Supervenience and social entities

What is the ontological status of social entities -- kinship systems, police departments, religious movements?  And what is the status of causal powers of social entities?  Do we need to "reduce" social entities to the compounds of individuals who make them up? And do we need to derive the causal properties of social entities from the characteristics and interactions of the individuals who make them up? In short, do we need to be reductionist about the social world?

This is a question that arises in many of the "special" sciences, including in particular psychology and neuroscience.  A basic premise of contemporary philosophy is that all phenomena are composed of physical entities, processes, and systems. The mind-body problem is the most immediate place where physicalism does some important work. Mind-body dualists held that mental states were independent from physical states, whereas physicalists insist that mental states are embodied in physical processes and systems.

It is plain enough that we make use of a vocabulary that doesn't appear to invoke physical states when we talk about people's actions and states of mind. When I decide to have tofu for dinner, or when I experience the taste of hot sesame oil, I am engaging in a mental act or qualitative state. "Deciding", "experiencing", and "having qualitative states" all appear to be terms that refer to private mental states. The physicalist takes it as a piece of ontological certainty, however, that these "mental" states are fully and entirely constituted by the physical substrates of the brain and nervous system.

So what do physicalists have in mind when they say that the phenomena to which these terms refer are really physical states? There are several possibilities:

• eliminative materialism: mental states do not exist, and we need to give definitions of mental terms that allow us to eliminate them in favor of physical terms [reductionism]
• non-eliminative materialism: mental states exist, but they are wholly and exhaustively caused by physical states
• epiphenomenalism: mental states are by-products of physical states without causal powers to influence subsequent physical states
• supervenience: mental states depend upon physical states and nothing else, but it is difficult and unnecessary to reduce facts about mental states to facts about physical states

Is there an analogous situation in the social sciences? Is individualism for the social sciences strongly analogous to physicalism for the natural sciences? Is there something ontologically dubious about referring to social entities and causes?

There is nothing peculiar about the idea that some entities are complex assemblages of other, simpler entities. Virtually every entity that we have an interest in is a compound of simpler entities -- genes, enzymes, or the insulin molecule depicted above.  A table has characteristics that depend on the physical features and arrangement of the materials that make it up, but those "table" characteristics are very different from the features of the composing elements -- hardness, stability, load-bearing capacity, etc. And there is no reason whatsoever to insist that "tables do not exist -- only bits of wood exist."  Tables are identifiable composite objects, and they have causal properties that we can invoke in explanations.  So the fact that there are characteristics of the composite that are dissimilar from the characteristics of the elements is not peculiar.  And this is entirely true of social entities as well.  The efficiency or corruptibility of a tax-collecting bureau is not a characteristic of the individuals who compose it; it is rather a system-level characteristic that derives from the incentives, oversight mechanisms, and physical infrastructure of the organization.

So composite entities are not suspect in general. However, there are a couple of challenging questions that we need to confront about composite entities.  First, can we explain the properties of the composite by knowing everything about the properties of the elements and the nature of their arrangement and interactions?  Can we derive the properties of the whole from the properties of the components?  Take metallurgy: can we derive the properties of the alloy from the physical characteristics of the tin and copper which make it up?  Or are there "emergent" properties that somehow do not depend solely on the properties of the components?

Second, can we attribute causal powers to composite entities directly, or do we need to disaggregate causal claims about the aggregate onto some set of claims about the causal powers of the elements?  Do we need to disaggregate the load-bearing capacity of the table onto a set of facts about the properties of the elements (legs, table top) and their configuration?  It is certainly true that we can derive the load-bearing capacity of the table from this set of facts; this is what civil engineers do in modeling bridges, for example.  The philosophical question is whether we ought to regard this causal property as simply a way of summarizing the underlying physics of the table, or as a stable causal property in its own right.

One appealing answer that has been offered to the question of the relationship between levels of entities is the theory of supervenience.  This theory is largely the work of philosopher Jaegwon Kim over the past thirty years. Here is a recent synthesis of his views (Physicalism, or Something Near Enough). He summarizes the basic idea in these terms:

It will suffice to understand [supervenience] as the claim that what happens in our mental life is wholly dependent on, and determined by, what happens with our bodily processes. (14)

And here is how Julie Zahle puts the point in her contribution to Turner and Risjord's Philosophy of Anthropology and Sociology:

Social entities, their properties, actions, etc. may be said to supervene upon individuals, their actions, and so on, insofar as: (1) there can be no difference at the level of social wholes, their properties, actions, etc., unless there is also a difference at the level of individuals, their properties, actions, and so on; (2) individuals, their actions, etc. fix or determine what kinds of social wholes, properties, etc. are instantiated. (327)

Does the idea of supervenience help answer the question of the ontological status of social entities?  Is it helpful to judge that social entities supervene upon facts about individuals and nothing else?  And does this leave room for the idea of social causation and relative explanatory autonomy?  Are we able to acknowledge the dependence of the social world on facts about individuals without abandoning the idea that there is social causation and social science?

Perhaps surprisingly, Kim thinks that the theory will not assist us in the last two ways, at least when it comes to psychology:

This view [supervenience] provides the burgeoning science of psychology and cognition with a philosophical rationale as an autonomous science in its own right: it investigates these irreducible psychological properties, functions, and capacities, discovering laws and regularities governing them and generating law-based explanations and predictions. It is a science with its own proper domain untouched by other sciences, especially those at the lower levels, like biology, chemistry, and physics. 

This seductive picture, however, turns out to be a piece of wishful thinking, when we consider the problem of mental causation--how it is possible, on such a picture, for mentality to have causal powers, powers to influence the course of natural events. (15)

So I am in a quandary at the moment: I favor the idea of "relatively autonomous social explanations" (link), I like the idea of regarding social entities as legitimate compound entities that don't require elimination, and I think of the theory of supervenience as providing some authority for these views.  And yet Kim himself seems to reject this line of thought when it comes to the special sciences of psychology and cognitive science.  Kim seems to want to argue that higher level sciences cannot claim relative autonomy; in this respect his own view seems to be reductionist.

