical perspectives on mechanisms", "The nature of mechanisms", "Mechanisms and the philosophy of science", and "Disciplinary perspectives on mechanisms." Each section consists of contributions by talented experts on genuinely interesting topics.

A good introduction to the general topic of mechanisms is the introduction to the volume by Glennan and Illari, and more especially their article, "Varieties of mechanisms." They directly confront one of the large issues in the field, the wide dispersion of definitions and applications of the idea of a causal mechanism. They correctly observe that the concept of mechanism is used fairly differently in various areas of science and philosophy, but they argue that there is a common core of elements that underlie most or all of these usages. The variety that exists is the result of differences in the nature of the phenomena across different areas of scientific investigation, and differences in methodology in use in various sciences. They provide a rather general definition of a mechanism:

A mechanism for a phenomenon consists of entities (or parts) whose activities and interactions are organized so as to be responsible for the phenomenon. (92)

They then attempt to provide a basis for classifying different kinds of mechanisms according to several different criteria. The dimensions of variation they identify include the kind of phenomenon produced, the kind of entities and activities constituting the mechanism, the way in which entities and activities are organized, and the etiology of the mechanism.

Also interesting is Petri Ylikoski's contribution, "Social mechanisms." Ylikoski structures his exposition of the theory of social mechanisms around the Coleman boat diagram (link). To provide a mechanism for a social phenomenon is to provide an account at the level of the actors of how a macro-level event or entity causally brings about another macro-level event or entity. Ylikoski insists that this is a matter of explanatory adequacy rather than reductive analysis, and is therefore not ontologically reductionist. But it does fundamentally imply that social mechanisms occur at the level of interactions among actors. In prior posts I have argued against this presupposition (link). I argue that it is perfectly intelligible to suggest that there are meso-level causal mechanisms. Ylikoski also underlines the affinity that exists between social mechanisms and agent-based modeling: a good ABM demonstrates the process through which a set of conditions at the micro-level aggregate to a certain kind of macro-level outcome. See this earlier post for a small amount of doubt about the adequacy of ABM models to perform this kind of social aggregation for realistic social scenarios; link. (Several of these points are developed in my New Directions in the Philosophy of Social Science.)

Povich and Craver address the topic of the relationship that exists between mechanisms, levels, emergence, and reduction in their contribution, "Mechanistic levels, reduction, and emergence". This is a key question within the philosophy of social science. And the idea of  mechanism seems to have great relevance to the idea of various levels of phenomena. At the level of the organization we see, perhaps, chronic inefficiency in the use of certain kinds of resources. In searching for the mechanisms that cause this inefficiency we may choose to drop down a level and examine the incentives and constraints that guide the behavior of individuals within the organization. And we arrive at a theory of the individual-level mechanism that produces the meso-level outcome. This is a mechanism that falls along strut 3 of Coleman's boat; it is an aggregative mechanism. But not all social mechanisms have this nature. If we want to know why rebellious segments of an agrarian society locate themselves in remote, mountainous areas, it is enough to know a few meso-level facts about the functioning of traditional military forces and the meso-level fact that mountainous terrain gives a tactical advantage to rebel commanders. This appears to be a meso-level mechanism from start to finish.

A particularly intriguing and original contribution is Abrahamsen, Sheredos, and Bechtel's "Explaining visually using mechanism diagrams." We tend to think of scientific explanations as mathematical demonstrations or text-based derivations of outcomes. Abrahamsen, Sheredos, and Bechtel point out that visual diagrams play a crucial role in the presentation of many scientific results; and these diagrams are not merely heuristic or illustrative. A visual presentation serves to designate how the hypothesized mechanism works: what its parts are, how the parts influence each other, and how the functioning of the mechanism over time produces the outcome in question. The authors make an admirable attempt to provide a philosophy-of-science analysis of the components and logic of a visual diagram as an expository device for presenting a causal mechanism or process. They highlight the logical problems of representing entities, spatial location, and temporal duration within a diagram in a way that permits the viewer to gain an accurate understanding of the hypothesized mechanism or process. And they note that it is a conceptually simple step to introduce computational modeling into the graphical representation described here, so the processes in question can step through their interactions on-screen.

Taken together, the essays collected here constitute a valuable contribution to the literature on mechanisms and explanation. The handbook also gives the reader a concrete experience of how deeply varied the mechanisms literature is, leading to very interesting questions about cross-disciplinary communication. It appears to be genuinely challenging to formulate an abstract analysis of the idea of a causal mechanism that will mean approximately the same thing to researchers trained within significantly different research traditions. Unlike many handbooks, this collection warrants reading cover to cover. Researchers who believe that the mechanisms approach provides a valid way of understanding the metaphysics of causal inquiry and explanation will find every article stimulating and helpful.

(Here are a couple of prior posts on the challenge of providing a classification scheme for social mechanisms; link, link.)

Contingency and explanation

Social change and historical events are highly contingent processes, in a specific sense: they are the result of multiple causal influences that "could have been otherwise" and that have conjoined at a particular point in time in bringing about an event of interest. Contrast this situation with what we are looking for when we seek an explanation of a change or event. When we explain an event, we show how and why it was not random or accidental; we identify a set of circumstances that made it necessary or likely in the given circumstances. Contingency and explanation therefore seem to be in tension with each other: a wholly contingent world is perhaps one in which explanation of particular occurrences is impossible.