What seems clear to me can be summarized in just a few points:

• Social entities and facts are determined and constituted by facts about individuals, their beliefs, their relations, and their actions.  So social entities and facts do in fact supervene upon facts about individuals.
• Social entities do have causal properties that can be discovered without needing to eliminate them in terms of properties of individuals.  
• The requirement of microfoundations is crucial because it establishes the intellectual discipline required by the first point: we must be able to validate that the claims we make about social properties and causal powers can be provided microfoundations at the level of socially situated individuals.
• There is a legitimate and defensible level of explanation at which social scientists can hypothesize social properties and causal capacities; so there is a place for a "relatively autonomous" social science.  We are not forced to be reductionist.
• The "social" is not inherently puzzling in the way that the "mental" is.  Social entities are more analogous to chairs and proteins than they are to thoughts and qualia: they are complex entities whose system-level characteristics are the ultimate effects of the interactions and properties of the individual elements that constitute them.  We often cannot trace out exactly how the properties of the whole derive from the properties of the components; but we don't need to do so except in unusual circumstances.

Social complexity

Social ensembles are often said to be "complex". What does this mean?

Herbert Simon is one of the seminal thinkers in the study of complexity. His 1962 article, "The Architecture of Complexity" (link), put forward several ideas that have become core to the conceptual frameworks of people who now study social complexity. So it is worthwhile highlighting a few of the key ideas that were put forward in that article. Here is Simon's definition of complexity:

Roughly, by a complex system I mean one made up of a large number of parts that interact in a nonsimple way. In such systems, the whole is more than the sum of the parts, not in an ultimate, metaphysical sense, but in the important pragmatic sense that, given the properties of the parts and the laws of their interaction, it is not a trivial matter to infer the properties of the whole. In the face of complexity, an in-principle reductionist may be at the same time a pragmatic holist. (468)

Notice several key ideas contained here, as well as several things that are not said. First, the complexity of a system derives from the "nonsimple" nature of the interaction of its parts (subsystems). A watch is a simple system, because it has many parts but the behavior of the whole is the simple sum of the direct mechanical interactions of the parts. The watchspring provides an (approximately) constant impulse to the gearwheel, producing a temporally regular motion in the gears. This motion pushes forward the time registers (second, minute, hour) in a fully predictable way. If the spring's tension influenced not only the gearwheel, but also the size of the step taken by the minute hand; or if the impulse provided by the spring varied significantly according to the alignment of the hour and second hands and the orientation of the spring -- then the behavior of the watch would be "complex". It would be difficult or impossible to predict the state of the time registers by counting the ticks in the watch gearwheel. So this is a first statement of the idea of complexity: the fact of multiple causal interactions among the many parts (subsystems) that make up the whole system.

A second main idea here is that the behavior of the system is difficult to predict as a result of the nonsimple interactions among the parts. In a complex system we cannot provide a simple aggregation model of the system that adds up the independent behaviors of the parts; rather, the parts are influenced in their behaviors by the behaviors of other components. The state of the system is fixed by interdependent subsystems; which implies that the system's behavior can oscillate wildly with apparently similar initial conditions. (This is one explanation of the Chernobyl nuclear meltdown: engineers attempted to "steer" the system to a safe shutdown by manipulating several control systems at once; but these control systems had complex effects on each other, with the result that the engineers catastrophically lost control of the system.)

A third important point here is Simon's distinction between "metaphysical reducibility" and "pragmatic holism." He accepts what we would today call the principle of supervenience: the state of the system supervenes upon the states of the parts. But he rejects the feasibility of performing a reduction of the behavior of the system to an account of the properties of the parts. He does not use the concept of "emergence" here, but this would be another way of putting his point: a metaphysically emergent property of a system is one that cannot in principle be derived from the characteristics of the parts. A pragmatically emergent property is one that supervenes upon the properties of the parts, but where it is computationally difficult or impossible to map the function from the state of the parts to the state of the system. This point has some relevance to the idea of "relative explanatory autonomy" mentioned in an earlier posting (link). The latter idea postulates that we can sometimes discover system properties (causal powers) of a complex system that are in principle fixed by the underlying parts, but where it is either impossible or unnecessary to discover the specific causal sequences through which the system's properties come to be as they are.

Another key idea in this article is Simon's idea of a hierarchic system.

By a hierarchic system, or hierarchy, I mean a system that is composed of interrelated subsystems, each of the latter being, in turn, hierarchic in structure until we reach some lowest level of elementary subsystem. (468) 

I have already given an example of one kind of hierarchy that is frequently encountered in the social sciences: a formal organization. Business firins, governments, universities all have a clearly visible parts-within-parts structure. (469)

Here the idea is also an important one. It is a formal specification of a particular kind of ensemble in which structures at one level of aggregation are found to be composed separately of structures or subsystems at a lower level of aggregation. Simon offers the example of a biological cell that can be analyzed into a set of exhaustive and mutually independent subsystems nested within each other. It is essential that there is a relation of enclosure as we descend the hierarchy of structures: the substructures of level S are entirely contained within it and do not serve as substructures of some other system S'.

It is difficult to think of biological examples that violate the conditions of hierarchy -- though we might ask whether an organism and its symbiote might be best understood as a non-hierarchical system. But examples are readily available in the social world. Labor unions and corporate PACs play significant causal roles in modern democracies. But they are not subsystems of the political process in a hierarchical sense: they are not contained within the state, and they play roles in non-state systems as well. (A business lobby group may influence both the policies chosen by a unit of government and the business strategy of a healthcare system.)

Simon appears to believe that hierarchies reduce the complexity of systems; and they support the feature of what we would now call "modularity", where we can treat the workings of a subsystem as a self-enclosed unit that works roughly the same no matter what changes occur in other subsystems.

Simon puts this point in his own language of "decomposability." A system is decomposable if we can disaggregate its behavior onto the sum of the independent behaviors of its parts. A system is "nearly decomposable" if the parts of the system have some effects on each other, but these effects are small relative to the overall workings of the system.

At least some kinds of hierarchic systems can be approximated successfully as nearly decomposable systems. The main theoretical findings from the approach can be summed up in two propositions:

(a) in a nearly decomposable system, the short-run behavior of each of the component subsystems is approximately independent of the short-run behavior of the other components; (b) in the long run, the behavior of any one of the components depends in only an aggregate way on the behavior of the other components. (474)

He illustrates this point in the case of social systems in these terms:

In the dynamics of social systems, where members of a system communicate with and influence other members, near decomposability is generally very prominent. This is most obvious in formal organizations, where the formal authority relation connects each member of the organization with one immediate superior and with a small number of subordinates. Of course many communications in organizations follow other channels than the lines of formal authoritv. But most of these channels lead from any particular individual to a very limited number of his superiors, subordinates, and associates. Hence, departmental boundaries play very much the same role as the walls in our heat example. (475)

And in summary:

We have seen that hierarchies have the property of near-decomposability. Intra-component linkages are generally stronger than intercomponent linkages. This fact has the effect of separating the high-frequency dynamics of a hierarchy -- involving the internal structure of the components-- from the low frequency dynamics-involving interaction among components. (477)

So why does Simon expect that systems will generally be hierarchical, and hierarchies will generally be near-decomposable?  It turns out that this is an expectation that derives from the notion that systems were created by designers (who would certainly favor these features because they make the system predictable and understandable) or evolved through some process of natural selection from simpler to more complex agglomerations.  So we might expect that hydroelectric plants and motion detector circuits in frogs' visual systems are hierarchical and near-decomposable.  

But here is an important point about social complexity.  Neither of these expectations is likely to be satisfied in the case of social systems.  Take the causal processes (sub-systems) that make up a city. And consider some aggregate properties we may be interested in -- emigration, resettlement, crime rates, school truancy, real estate values.  Some of the processes that influence these properties are designed (zoning boards, school management systems), but many are not.  Instead, they are the result of separate and non-teleological processes leading to the present.  And there is often a high degree of causal interaction among these separate processes.  As a result, it might be more reasonable to expect, contrary to Simon's line of thought here, that social systems are likely to embody greater complexity and less decomposability than the systems he uses as examples.

(A recent visit to the Center for Social Complexity at George Mason University (link) was very instructive for me.  There is a great deal of very interesting work underway at the Center using agent-based modeling techniques to understand large, complicated social processes: population movements, housing markets, deforestation, and more.  Particularly interesting is a blog by Andrew Crooks at the Center on various aspects of agent-based modeling of spatial processes.)  

Social things, kinds and meso causes

Consider a social entity -- say the IBM corporation -- and consider the group of individuals who currently make up the entity. What is the relation between the social entity and the individuals? There are several things that are plainly true: the entity is composed of the individuals. The behavior of the entity supervenes upon the individuals. But other issues are less clear. Does the corporation have causal properties of its own? Is this corporation an instance of a broader class of social entities with similar properties? Do we need to explain the corporation by deriving its properties from the properties of its constituents?

There are analogous issues in other areas of the "special" sciences. Take weather, for example. We have a fairly complex vocabulary of weather ontology and causation: El Nino, warm front, high pressure cell, hurricane, windshear, etc. Two things seem evident. First, weather phenomena are composed of lower-level physical components: bodies of water at a range of temperatures, non-homogeneous gases (atmosphere), inputs of energy (sunlight), and the like. Weather phenomena are "nothing but" ensembles of micro-cells of gases, water, land masses, and energy. Weather states might be exhaustively described as a set of micro-states with no reference to weather vocabulary; we could, if we wanted, reduce weather statements to micro physical statements. At the same time, we can legitimately describe weather phenomena at the meso or macro levels -- storms, etc. There is no compelling scientific reason to insist on reduction or elimination. And we can explain weather outcomes in terms of causal statements that invoke other meso factors. The fact that there are no autonomous weather phenomena does not mean there are no autonomous weather explanations.

There isn't much of a problem in defining or identifying entities at the level of social aggregates. The IBM corporation is no more ontologically suspect than Hurricane Irene. Both are high-level entities composed of lower level things and processes. What is more problematic is when we consider whether there are "kinds" of social entities; whether there are law-like generalizations that are true of those kinds of things; and whether kinds of social things have distinctive causal properties and processes.

Candidate social kinds might include armies, bureaucracies, and religions. I take the view, however, that none of these concepts identifies a set of social entities that have enough in common to call them a kind. They are heterogeneous in all the ways that have made them intriguing to historical sociologists. Caesar's army, Rommel's army, and Ho Chi Minh's army had some things in common; but they also had enormously important differences at the level of organization, technology, and leadership structure. So it isn't plausible to look at them as constituting a kind of social entity.

This high degree of heterogeneity among the items classified under social concepts also provides the basis for a negative answer to the question about laws as well. Armies are not like metals; they don't have a common set of generating processes, and they don't give rise to significant regularities. We can't say things like "all armies fight to exhaustion within 12 months."

We can, however, identify common processes and challenges that confront all armies, and we can consequently tease out some causal mechanisms and processes that recur across armies in a wide range of contexts. This is the thrust of McAdam, Tarrow and Tilly's analysis of mechanisms underlying contention.

This discussion illustrates three rather different points. First, there is the level of descriptive concepts we choose to use. Second is the issue of whether explanations need to proceed from more fundamental to more complex. And third is the issue of the ontological status of composite entities.

These topics have been addressed under the rubric of inter-theoretic reduction for almost forty years. Jerry Fodor's "Special Sciences (Or: The Disunity of Science as a Working Hypothesis" (1974) (link) and "Special Sciences: Still Autonomous After all These Years" (1997) (link) present core arguments for the autonomy of the special sciences. He distinguishes between token-token identity and type-type identity, pointing out that it might be the case that the first kind of identity obtains while the second does not. His central argument is that the possibility of multiple functional realizability demonstrates that it will not be possible to reduce higher-level laws to lower-level laws.

Jaegwon Kim has addressed the issue of physicalism throughout his career, most recently in Physicalism, or Something Near Enough. He holds that the concept of supervenience provides a basis for finessing the demand for reduction from higher kevel to lower level. Julie Zahle raises some of these problems in her contribution to Philosophy of Anthropology and Sociology: A Volume in the Handbook of the Philosophy of Science Series, "Holism and Supervenience."

My general intuition is that the issue of inter-theoretic reduction and an insistence on a strong version of methodological individualism are much less important to the philosophy of social science than they have appeared to be. Once we have sufficiently understood the ways that individuals, institutions, networks, and values work at the local level, we are in a good position to characterize meso-level social entities like organizations and value systems. And we are intellectually empowered to try to discover the dynamic properties of these systems of actors and social arrangements. So meso level causal properties seem entirely legitimate in the social sciences.

Relative explanatory autonomy

In an earlier post I indicated a degree of disagreement with the premises of analytical sociology concerning the validity of methodological individualism (link). This disagreement comes down to three things.

First, for reasons I've referred to several times here and elsewhere (link), I prefer to refer to methodological localism rather than methodological individualism.

This theory of social entities affirms that there are large social structures and facts that influence social outcomes. But it insists that these structures are only possible insofar as they are embodied in the actions and states of socially constructed individuals. The “molecule” of all social life is the socially constructed and socially situated individual, who lives, acts, and develops within a set of local social relationships, institutions, norms, and rules. (link)

I believe that the ideas of localism and socially constructed, socially situated actors do a better job of capturing the social molecule that underlies larger social processes than the simple idea of an "individual". Structural individualism seems to come to a similar idea, but less intuitively.