The appearance of contradiction lessens when we realize that "contingent" is not the same as "random" or "uncaused". (See an earlier post for an effort to disentangle a number of related causal concepts; link.) When a uranium atom decays at a particular moment, this is a truly random event. There is no underlying cause that brought about the decay of the nucleus at this particular moment. When a race riot occurs in in Detroit on July 23, 1967, this was a contingent occurrence -- it did not have to happen; but it was not random, spontaneous, or uncaused. Rather, there were multiple causal factors and processes, along with a number of accidental and spontaneous events, leading to a pathway of social actions that resulted in largescale confrontation, arson, violence. Here is how the Kerner Commission described the occurrence of major race riots in the United States (link):

Disorder did not erupt as a result of a single "triggering" or "precipitating" incident. Instead, it was generated out of an increasingly disturbed social atmosphere, in which typically a series of tension-heightening incidents over a period of weeks or months became linked in the minds of many in the Negro community with a reservoir of underlying grievances. At some point in the mounting tension, a further incident--in itself often routine or trivial--became the breaking point and the tension spilled over into violence.

We can understand this account as depending on a distinction between proximate and distal causes; distal causes (a pattern of police brutality, say) set the stage for racial tension, which makes an outbreak of violence more likely; and a precipitating (proximate) event triggers the outburst. The point in this paragraph is that the triggering cause is not the sole cause, or even the most important cause. But all these factors are causally relevant to the outcome. We say that the riot was contingent because there are many ways in which the tensions created by the background conditions could have been defused -- a progressive mayor could have enacted a police reform along with a jobs program, a charismatic leader like Dr. King could have emerged in Detroit who helped to channel tension into electoral politics rather than an outbreak of violence, the Federal government could have been more successful in its civil rights reforms and its War on Poverty. Or the raid on the blind pig could have happened in a driving rainstorm, with the result that no crowd gathered. So the outcome was not preordained. It was contingent, but it was caused.

So it is not the case that a contingent world is one in which nothing can be explained. A chaotic and random world has that property; but contingency is not chaos. Rather, for many historical and social events we can identify a set of background or standing conditions that elevated the probability of the event, we can sometimes identify independent causal processes underway at the same time that interact to further elevate the probability of the event; and we can identify one or more unrelated and random events that served as a trigger to the occurrence of the event of interest.

This is one reason why the strategy of seeking out causal mechanisms in the social world is an appealing approach to social explanation. Appeal to causal mechanisms allows us to make sense of both important features of the social world: that processes and events are contingent, and that many processes and events are amenable to causal explanation.

When researchers set their goals on identifying general causes for groups of social phenomena, they often have in mind the idea that there are similarities in the background standing causal conditions that serve to increase the likelihood of a certain kind of event -- revolution, riot, economic crisis, or period of rapid innovation. And indeed, there are credible hypotheses about such conditions; this is the underlying rationale for the application of Mill's methods to causal reasoning in the social sciences. It is indeed perfectly credible that there are pervasive social conditions that make certain kinds of social events more likely -- a good university system and rapid technological innovation, a defeat in war and political turmoil, the pervasiveness of Protestantism and the hockey stick of market activity.

This insight is closely related to the distinction that Bhaskar and critical realists draw between closed and open systems. In an open system we cannot predict future states of the system because we cannot achieve causal closure; there is always the possibility of another kind of causal influence or mechanism that can offset the workings of the known mechanisms.

The page from the Washington Times above draws attention to an event that was itself highly contingent and yet explicable (the bungled but eventually successful effort to assassinate Archduke Ferdinand); leading to an important historical event (the outbreak of World War I) which was also both contingent and explicable.

Ideologies, policies, and social complexity

The approach to social and historical research that I favor is one that pays attention to the heterogeneity and contingency of social processes. It advises that social and historical researchers should disaggregate the large patterns they start with and try to identify the multiple underlying mechanisms, causes, motivations, movements, and contingencies that came together to create higher-level outcomes. Social research needs to focus on the micro- or meso-level processes that combined to create the macro world that interests us. The theory of assemblages fits this intellectual standpoint very well, since it emphasizes contingency and heterogeneity all the way down. The diagram above was chosen to give a visual impression of the complexity and interconnectedness of factors and causes that are associated with this approach to the social world.

According to the premises of this approach, we are not well served by imagining that there are simple, largescale forces that drive the outcomes in history. Examples of efforts at overly simplified explanations like these include:

• Onerous conditions of the Treaty of Versailles caused the collapse of the Weimar Republic.
• The Chinese Revolution succeeded because of post-Qing exploitation of the peasants.
• The Industrial Revolution occurred in England because of the vitality of English science.

Instead, each of these large outcomes is the result of a large number of underlying processes, motivations, social movements, and contingencies that defy simple summary. To understand the Mediterranean world over the sweep of time, we need the detailed and granular research of a Fernand Braudel rather than the simplified ideas of Johann Heinrich von Thunen in the economic geography of central place theory.

In situations of this degree of underlying complexity, it is pointless to ask for a simple answer to the question, "what caused outcome X?" So the Great Depression wasn't the outcome of capital's search for profits; it was instead the complex product of interacting forms of private business activity, financial institutions, government action, legislation, war, and multiple other forces that conjoined to create a massive and persistent economic depression.

This approach has solid intellectual and ontological foundations. This is pretty much how the social world works. But this ontological vision about the nature of the social world is hard to reconcile with the large intellectual frameworks on the left and on the right that are used to diagnose our times and sometimes to prescribe solutions to the problems identified.

An ideologue is a thinker who seeks to subsume the sweep of history or current events under an overarching narrative with simple explanatory premises and interpretive schemes. The ideologue wants to portray history as the unfolding of a simple set of forces or drivers -- whether markets, classes, divine purposes, or philosophies. And the ideologue is eager to force the facts into the terms of the narrative, and to erase inconvenient facts that appear to conflict with the narrative.

Consider Lenin, von Hayek, and Ronald Reagan. Each had a simplified mental framework that postulated a set of ideas about how the world worked. For Lenin it was expressed in a few paragraphs about class, the economic structure of capitalism, and the direction of history. For von Hayek it was the idea that free economic activity within idealized markets lead to the best possible outcomes for the whole of society. For Reagan it was a combination of von Hayek and the simplified notions of realpolitik associated with Kennan, Morgenthau, or Kissinger.

There are two problems for these kinds of approaches to understanding the social world. First is the indifference ideologues express to the role of facts and empirical validation in their thinking. This is an epistemic shortcoming. But second, and equally problematic, is their insistence on representing the social world as a fundamentally simple process, with a few driving forces whose impact can be forecast. This is an ontological shortcoming. The social world is not simple, and there are not a small number of dominant forces whose effects overshadow the myriad of other socially relevant processes and events that make up a given situation.

Ideologues are insidious for serious historians, since they denigrate careful efforts to discover how various events actually unfolded, in favor of the demands of a particular interpretation of history. It is not possible to gain adequate or insightful historical knowledge from within the framework of a rigid and dogmatic ideology. But even more harmful are policy makers driven by ideologies. An ideological policy maker is an actor who takes the simplistic assumptions of an ideology and attempts to formulate policy interventions based on those assumptions. Ideology-based policies are harmful, of course, because the world has its own properties independent from our theories, and interventions based on false hypotheses about how the world works are unlikely to bring about their intended results. Policies need to be driven by theories that are fact-based and approximately true. And policy makers and officials need to be rejected when they flout science and fact-based inquiry in favor of pet theories and ideologies.

A hard question that this line of thought poses and that I have not addressed here is whether policies can be formulated at all within the context of a fundamentally heterogeneous and contingent world. It might be argued that policy formation requires fairly simple cause-and-effect relationships in order to justify the idea of an intervention; and complexity makes it unlikely that such relationships exist. I believe policies can be formulated within this ontological framework; but I agree that the case must be made. A few earlier posts are relevant to this topic (link, link, link, link, link).

Causal diagrams and causal mechanisms

There is a long history of the use of directed causal diagrams to represent hypotheses about causation. Can the mathematics and graphical systems created for statistical causal modeling be adapted to represent and evaluate hypotheses about causal mechanisms and outcomes?

In the causal modeling literature the structure of a causal hypothesis is something like this: variable T increases/ decreases the probability of the occurrence of outcome E. This is the causal relevance criterion described by Wesley Salmon in Scientific Explanation and the Causal Structure of the World. It is a fundamentally statistical understanding of causality.

Here is a classic causal path model by Blau and Duncan indicating the relationships among a number of causal factors in bringing about an outcome of interest -- "respondent's first job".

This construction aims at joining a qualitative hypothesis about the causal relations among a set of factors with a quantitative measurements of the correlations and conditional probabilities that support these causal relations. The whole construction often takes its origin in a multivariate regression model.

Aage Sørensen describes the underlying methodological premise of quantitative causal research in these terms in his contribution to Frontiers of Sociology (Annals of the International Institute of Sociology Vol. 11):

Understanding the association between observed variables is what most of us believe research is about. However, we rarely worry about the functional form of the relationship. The main reason is that we rarely worry about how we get from our ideas about how change is brought about, or the mechanisms of social processes, to empirical observation. In other words, sociologists rarely model mechanisms explicitly. In the few cases where they do model mechanisms, they are labeled mathematical sociologists, not a very large or important specialty in sociology. (370)

My question here is whether this scheme of representation of causal relationships and the graphical schemes that have developed around it are useful for the analytics of causal mechanisms.

The background metaphysics assumed in the causal modeling literature is Humean and "causal-factor" based; such-and-so factor increases the probability of occurrence of an outcome or an intermediate variable, the simultaneous occurrence of A and B increase the probability of the outcome, etc. Quoting Peter Hedstrom on causal modeling:

In the words of Lazarsfeld (1955: 124-5), "If we have a relationship between x and y; and if for any antecedent test factor the partial relationships between x and y do not disappear, then the original relationship should be called a causal one." (Dissecting the Social: On the Principles of Analytical Sociology)

The current iteration of causal modeling is a directed acyclic graph (DAG). Felix Elwert provides an accessible introduction to directed acyclic graphs in his contribution to Handbook of Causal Analysis for Social Research (link). Here is a short description provided by Elwert:

DAGs are visual representations of qualitative causal assumptions: They encode researchers’ expert knowledge and beliefs about how the world works. Simple rules then map these causal assumptions onto statements about probability distributions: They reveal the structure of associations and independencies that could be observed if the data were generated according to the causal assumptions encoded in the DAG. This translation between causal assumptions and observable associations underlies the two primary uses for DAGs. First, DAGs can be used to prove or disprove the identification of causal effects, that is, the possibility of computing causal effects from observable data. Since identification is always conditional on the validity of the assumed causal model, it is fortunate that the second main use of DAGs is to present those assumptions explicitly and reveal their testable implications, if any. (246)

A DAG can be interpreted as a non-parametric structural equation model, according to Elwert. (Non-parametric here means simply that we do not assume that the data are distributed normally.) Elwert credits the development of the logic of DAGs to Judea Pearl and Peter Spirtes, along with other researchers within the causal modeling community.