Second, the requirement of providing microfoundations for social assertions is preferable to methodological individualism because it is not inherently reductionist (link). A microfoundation is:

a specification of the ways that properties, structural features, and causal powers of a social entity are produced and reproduced by the actions and dispositions of socially situated individuals. (link)

We need to be confident that our theories and concepts about social structures, entities, and forces appropriately supervene upon facts about individuals; but we don't need to rehearse those links in every theory or explanation. In other words, we can make careful statements about macro-macro and macro-meso links without proceeding according to the logic of Coleman's boat -- up and down the struts. Jepperson and Meyer make this point in "Multiple Levels of Analysis and the Limitations of Methodological Individualisms" (link), and they offer an alternative to Coleman's macro-micro boat that incorporates explanations referring to meso-level causes (66).

Third, these points leave room for a meta-theory of relative explanatory autonomy for social explanations. The key insight here is that there are good epistemic and pragmatic reasons to countenance explanations at a meso-level of organization, without needing to reduce these explanations to the level of individual actors. Here is a statement of the idea of relative explanatory autonomy, provided by a distinguished philosopher of science, Lawrence Sklar, with respect to areas of the physical sciences:

Everybody agrees that there are a multitude of scientific theories that are conceptually and explanatorily autonomous with respect to the fundamental concepts and fundamental explanations of foundational physical theories. Conceptual autonomy means that there is no plausible way to define the concepts of the autonomous theories in terms of the concepts that we use in our foundational physics. This is so even if we allow a rather liberal notion of “definition” so that concepts defined as limit cases of the applicability of the concepts of foundational physics are still considered definable. Explanatory autonomy means that there is no way of deriving the explanatory general principles, the laws, of the autonomous theory from the laws of foundational physics. Once again this is agreed to be the case even if we use a liberal notion of “derivability” for the laws so that derivations that invoke limiting procedures are still counted as derivations. (link)

The idea of relative explanatory autonomy has been invoked by cognitive scientists against the reductionist claims of neuro-scientists. Of course cognitive mechanisms must be grounded in neurophysiological processes. But this doesn't entail that cognitive theories need to be reduced to neurophysiological statements. Sacha Bem reviews these arguments in "The Explanatory Autonomy of Psychology: Why a Mind is Not a Brain" (link). Michael Strevens summarizes some of these issues in "Explanatory Autonomy and Explanatory Irreducibility" (link). And here Geoffrey Hellman addresses the issues of reductionism and emergence in the special sciences in "Reductionism, Determination, Explanation".

These arguments are directly relevant to the social sciences, subject to several important caveats. First is the requirement of microfoundations: we need always to be able to plausibly connect the social constructs we hypothesize to the actions and mentalities of situated agents. And second is the requirement of ontological and causal stability: if we want to explain a meso-level phenomenon on the basis of the causal properties of other meso-level structures, we need to have confidence that the latter properties are reasonably stable over different instantiations. For example, if we believe that a certain organizational structure for tax collection is prone to corruption of the ground-level tax agents and want to use that feature as a cause of something else -- we need to have empirical evidence supporting the assertion of the corruption tendencies of this organizational form.

Explanatory autonomy is consistent with our principle requiring microfoundations at a lower ontological level. Here we have the sanction of the theory of supervenience to allow us to say that composition and explanation can be separated. We can settle on a level of meso or macro explanation without dropping down to the level of the actor. We need to be confident there are microfoundations, and the meso properties need to be causally robust. But if this is satisfied, we don't need to extend the explanation down to the actors.

Woven throughout this discussion are the ideas of reduction and emergence. An area of knowledge is reducible to a lower level if it is possible to derive the statements of the higher-level science from the properties of the lower level. A level of organization is emergent if it has properties that cannot be derived from features of its components. The strong sense of emergence holds that a composite entity sometimes possesses properties that are wholly independent from the properties of the units that compose it. Vitalism and mind-body dualism were strong forms of emergentism: life and mind were thought to possess characteristics that do not derive from the properties of inanimate molecules. Physicalism maintains that all phenomena -- including living systems -- depend ultimately upon physical entities and structures, so strong emergentism is rejected. But physicalism does not entail reductionism, so it is scientifically acceptable to provide explanations that presuppose relative explanatory autonomy.

Once we have reason to accept something like the idea of relative explanatory autonomy in the social sciences, we also have a strong basis for rejecting the exclusive validity of one particular approach to social explanation, the reductionist approach associated with methodological individualism and Coleman's boat. Rather, social scientists can legitimately aggregate explanations that call upon meso-level causal linkages without needing to reduce these to derivations from facts about individuals. And this implies the legitimacy of a fairly broad conception of methodological pluralism in the social science, constrained always by the requirement of microfoundations.

Microfoundationalism

detail: Lynn Cazabon photo

The philosophy of social science encompasses several important tasks, and key among them is to provide theories of social ontology and social explanation. What is the nature of social entities? What is needed in order to substantiate a claim of social causation? What constitutes an acceptable social explanation?

The concept of microfoundations is relevant to each of these domains. A microfoundation is:

a specification of the ways in which the properties and structure of a higher-level entity are produced by the activities and properties of lower-level entities.

In the case of the social sciences, this amounts to:

a specification of the ways that properties, structural features, and causal powers of a social entity are produced and reproduced by the actions and dispositions of socially situated individuals.

This concept is relevant to social ontology in this way. Social entities are understood to be compositional; they are assemblages constituted and maintained by the mentality and actions of individuals. So providing an account of the microfoundations of a structure or causal connection -- say a paramilitary organization or of the causal connection between high interest rates and the incidence of alcohol abuse -- is a specification of the composition of the social-level fact. It is a description of the agent-level relationships and patterns of behavior that cohere in such a way as to bring about the higher-level structure or causal relationship.

The concept of microfoundations is directly relevant to explanation. If we assert a causal or explanatory relation between one social entity or condition and another, we must be prepared to offer a credible sketch of the ways in which this influence is conveyed through the mentalities and actions of individuals.