Johannes Textor and a team of researchers have implemented DAGitty, a platform for creating and using DAGs in appropriate fields, including especially epidemiology (link). A crucial feature of DAGitty is that it is not solely a graphical program for drawing graphs of possible causal relationships; rather, it embodies an underlying logic which generates expected statistical relationships among variables given the stipulated relationships on the graph. Here is a screenshot from the platform:

The question to consider here is whether there is a relationship between the methodology of causal mechanisms and the causal theory reflected in these causal diagrams. 

It is apparent that the underlying ontological assumptions associated with the two approaches are quite different. Causal mechanisms theory is generally associated with a realist approach to the social world, and generally rejects the Humean theory of causation. The causal diagram approach, by contrast, is premised on the Humean and statistical approach to causation.  A causal mechanisms hypothesis is not fundamentally evaluated in terms of the statistical relationships among a set of variables; whereas a standard causal model is wholly intertwined with the mathematics of conditional correlation.

Consider a few examples. Here is a complex graphical representation of a process understood in terms of causal mechanisms from McGinnes and Elandy, "Unintended Behavioural Consequences of Publishing Performance Data: Is More Always Better?" (link):

Plainly this model is impossible to evaluate statistically by attempting to measure each of the variables; instead, the researchers proceed by validating the individual mechanisms identified here as well as the direction of influence they have on other intermediate outcomes. The outcome of interest is "quality of learning" at the center of the graph; and the diagram attempts to represent the complex structure of causal influences that exist among several dozen mechanisms or causal factors.

Here is another example of a causal mechanisms path diagram, this time representing the causal system involved in drought and mental health by Vins, Bell, Saha, and Hess (link).

Here too the model is not offered as a statistical representation of covariance among variables; rather, it is a hypothetical sketch of the factors which play in mechanisms leading from drought to depression and anxiety in a population. And the assessment of the model should not take the form of a statistical evaluation (a non-parametric structural equation model), but rather a piecemeal verification of the validity of the specific mechanisms cited. (John Gerring argues that this is a major weakness in causal mechanisms theory, however, in "Causal Mechanisms? Yes, But ..." (link).)

It seems, therefore, that the superficial similarity between a causal model graph (a DAG) and a causal mechanisms diagram is only skin-deep. Fundamentally the two approaches make very different assumptions about both ontology (what a causal relationship is) and epistemology (how we should empirically evaluate a causal claim). So it seems unlikely that it will be fruitful for causal-mechanisms theorists to attempt to adapt methods like DAGs to represent the causal claims they want to advance and evaluate.

Mechanisms, experiments, and policies

The social mechanisms approach to the social sciences aligns well with two key intellectual practices, experiments and policies. In an experiment we are interesting in testing whether a given factor has the effect it is thought to have. In a policy design we are interested in affecting an outcome of interest by manipulating some of the background conditions and factors. In both instances having a theory of the mechanisms in play in a domain permits us to frame our thinking better when it comes to designing experiments and policies.

Let's say that we are interested in reducing the high school dropout rate in a high-poverty school. We may have a hypothesis that one important causal factor that leads to a higher likelihood of dropping out is that high-poverty students have a much greater burden of family and social problems than students in low-poverty populations. We might describe the mechanism in question in these terms:

H1: (a) high burden of social/familial problems => (b) student has higher likelihood of becoming discouraged => (c) student has higher likelihood of stopping attending => (d) student has a higher likelihood of dropping out of high school

We can evaluate this hypothesis about one of the mechanisms of dropping out of high school in several ways. First, we note that each clause invokes a likelihood. This means that we need to look at sets of students rather than individual students. Single cases or individual pairs of cases will not suffice, since we cannot make any inference from data like these:

A. Individual X has high burden of social/familial problems; Individual X does not become discouraged; Individual X does not drop out of high school.

B. Individual Y has a low burden of social/familial problems; Individual Y does become discouraged; Individual Y does drop out of high school.

Observations A and B are both compatible with the possible truth of the mechanisms hypothesis. Instead, we need to examine groups of individuals with various configurations of the characteristics mentioned in the hypothesis. If H1 is true, it can only be evaluated using population observations.

In theory we might approach H1 experimentally: randomly select two groups G1 and G2 of individuals; expose G1 to a high burden of social/familial problems while G2 is exposed to a low burden of social/familial problems; and observe the incidence of dropping out of high school. This would be to treat the hypothesis through an experiment based on the logic of random controlled trials. The difficulty here is obvious: we are harming the individuals in G1 in order to assess the causal consequences of the harmful treatment. This raises an irresolvable ethical problem. (Here is a discussion of Nancy Cartwright's critique of the logic of RCT methodology in Evidence Based Policy; link.)

A slightly different experimental design would pass the ethics test. Select two schools S1 and S2 with comparable levels of high-poverty students and high burdens of social/familial problems for the individuals at the schools and comparable historical dropout rates. Now expose the students at S1 to a "treatment" that reduces the burden of social/familial problems (provide extensive social work services in the school that students can call upon). This design too conforms to the logic of a random controlled trial. Continue the treatment for four academic years and observe the graduation rates of the two schools. If H1 is true, we should expect that S1 will have a higher graduation rate than S2.

A third approach takes the form of a "quasi-experiment". Identify pairs of schools that are similar in many relevant respects, but differ with respect to the burden of social/familial problems. This is one way of "controlling" for the causal influence of other observable factors -- family income, race, degree of segregation in the school, etc. Now we have N pairs of matched schools and we can compute the graduation rate for the two components of the matches; that is, graduation rates for "high burden school" and "low burden school". If we find that the high burden schools have a lower graduation rate than the low burden schools, and if we are satisfied that the schools do not differ systematically in any other dimension, then we have a degree of confirmation for the causal hypothesis H1. But Stanley Lieberson in Making It Count poses some difficult challenges for the logic of this kind of experimental test; he believes that there are commonly unrecognized forms of selection bias in the makeup of the test cases that potentially invalidates any possible finding (link).