Much turns, however, on what precisely we mean to require of a satisfactory explanation: a full specification of the microfoundations in every case, or a sketch of the way that a given social-level process might readily be embodied in individual-level activities. If we go with the second version, we are licensing a fair amount of autonomy for the social-level explanation; whereas if we go with the first version, we are tending towards a requirement of reductionism from higher to lower levels in every case. I am inclined to interpret the requirement in the second way; it doesn't seem necessary to disaggregate every claim like "organizational deficiencies at the Bhopal chemical plant caused the devastating chemical spill" onto specific individual-level activities. We understand pretty well, in a generic way, what the microfoundations of organizations are, and it isn't necessary to provide a detailed account in order to have a satisfactory explanation.

The ontological position associated with microfoundationalism falls in the general area of methodological individualism and reductionism, in that it insists on the compositional nature of the social. However, there is a recursive aspect of the theory that distinguishes it from strict reductionism. The individuals to which microfoundations are traced are not a-social; rather, their psychology, beliefs and motives are constituted and shaped by the social forces they and others constitute. So the microfoundational account of the workings of a paramilitary organization may well refer to the locally embodied effects of that organization on the current psychology of the members of the organization; and their behavior in turn reproduces the organization in the next iteration. This is why I prefer the idea of methodological localism over that of methodological individualism (link).

The theory of microfoundations is also very consistent with the idea of social mechanisms. When we ask about the microfoundations of a social process, we are asking about the mechanisms that exist at a lower level that create and maintain the social process.

One way of motivating the theory of microfoundations is to observe that it is a prescription against "magical thinking" in the social realm. There is no "social stuff" that has its own persistent causal and structural characteristics; rather, all social phenomena are constituted by patterns of behavior and thought of populations of individual human beings. And likewise, social events and structures do not have inherently social causal properties; rather, the causal properties of a social structure or event are constituted by the patterns of behavior and thought of the individuals who constitute them and nothing else.

The theory of supervenience is often invoked to express the idea that social entities and properties are constituted by individuals. (Jaegwon Kim is the primary creator of the theory of supervenience in the philosophy of mind; Supervenience and Mind: Selected Philosophical Essays.) This basic idea is expressed as:

No difference at level N without some difference at lower level K.

The advantage of the theory of supervenience is that it provides a way of recognizing the compositional nature of higher-level entities without presupposing explanatory reductionism from one level to the lower level.

The explicit idea of microfoundations appears to have been first developed in the domain of microeconomics; there it referred to the necessity of deriving macroeconomic phenomena from the premises of rational economic behavior (Weintraub, Microfoundations: The Compatibility of Microeconomics and Macroeconomics). (Here is an interesting article by van den Bergh and Gowdy on recent analysis of the microfoundations debate in economics.) Maarten Jansen describes the theory of microfoundations in economics in his entry in The New Palgrave Dictionary of Economics and the Law:

The quest to understand microfoundations is an effort to understand aggregate economic phenomena in terms of the behavior of individual economic entities and their interactions. These interactions can involve both market and non-market interactions. The quest for microfoundations grew out of the widely felt, but rarely explicitly stated, desire to stick to the position of methodological individualism ..., and also out of the growing uneasiness among economists in the late 1950s and 1960s with the co-existence of two subdisciplines, namely microeconomics and macroeconomics, both aiming at explaining features of the economy as a whole. Methodological individualism, as explained in the entry on the topic, is the view according to which proper explanations in the social sciences are those that are grounded in individual motivations and their behavior.

The idea of microfoundations is now important in many areas of the social sciences, including especially sociology and political science. Particularly important were ideas formulated by James Coleman in Foundations of Social Theory. Coleman doesn't use the term "microfoundations" explicitly in this work, but his analysis of the relationship between the macro and the micro seems to imply a requirement of providing microfoundations as a condition on good explanations in the social sciences. The Coleman boat (link) seems to be a graphical way of representing the microfoundations of a macro-level fact.

A second mode of explanation of the behavior of social systems entails examining processes internal to the system, involving its component parts, or units at a level below that of the system. The prototypical case is that in which the component parts are individuals who are members of the social system. In other cases the component parts may be institutions within the system or subgroups that are part of the system. In all cases the analysis can be seen as moving to a lower level than that of the system, explaining the behavior of the system by recourse to the behavior of its parts. This mode of explanation is not uniquely quantitative or uniquely qualitative, but may be either. ... I call [this] the internal analysis of system behavior. (2)

Coleman's view here is complex, though, and isn't entirely unambiguous. Consider this qualification a few pages later, which refers unexpectedly to "emergent phenomena" and intermediate levels of explanatory mechanisms between the macro and the micro:

Those readers familiar with debates and discussions on methodological holism and methodological individualism will recognize that the position taken above on explanation is a variant of methodological individualism. But it is a special variant. No assumption is made that the explanation of systemic behavior consists of nothing more than individual actions and orientations, taken in aggregate. The interaction among individuals is seen to result in emergent phenomena at the system level, that is, phenomena that were neither intended nor predicted by the individuals. Furthermore, there is no implication that for a given purpose an explanation must be taken all the way to the individual level to be satisfactory. The criterion is instead pragmatic. (5)

Other more explicit advocates of the microfoundations principle are Jon Elster, John Roemer, Adam Przeworski, and other contributors to the theories of analytical Marxism (Analytical Marxism). Here is how I attempted to synthesize some of this thinking in 1994:

Marxist thinkers have argued that macro-explanations stand in need of microfoundations: detailed accounts of the pathways by which macro-level social patterns come about. These theorists have held that it is necessary to provide an account of the circumstsances of individual choice and action that give rise to aggregate patterns if macro-explanations are to be adequate. Thus in order to explain the policies of the capitalist state it is not sufficient to observe that this state tends to serve capitalist interests; we need to have an account of the processes through which state policies are shaped or controlled so as to produce this outcome. ("Microfoundations of Marxism," reprinted in D. Little, Microfoundations, Methods, and Causation, 4)

As noted in a prior post, the idea of microfoundations is also a core constituent of the methodology of analytical sociology (Peter Hedström, Dissecting the Social: On the Principles of Analytical Sociology).

In short, a fairly wide range of social science research today embraces the general idea of providing microfoundations for macro-level assertions. And this seems to be a very reasonable requirement, given what we know about how social entities, processes, and forces are composed.

Structures and structuration

Several recent posts have focused on new thinking about how to characterize "agency". Much of that thinking is aimed at dissolving the distinction between agency and structure. So what remains to be said about "structure"? Has the structure side of the debate developed much in the past decade or so?