So far we have looked at ways of experimentally evaluating the link between (a) and (d). But H1 is more complex; it hypothesizes that social/familial problems exercise their influence through two behavioral stages that may themselves be the object of intervention. The link from (b) to (c) is an independent hypothetical causal relation, and likewise the link from (c) to (d). So we might attempt to tease out the workings of these links in the mechanism as well. Here we might design our experiments around populations of high burden students, but attempt to find ways of influencing either discouragement or the link from discouragement to non-attendance (or possibly the link from non-attendance to full dropping out).

Here our intervention might go along these lines: the burden of social/familial problems is usually exogenous and untreatable. But within-school programs like intensive peer mentoring and encouragement might serve to offset the discouragement that otherwise results from high burden of social/familial problems. This can be experimentally evaluated using one or another of the designs mentioned above. Or we might take discouragement as a given but find an intervention that prevents the discouraged student from becoming a truant -- perhaps a strong motivational incentive dependent on achieving 90% attendance during a six-week period.

In other words, causal hypotheses about causal mechanisms invite experimental and quasi-experimental investigation.

What about the other side of the equation; how do hypotheses about mechanisms contribute to policy intervention? This seems even more straightforward than the first question. The mechanism hypothesis points to several specific locations where intervention could affect the negative outcome with which we are concerned -- dropping out of high school in this case. If we have experimental evidence supporting the links specified in the hypothesis, then equally we have a set of policy options available to us. We can design a policy intervention that seeks to do one or more of the following things: reduce the burden of social/familial problems; increase the level of morale of students who are exposed to a high burden; find means of encouraging high-burden students to persevere; and design an intervention to encourage truants to return to school. This suite of interventions touches each of the causal connections specified in the hypothesis H1.

Now, finally, we are ready to close the circle by evaluating the success of interventions like these. Does the graduation rate of schools where the interventions have been implemented work out to be higher than those where the interventions were not implemented? Can we begin to assign efficacy assessments to various parts of the policy? Can we arrive at secondary hypotheses about why this policy intervention ("reduce the burden of social/familial issues") doesn't succeed, whereas another policy intervention ("bolster morale among high-risk students") does appear to succeed?

The upshot is that experiments and policies are opposite sides of the same coin. Both proceed from the common assumption that social causes are real; that we can assess the causal significance of various factors through experimentation and controlled observation; and that we can intervene in real-world processes with policy tools designed to exert influence at key junctures in the causal process.

George and Bennett on case study methodology

Establishing causal relationships within the fabric of the social world is more challenging than in the biological or physical-chemical domains. The reasons for this difficulty are familiar — the high degree of contextuality and contingency that is characteristic of social change, the non-deterministic character of social causation, and the fact that most social outcomes are the result of unforeseen conjunctions of independent influences, to name several.

Alexander George and Andrew Bennett argue for the value of a case-study method of social research in Case Studies and Theory Development in the Social Sciences. The idea here is that social researchers can learn about the causation of particular events and sequences by examining them in detail and in comparison with carefully selected alternative examples.

Here is how they describe the case-study method:

The method and logic of structured, focused comparison is simple and straightforward. The method is “structured” in that the researcher writes general questions that reflect the research objective and that these questions are asked of each case under study to guide and standardize data collection, thereby making systematic comparison and cumulation of the findings of the cases possible. The method is “focused” in that it deals only with certain aspects of the historical cases examined. The requirements for structure and focus apply equally to individual cases since they may later be joined by additional cases. (67)

George and Bennett believe that the techniques and heuristics of the case study approach permit the researcher to arrive at rigorous and differentiated hypotheses about underlying social processes. In particular, they believe that the method of process-tracing has substantial power in social research, permitting the researcher to move from the details of a particular historical case to more general hypotheses about causal mechanisms and processes in other contexts as well (6). They discourage research strategies based on the covering-law model, in which researchers would seek out high-level generalizations about social events and outcomes: “highly general and abstract theories … are too general to make sharp theoretical predictions or to guide policy” (7). But they also note the limits of policy relevance of “independent, stable causal mechanisms” (7), because social mechanisms interact in context-dependent ways that are difficult or impossible to anticipate. It is therefore difficult to design policy interventions based on knowledge of a few relevant and operative mechanisms within the domain of behavior the policy is expected to govern, since the workings of the mechanisms in concrete circumstances are difficult to project.

Fundamentally they align with the causal mechanisms approach to social explanation. Here is how they define a causal mechanism:

We define causal mechanisms as ultimately unobservable physical, social, or psychological processes through which agents with causal capacities operate, but only in specific contexts or conditions, to transfer energy, information, or matter to other entities. In so doing, the causal agent changes the affected entity’s characteristics, capacities, or propensities in ways that press until subsequent causal mechanisms act upon it. (137)

And they believe that the case-study method is a suite of methodological approaches that permit identification and exploration of underlying causal mechanisms.

The case study approach – the detailed examination of an aspect of a historical episode to develop or test historical explanations that may be generalizable to other events – has come in and out of favor over the past five decades as researchers have explored the possibilities of statistical methods … and formal models. (5)

The case study method is designed to identify causal connections within a domain of social phenomena.