One of the important exponents of a structure-centered approach to social theory is Anthony Giddens. His 1979 book, Central Problems in Social Theory: Action, Structure, and Contradiction in Social Analysis, is one place where his views on structure are expressed. He separates himself from earlier versions of "structuralism," including Saussure and Levi-Strauss; but he advocates for the reality of social structures and the methodological appropriateness of attempting to arrive at empirically based theories of social structures in sociology and other areas of the social sciences. He emphasizes the fact of process rather than static organization in his theory of structures; to capture the "verbiness" of a process view, he prefers the language of "structuration" rather than "structuralism." This distinction having to do with the location of things in time -- a static snapshot versus a continuous process -- extends to his treatment of action as well:

'Action' or agency, as I use it, thus does not refer to a series of discrete acts combined together, but to a continuous flow of conduct. We may define action... as involving a 'stream of actual or contemplated causal interventions of corporeal beings in the ongoing process of events-in-the-world'. (55)

A recent and unapologetic treatment of the reality of social structures is presented in Dave Elder-Vass's The Causal Power of Social Structures (2010). Elder-Vass accepts the point that agency and structure are inseparable; neither functions as a solely sufficient cause of social outcomes. But he argues strongly for the idea that social structures have causal powers that are not reducible to facts about individuals. He places his analysis generally within the tradition of critical realism (Bhaskar, A Realist Theory of Science). And he relies heavily on the theory of supervenience to solve the riddle of how structures can be composed of individual-level activity and yet possess autonomous causal powers (Kim, Mind in a Physical World: An Essay on the Mind-Body Problem and Mental Causation).

Elder-Vass rejects the ontology of methodological individualism, which he regards as a species of reductionism: social properties need to be reducible to features of individuals. And yet he fully and unambiguously embraces the obvious fact that social structures must be composed of individuals in relations to each other. His way out of this apparent contradiction is to argue that social structures possess emergent causal powers: causal characteristics that pertain to the whole but not to the parts or their ensembles. Here is how he characterizes emergence:

A thing ... can have properties or capabilities that are not possessed by its parts. Such properties are called emergent properties. (4)

And later:

An emergent property is one that is not possessed by any of the parts individually and that would not be possessed by the full set of parts in the absence of a structuring set of relations between them. (17)

It is obvious and unexceptional that there are "emergent" properties in this limited sense. However, I'm not sure it captures the full concept of emergence. It would appear that holists have another idea in mind as well: the idea that the properties of the whole cannot be derived from knowledge of the properties of the parts and their relations to each other. ((E-V discusses this angle under the topic of "eliminative reductionism"; 24, 54.) A square figure is composed of four lines; the figure possesses area, whereas the lines do not. Sugar is sweet, but its component parts -- carbon, hydrogen, oxygen -- are not. So is "sweetness" an emergent property? Apparently it is not; because the sweetness of the compound can in fact be explained by the knowledge we have of the chemistry of the components, their molecular bonds, and the human sensors that register "sweetness." The property of sweetness can be derived from knowledge of the basic chemistry and the workings of the sensory system. So sweetness can be "reduced" to facts about the molecule and the sensor. It is not a novel causal power.

Here is why Elder-Vass thinks the concept of emergence is valuable in the context of social structures:

The value of the concept of emergence lies in its potential to explain how an entity can have a causal impact on the world in its own right: a causal impact that is not just the sum of the impacts its parts would have if they were not organised into this kind of whole. (5)

But really his central insight is that composites may have causal powers that are relatively autonomous from the powers of their parts:

Reductionist thinkers have argued that if we can explain how a causal power works in terms of lower-level forces, the original power itself becomes redundant to any explanation of its effects. By contrast, I argue in chapter 3 that when we explain a causal power, we do not explain it away. (6)

And here is his central point:

I shall argue that social entities [structures] are causally effective in their own right, with causal powers that are distinct from those of human individuals. But I shall also examine the mechanisms that underpin these causal powers, thus recognizing the contributory role that human individuals make to the functioning of social structures. (6)

(This last point is what I and others refer to as providing "microfoundations" for claims about social causation.)

So this seems to be Elder-Vass's core ontology of structures: structures are composed of individuals in relation to each other; structures have "emergent" causal powers that are not simply the sum of the causal powers of the component individuals; these emergent powers derive from the relations between the components; and for any particular causal power of the structure it must be possible to provide an explanation of the power in terms of mechanisms involving the individuals and their relations.

The bulk of the book takes the form of an effort to work out this view in detail with respect to two important types of social structures: organizations and normative circles. A bank is an organization (which in turn fits into a web of other organizations). A normative circle is a set of overlapping sets of individuals who both embody and conform to a normative rule or ideal. (E-V traces the idea here back to Durkheim and Simmel.) More on this later!

I suggested above that the concept of explanatory autonomy might work better than emergence. So what is a good definition of "relative autonomy" of causal powers at one level or another, with respect to the underlying entities? I think there are many areas of the sciences where we identify a set of causal processes that are sufficiently regular and predictable, that we don't feel obliged to drop down to the next level in order to achieve an adequate explanation. A level is "explanatorily autonomous" if entities at that level conform to a regular set of causal relations. Thinking of the brain as a computing system is an example. If we analyze the visual system in terms of a set of receptors, aggregators, detectors, analyzers, etc., and if this model allows us to explain and predict the organism's perceptual capacities (including mistakes), then we don't need to have a full theory of the neurophysiology that underlies. On this example, "perception as computational system" is relatively autonomous with respect to the underlying neurophysiology.

Arguments for social holism

The topic of methodological individualism (MI) came up in a recent posting, and I underlined the connection between MI and some version of reductionism. Here I'd like to take a different approach and ask the question, what considerations can be offered in support of some version of social holism?

Here are a couple of arguments that avoid the accusation of "spooky holism". (By spooky I mean "disembodied.")

First is a very reasonable point deriving from pragmatic objections to reductionism. If we know on ontological grounds that the behavior of the whole depends upon the features and behavior of the constituent parts and nothing else -- the heart of the theory of supervenience -- but also know that it is entirely hopeless to attempt to calculate the one based on facts about the other -- then perhaps it is justified to consider the whole as if it embodied causal processes at the macro-level. So there is a pragmatic argument available that recommends the autonomy of social facts based on the infeasibility of derivability.

Second is the plausibility of the idea that there are large historical or social forces that are for all intents and purposes beyond the control of any of the individuals whom they influence. The fact that a given population exists as a language community of German speakers or Yoruba speakers has an effect on every child born into that population. The child's brain is shaped by this social reality, quite independently from facts about the child's agency or individuality. The grammar of the local language is an autonomous social fact in this context -- even though it is a fact that is embodied in the particular brains and behaviors of the countless individuals who constitute this community. But this is probably similarly true when we turn to systems of attitudes, norms, or cognitive systems of thinking. (See a posting on social practices on this subject.)

It is obvious but perhaps trivial to observe that the vector of influence flows through individuals who possess the grammar, norms, or folk beliefs -- this is the ontological reality captured by the microfoundations thesis. But perhaps a point in favor of a modest holism is this: the fact of the commonality of Yoruba grammar can be viewed as if it were an autonomous fact -- even though we know it depends on the existence of Yoruba speakers. But the point of the holism thesis here remains: that the social fact of the current grammar is coercive with respect to current Nigerian children in specific communities. And, perhaps, likewise with respect to other aspects of social cognition and norms. And this takes us some distance towards Durkheim's central view -- the autonomy of social facts.

Now what about social structures? Can some instances of social structures be treated as if they were autonomous with respect to the individuals whom they affect? Here is how the home mortgage system works -- we can specify a collection of rules and practices X, Y, Z that regulate the transactions that occur within this system. The individual who wants to borrow from a bank or other financial institution is simply subject to these rules and practices. He/she doesn't have the option of rewriting the rules in a more rational or fair or socially progressive way; at a given point in time the rules and practices are fixed independently from the wishes or intentions of the people involved in the institution. Once again it is trivially true that these rules are embodied; but they function as if they were autonomous. And this is true for institutions at the full range of scope, from the local to the global.

The advocate for a modest social holism might maintain two plausible positions: first, that all social facts are embodied in the states of mind and behavior of individuals; but second, that some social facts (institutions, social practices, systems of rules) have explanatory autonomy independent from any knowledge we might be able to provide about the particular ways in which these facts are embodied in individuals. The first is an ontological point and the second is a point about explanation.

These points in favor of a modest holism are compatible with other important points about social entities -- the points about heterogeneity, plasticity, and opportunistic transformation that have been made elsewhere in this blog. In other words, we aren't forced to choose between "agent" and "structure"; rather, agents influence structures and structures influence agents.

These arguments suggest two things. First, holism and individualism are not so sharply opposed as perhaps they appear.

But more important, two styles of social explanation are validated and compatible: the compositional or aggregative model of explanation -- explain the outcome as the aggregative consequence of the behavior of large numbers of individuals -- and constraining or filtering explanations -- the structuring of individual behavior that is created by the workings of social institutions. The first model of explanation corresponds well to the assumptions of methodological individualism, while the second corresponds to the idea that structures and large social factors cause patterns of individual behavior. And neither has antecedent priority over the other. (Some of this variety of explanatory strategies is highlighted in an earlier posting on explanation.)

Methodological individualism

Methodological individualism (MI) is a doctrine in the philosophy of the social sciences about the relationship between "society" and individuals. The idea can be formulated in several related but somewhat different ways: social facts are constituted by facts about individuals; social entities are composed of individuals and their properties and relations; social structures and entities are "nothing but" ensembles of individuals and their behaviors; social explanations must be derivable from facts about individuals; scientific statements about "society" must be reducible to statements about individuals and their properties and relations; social laws or generalizations must be derivable from general facts about individuals. And there are probably other possible formulations as well.

So the doctrine of methodological individualism often represents a form of reductionism from one area of scientific theory to another: theories about social entities and properties must be reducible to theories and statements about individuals and their properties. This line of thought is parallel to materialism in the philosophy of mind or anti-vitalism in the philosophy of biology. Mental states must be reducible to a set of facts about neurophysiology; statements about living organisms must be reducible to a set of facts about the molecular chemistry and physiology of cells. (See Ingo Brigandt and Alan Love's article in the Stanford Encyclopedia of Philosophy on reduction in biology.)

A somewhat less restrictive view than reductionism is the theory of "supervenience" between levels of scientific description. According to the theory of supervenience, facts at one level of description are fixed or determined by facts at a lower level of description. To say that X supervenes upon Y is to say that there is no difference between states of affairs concerning X for which there is not also a difference in states of affairs concerning Y. This is a less restrictive doctrine because it doesn't require that we provide derivations of the facts of X from facts of Y. (Jaegwon Kim is the primary innovator here; see Physicalism, or Something Near Enough for a recent formulation of the theory.)

Methodological individualism is the limit version of a family of perspectives on social explanation that we might refer to as "agent-centered" approaches to social explanation. Here the general idea is that we explain social outcomes as an aggregate result of the actions, choices, and mentalities of individuals. Individuals' behavior and choice constitute the causal dynamics of social outcomes. The idea of "microfoundations" has played an important role in recent thinking about the relationship between social facts and individual facts. (See Microfoundations, Methods, and Causation for more on the idea of microfoundations for social explanations.)

But agent-centered approaches can give more "social-ness" to the individual than the founding statements of MI would permit. For example, the position of "methodological localism" identifies socially constituted and socially situated individuals as the foundation of social explanation; but this position explicitly denies the idea that all social facts are reducible to bare psychological facts about individuals. Rather, individuals are themselves constituted and constrained by previously established social conditions. (See "Levels of the Social" for more about methodological localism.)

The idea of methodological individualism is one that has appealed to philosophers and social thinkers for almost as long as there has been systematic thinking about social science. Modern philosophy of social science began in the nineteenth century, and John Stuart Mill's theories of social knowledge contained the assumption of methodological individualism (The Logic of the Moral Sciences). Max Weber also put forward the doctrine in The Methodology Of The Social Sciences. A classic statement was presented by J. W. N. Watkins (1968), "Methodological Individualism and Social Tendencies" in May Brodbeck, Readings in the Philosophy of the Social Sciences.

The logical contrary of MI is the idea of social holism, most explicitly advocated by Emile Durkheim in Rules of Sociological Method. Holism is a form of anti-reductionism; it maintains that there are facts about the social world that do not reduce to facts about individuals. Society is autonomous with respect to the individuals who "make it up". There are social forces (e.g. systems of norms) that exercise causal power over individuals, instead of norms being constituted by the psychological states of individuals. Other varieties of social holism are possible as well. Structuralism is the view that social structures exercise autonomous causal properties, as expressed in such authors as Levi Strauss, Althusser, and Foucault.