Scientific realists who have emphasized that explanation requires not merely correlational data, but also knowledge of intervening causal mechanisms, have not yet had much to say on methods for generating such knowledge. The method of process-tracing is relevant for generating and analyzing data on the causal mechanisms, or processes, events, actions, expectations, and other intervening variables, that link putative causes to observed effects. (214)

How is that to be accomplished? The most important tool that George and Bennett describe is the method of process tracing. "The process-tracing method attempts to identify the intervening causal process--the causal chain and causal mechanism--between an independent variable (or variables) and the outcome of the dependent variable" (206). Process tracing requires the researcher to examine linkages within the details of the case they are studying, and then to assess specific hypotheses about how these links might be causally mediated. 

Suppose we are interested in a period of violent mobilization VM in the countryside at time t, and we observe a marked upswing of religious participation RP in the villages where we have observations. We might hypothesize that the surge of religious participation contributed causally to the political mobilization that ensued. But a process-tracing methodology requires that we we consider as full a range of alternative possibilities as we can: that both religious and political activism were the joint effect of some other social process; that religious participation was caused by political mobilization rather than caused that mobilization; that the two processes were just contingent and unrelated simultaneous developments. What can we discover within the facts of the case that would allow us to disentangle these various causal possibilities? If RP was the cause of VM, there should be traces of the influence that VM exerted within the historical record -- priests who show up in the interrogation cells, organizational linkages that are uncovered through archival documents, and the like. This is the work of process tracing in the particular case. And I agree with George and Bennett that there is often ample empirical evidence available in the historical record to permit this kind of discovery.

Finally, George and Bennett believe that process-tracing can occur at a variety of levels:

The simplest variety of process-tracing takes the form of a detailed narrative or story presented in the form of a chronicle that purports to throw light on how an event came about.... A substantially different variety of process-tracing converts a historical narrative into an analytical causal explanation couched in explicit theoretical forms.... In another variety of process-tracing, the investigator constructs a general explanation rather than a detailed tracing of a causal process. (210-211)

One of the strengths of the book is an appendix presenting a very good collection of research studies that illustrate the case study methodology that they explore. There are examples from American politics, comparative politics, and international relations. These examples are very helpful because they give substance to the methodological ideas presented in the main body of the book.

Mechanisms thinking in international relations theory

source: Alex Cooley, "America and Empire" (link)

One of the most fundamental ideas underlying the philosophy of social science expressed here and elsewhere is the view that social explanations should seek out the causal mechanisms that underly the social phenomena of interest. So now we need to be able to say a lot more about what social mechanisms are, and how they relate to each other. Quite a bit of my own thinking has been devoted to this subject, and in a recent post I proposed that it would be useful to begin to compile an inventory of social mechanisms currently in use in the social sciences (link). There I suggested that it would be useful to find a motivated way of classifying the mechanisms that we discover.

Interest in mechanisms is taking hold in some sub-disciplines of political science. An especially clear statement of the appeal of the mechanisms theory of explanation for political science is offered by Andrew Bennett in "The Mother of All Isms: Causal Mechanisms and Structured Pluralism in International Relations Theory" (link). (Bennett is also co-author with Alexander George of the excellent book on case-study methodology, Case Studies and Theory Development in the Social Sciences.) In the current article Bennett reviews the progression that has occurred in IR theory from positivism and the covering law model, to the idea of high-level "paradigms" of explanation, to the idea of a diverse set of causal mechanisms as the foundation of explanation in the field. He calls the latter position "analytic eclecticism", and he argues that it is a powerful and flexible way of thinking about the processes and research questions that make up the subject matter of IR theory.

In order to advance the value of mechanisms theory for working political scientists, Bennett argues that it will be helpful to attempt to classify the large number of mechanisms currently in use in IR theory in terms of a small number of dimensions. He proposes two dimensions in terms of which to analyze social mechanisms, which can be summarized as content and structure. The content dimension asks the question, what substantive social entities or properties are invoked by the mechanism? And the structure dimension asks the question, what is the nature of the relationship invoked by the mechanism? He proposes three large types of content: material power, functional efficiency, and legitimacy. And he suggests that there are four basic structures that can be formed: agent to agent, structure to agent, agent to structure, and structure to structure. (Notice that this corresponds exactly to the four arrows in Coleman's boat, including the Type 4 "structure to structure" connection.) Here is how Bennett motivates this classification scheme:

This tripartite division of categories of mechanisms usefully mirrors the three leading ‘isms’ in the IR subfield: (neo)realism (with a focus on material power); (neo)liberalism (institutional efficiency); and constructivism (legitimacy). It thereby provides a bridge to the vast literature couched in terms of the isms, preserving this literature’s genuine contributions toward better theories on mechanisms of power, institutions, and social roles. (472)

Here is the resulting classification of social mechanisms that Bennett offers:

Others have found this approach to be promising. Here is an elaboration on Bennett's classification by Mikko Huotari at the Mercator Institute for China Studies in Berlin:

(Thanks for sharing this classification, Mikko.)

I agree with Andrew in thinking that it is useful to find a non-arbitrary way of classifying mechanisms. It is quite worthwhile to make a start at this project. I'm not yet fully persuaded, however, by either of the axes that he proposes.

First, the content axis seems arbitrary -- legitimacy, material power, functional efficiency. Why choose these substantive characteristics rather than a dozen other possible content features? Is it simply that these correspond to the three primary "isms" of IR theory -- neorealism, neoliberalism, and constructivism (as he suggests earlier; 472)? But the thrust of the first part of the paper is that the "isms" are an unsatisfactory basis for guiding explanation in international relations theory; so why should we imagine that they serve to identify the crucial distinctions in content among social mechanisms? Would the content categories look different if we were taking our examples from feminist sociology, the sociology of organizations, or theories of legislatures? Bennett doesn't assert that these content categories are exhaustive; but if they are not, then somehow the tabulation needs to indicate that there is an extensible list on the left. And are these categories exclusive? Can a given mechanism fall both into the legitimacy group and the functional efficiency group? It would appear that this is possible; but in that case classification is difficult to carry out.