Arguments in favor of methodological individualism derive from several insights. First, there is the point that social facts are evidently constituted by the thoughts and behaviors of groups of individuals. Social movements are composed of individuals with specific psychologies and beliefs; organizations are composed of individuals; and, arguably, moralities and cultures are made up of individuals with specific beliefs and values. Second, there is the point that social "laws" are rare, exception-laden, and conditional; so there is a methodological reason to look for the more basic laws that may regulate social behavior -- at the level of individuals and their psychology. Third, there is a preference for ontological sparsity: if we can explain social facts in terms of facts about individuals, then we don't need to attribute ontological status to social facts and entities. Fourth, there is a "materialist" or anti-occultist preference that is appealing to many philosophers; the idea that social facts might be autonomous with respect to individuals gives an impression of occult causal powers or action at a distance. So there is a range of reasons to think that social outcomes are made up of or determined by the aggregate results of individuals and their interactions.

MI has been particularly appealing in certain disciplines of the social sciences -- especially economics and political science. In each case the perspective of "rational actor theory" has appeared to be a very promising line of explanation: explain the behavior of groups (consumers, voters) as the aggregate outcome of individuals making choices with a specific set of beliefs and preferences, within a specific set of constraints.

Other areas of the social sciences are less amenable to the theory of methodological individualism. Anthropology and sociology are disciplines that set the focus on the higher-level social conditions or causes that influence behavior -- a perspective that seems to be more comfortable with some form of holism. However, there is a range of opinion among practitioners of these disciplines as well, and there are anthropologists and sociologists who are more sympathetic to the impulses of methodological individualism.

It is important to highlight some points that MI does not entail. MI does not entail that individuals are egoists or purely self-regarding. It does not entail that individuals are not social. It does not entail that social facts do not have causal consequences -- for other social facts and for individual behavior. It is indeed possible to reframe almost all substantive sociological theory in terms that are consistent with the reasonable conditions implied by MI. Even Durkheim's central theories can be formulated in a way that is innocent with respect to the charge of "action at a distance". And, from the other direction, even a theorist with as clear a commitment to MI as Max Weber, is still able to make "macro" or "holistic" claims about the causal importance of factors such as religion or morality.

Are social facts reducible to something?

What is the relationship between facts about society and facts about individuals?

Reducibility means that the statements of one scientific discipline should be logically deducible from the truths of some other "more fundamental" discipline. It is sometimes maintained that the truths of chemistry ought in principle be derivable from those of quantum mechanics. A field of knowledge that is not reducible to another field R is said to be "autonomous with respect to R". Philosophers sometimes further distinguish "law-to-law" reduction, "type-to-type" reduction, "law-to-singular-fact" reduction, and "type-to-token" reduction.

Are social sciences such as economics, sociology, or political science reducible in principle to some other more fundamental field--perhaps psychology, neurophysiology, or the theory of rationality?

To begin to answer this question we must first decide what items might be reduced: statements, truths, laws, facts, categories, or generalizations. Second, we need to distinguish several reasons for failure of reduction: failure in principle, because events, types, and laws at the social level are simply not fixed by states of affairs at "lower" levels, and failure for reasons of limits on computation. (The motions of a five-body system of stars might be determined by the laws of gravitation even though it is practically possible to perform the calculations necessary to determine future states of the system.)

So now we can consider the question of social reduction in a reasonably clear form. Consider first the "facts" that pertain to a domain of phenomena--whether these facts are known or not. (I choose not to concentrate on laws or generalizations, because I am doubtful about the availability of strong laws of social phenomena.) Do the facts of a hypothetically complete theory of human psychology "fix in principle" the facts of economics or sociology, given appropriate information about boundary conditions?

One important approach to this problem is the theory of supervenience (Jaegwon Kim, Physicalism, or Something Near Enough). A level of description is said to supervene upon another level just in case there can be no differences of state at the first level without there being a difference of state in the second level. The theory is first applied to mental states and states of neurophysiology: "no differences in mental states without some difference in neurophysiology states." Supervenience theory implies an answer to the question of whether one set of facts "fixes in principle" the second set of facts. (It has been taken as obviously true that social facts supervene upon facts about individuals; how could it be otherwise? What other constitutive or causal factors might influence social facts, beyond the actions and ideas of individuals?) If the facts about social life supervene upon facts about the psychological states of individuals, then it follows that the totality of facts about individual psychology fixes in principle the totality of facts about social life. (Otherwise there would be the situation that there are two total social worlds corresponding to one total "individual psychology" world; so there would be a difference at the social level without a difference at the level of individual psychology.)

So this provides the beginnings of an answer to our question: if we believe that social facts supervene upon facts about individuals, then we are forced to accept that the totality of facts about individuals "fix" the facts about society.

However, supervenience does not imply "reducibility in principle", let alone "reducibility in practice" between levels. In order to have reducibility, it is necessary to have a system of statements describing features of the lower level which are sufficient to permit deductive derivation (or perhaps probabilistic inference) of all of the true statements contained in the higher-level domain. If it is a social fact that "collective action tends to fail when groups are large", then there would need to be set of statements at the level of individual psychology that logically entail this statement. Two additional logical features would appear to be required for reduction: a satisfactory set of bridge statements (linking the social term to some construction of individual-level terms; "collective action" to some set of features of individual agents, so there is a mapping of concepts and ontologies between the two domains), and at least some statements at the lower level that have the form of general laws or law-like probabilistic statements. (If there are no general statements at the lower level, then deductive inference will be limited to truth-functional deduction.)

Now it is time for a speculative leap: a judgment call on the question of whether we ought to look for reductive links between social facts and individual-level facts. My intuition is that it is not scientifically useful to do so, for several reasons. First is the point about computational limits: even if the outcome of a riot is "fixed" by the full psychological states of participants ex ante and their strategic interactions during the event--it is obviously impossible to gather that knowledge and aggregate it into a full and detailed model of the event. So deriving a description of the outcome from a huge set of facts about the participants is unpromising. Second, it is telling that we need to refer to the strategic interactions of participants in order to model the social event; this means that the social event has a dynamic internal structure that is sensitive to sub-events that occur along the way. (Jones negotiates with Smith more effectively than Brown negotiates with Black. The successful and failed negotiations make a difference in the outcome but are unpredictable and contingent.) Third, the facts at the social level rarely aggregate to simple laws or regularities that might have been derived from lower-level laws and regularities; instead, social outcomes are contingent and varied.

So for a variety of reasons, it is reasonable to take the view that social facts supervene upon facts about individuals, but that social explanations are autonomous from laws of psychology. (This final point might be paraphrased as "the laws of psychology underdetermine social outcomes.")