Second, the structure axis. Why is it crucial to differentiate mechanisms according to their place within an agent-structure grid? Why is this an illuminating or fundamental feature of the mechanisms that are enumerated? Would this dimension explode if we thought of social organization as a continuum from macro to meso to micro (along the lines of Jepperson and Meyer (link), as well as several earlier posts here (link))?

An early question that needs answer here is this: What do we want from a scheme of classification of social mechanisms? Should we be looking for a strict classification with exhaustive and mutually exclusive groupings? Or should we be looking for something looser -- perhaps more like a cluster diagram in which some mechanisms are closer to each other than they are to others?

We do have several other examples to think about when it comes to classifying mechanisms. In an earlier post I discussed Craver and Darden's account of mechanisms in biology, and highlighted the table of mechanisms that they provide (link). It is evident that the Craver-Darden table is much less ambitious when it comes to classification. They have loosely grouped mechanisms into higher-level types -- adaptation, repair, synthesis, for example; but they have not tried to further classify mechanisms in terms of the levels of the entities that are linked by the mechanism. So they offer one dimension of classification rather than two, and they leave it entirely open that there may be additional types to be added in the future. This is a fairly unexacting understanding of what is needed for a tabulation of mechanisms.

In Dynamics of Contention McAdam, Tarrow and Tilly offer a sort of classification of their own for the kinds of mechanisms they identify. They propose three types of mechanisms -- environmental, cognitive, and relational (kl 375):

• Environmental mechanisms mean externally generated influences on conditions affecting social life. Such mechanisms can operate directly: For example, resource depletion or enhancement affects people's capacity to engage in contentious politics (McCarthy and Zald, ed. 1987).
• Cognitive mechanisms operate through alterations of individual and collective perception; words like recognize, understand, reinterpret, and classify characterize such mechanisms. Our vignettes from Paris and Greenwood show people shifting in awareness of what could happen through collective action; when we look more closely, we will observe multiple cognitive mechanisms at work, individual by individual. For example, commitment is a widely recurrent individual mechanism in which persons who individually would prefer not to take the risks of collective action find themselves unable to withdraw without hurting others whose solidarity they value - sometimes at the cost of suffering serious loss.
• Relational mechanisms alter connections among people, groups, and interpersonal networks. Brokerage, a mechanism that recurs throughout Parts II and III of the book, we define as the linking of two or more previously unconnected social sites by a unit that mediates their relations with one another and/or with yet other sites. Most analysts see brokerage as a mechanism relating groups and individuals to one another in stable sites, but it can also become a relational mechanism for mobilization during periods of contentious politics, as new groups are thrown together by increased interaction and uncertainty, thus discovering their common interests.

This too is a one-dimensional classification. And it appears to be intended to be exhaustive and mutually exclusive. But it isn't clear to me that it succeeds in classifying all the mechanisms we might want to bring forward. Once again, this strikes me as a good beginning but not an exhaustive grouping of all social mechanisms.

My own preliminary grouping of mechanisms has even less structure (link). It groups mechanisms according to the subject matter or discipline from which they have emerged. But this does not serve to shed light on how these examples are similar or different from each other -- one of the key purposes of a classification.

I think this is a very useful research activity, and Andrew Bennett has done a service to the theory of social mechanisms in putting forward this effort at classification. Let's see what other schemes may be possible as well. A good scheme of classification may tell us something very important about the nature of how causation works in the social world.

Are there meso-level social mechanisms?

It is fairly well accepted that there are social mechanisms underlying various patterns of the social world — free-rider problems, communications networks, etc. But the examples that come readily to mind are generally specified at the level of individuals. The new institutionalists, for example, describe numerous social mechanisms that explain social outcomes; but these mechanisms typically have to do with the actions that purposive individuals take within a given set of rules and incentives.

The question here is whether we can also make sense of the notion of a mechanism that takes place at the social level. Are there meso-level social mechanisms? (As always, it is acknowledged that social stuff depends on the actions of the actors.)

This question is analogous to two other similar issues in other special sciences:

• Are there information-system level causal mechanisms in human cognition?
• Are there cellular-level causal mechanisms in biological systems?

Or, to the contrary, are all mechanisms in sociology, cognition science, and biology properly understood to be carried out at the level of individuals, neurons, and biochemistry?

Here is my version of a definition of a causal mechanism (link):

A causal mechanism is (i) a particular configuration of conditions and processes that (ii) always or normally leads from one set of conditions C to an outcome O (iii) through the properties and powers of the events and entities in the domain of concern. 

And here is the definition offered by Doug McAdam, Sidney Tarrow, and Chuck Tilly in Dynamics of Contention:

Mechanisms are a delimited class of events that alter relations among specified sets of elements in identical or closely similar ways over a variety of situations. (kl 354)

We should begin by asking what it is that we are looking for. What would a meso-level mechanism look like?

Here is a start: it would be a linkage between two conditions or entities, each of which is itself a meso-level structure or entity. So a meso-level causal mechanism is one in which both C and O are meso-level entities or conditions and where C leads to O "always or normally".

Earlier I argued that meso-level entities possess causal powers: regular dispositions to produce specified effects, grounded in the substrate of social activity (link). If some of those effects O_i are themselves meso-level outcomes or structures, then our question here is answered. Any pair {C,O_i} is itself a meso-level causal mechanism. If, on the other hand, the causal powers of meso-level entities are restricted to changes in the behavior of individuals, then meso-level mechanisms do not exist.

McAdam, Tarrow, and Tilly address a very similar question in Dynamics of Contention, and they argue for what they call "relational mechanisms":

Relational mechanisms alter connections among people, groups, and interpersonal networks. Brokerage, a mechanism that recurs throughout Parts II and III of the book, we define as the linking of two or more previously unconnected social sites by a unit that mediates their relations with one another and/or with yet other sites. Most analysts see brokerage as a mechanism relating groups and individuals to one another in stable sites, but it can also become a relational mechanism for mobilization during periods of contentious politics, as new groups are thrown together by increased interaction and uncertainty, thus discovering their common interests. (kl 376)

Having formulated the question in these terms, it seems that we can provide a credible affirmative answer: it is possible to identify a raft of social explanations in sociology that represent causal assertions of social mechanisms linking one meso-level condition to another. Here are a few examples:

• Al Young: decreasing social isolation causes rising inter-group hostility (link)
• Michael Mann: the presence of paramilitary organizations makes fascist mobilization more likely (link)
• Robert Sampson: features of neighborhoods influence crime rates (link)
• Chuck Tilly: the availability of trust networks makes political mobilization more likely (link)
• Robert Brenner: the divided sovereignty system of French feudalism impeded agricultural modernization (link)
• Charles Perrow: legislative control of regulatory agencies causes poor enforcement performance (link)

We might also consider the possibility of compound meso-level mechanisms, in which M₁ produces M₂ which in turn produces M₃. Does the sequence also qualify as a mechanism? That depends on the strength of the relationships that exist at each link; if the conditional probabilities of the links fall low enough, then the compound probability of the chain is no longer sufficient to satisfy condition (ii) above ("initial condition normally leads to the outcome").

Essentially this question comes down to the tightness of the linkages that exist among the sub-components of social systems. If there are sub-components within bureaucracies that maintain their properties and are tightly linked to specified outcomes, then these can play a role within meso-level causal mechanism narratives. If, on the other hand, the effects of a given subcomponent of a social system vary widely over time and space, then that type of component does not play a useful role in a causal mechanisms analysis. So the question of how extensive meso-level causal mechanisms are is itself an empirical one; it depends on the specific features of the social world.

So it seems as though we can offer two related conclusions about the causal reality of meso-level entities: meso-level structures possess causal powers, and there are causal mechanisms that invoke meso-level entities as both input and output.

Social powers?

I am one of those people who think that causal claims are the foundation of almost all explanations. When we ask for an explanation of something, we generally want to know why and how it came to be, and this means looking into its causal history. Moreover, I have believed for many years that this means looking for a set of causal mechanisms whose workings contribute to the outcome. And I subscribe to the anti-Humean idea that a causal relation involves some kind of necessity from cause to effect -- there is something in the substrate that necessitates the transition from cause to effect. The cause forces the effect to occur. (These ideas were first expressed in Varieties of Social Explanation.)

This means that my philosophy of social science has affinities to both large bodies of thought about causation today -- mechanisms and powers. The connection to mechanisms is explicit. The connection to powers is less direct but no less genuine. Essentially it comes down to the idea of necessity -- the idea that the properties of the causing thing, in the setting under consideration, actively produce its effects. This is what Ruth Groff refers to as an anti-passivist philosophy of causation.

One thing that makes me a little nervous about the current powers literature, though, is a kind of essentialism that it often seems to bring along. Rom Harré expressed this in his early formulations: it is the essential properties of a thing that create its causal powers. Here is how Stephen Pratten describes Harré's view (link):

Causal powers are, for Harré and Madden, properties of concrete powerful particulars which they possess in virtue of their essential natures.They analyse the ascription of causal powers to a thing in the following way: ‘ “X has the power to A” means “X will/can do A, in the appropriate circumstances in virtue of its intrinsic nature” ' (1975: 86).

And current powers theorists make similar claims. But I don't think things have an essential nature in any rigorous sense. So I'd rather see a powers theory whose formulation avoids reference to essential characteristics.

This is particularly important in the realm of the greatest interest to me, the social world. I believe that social entities are plastic and heterogeneous, and I don't think there are social kinds in a strong metaphysical sense. This entails that social entities do not have essential properties. So if powers theory depends on essentialism, then it seems not to apply in my understanding of the nature of the social world.

Fortunately essentialism is not essential! We can formulate an account of the causal powers of a social thing in terms of its contingent and changing properties and we don't have to hypostatize social things.

The way this works is that we do understand how the substrate of causal interconnection works in the social world. Social causation always works through the thoughts and actions of socially situated purposive actors. Individuals form representations of the world around them, both social and natural, they form relationships with other actors, and they act accordingly. So social structures acquire causal powers by shaping and incentivizing the individuals they touch.

So when we say that a certain social entity, structure, or institution has a certain power or capacity, we know what that means: given its configuration, it creates an action environment in which individuals commonly perform a certain kind of action. This is the downward strut in the Coleman's Boat diagram (link).

This construction has two important consequences. First, powers are not "irreducible" -- rather, we can explain how they work by analyzing the specific environment of formation and choice they create. And second, they are not essential. Change the institution even slightly and we may find that it has very different causal powers and capacities. Change the rules of liability for open range grazing and you get different patterns of behavior by ranchers and farmers (Order without Law: How Neighbors Settle Disputes).