Atomism versus holism in social ontology

Frank Jackson and Philip Pettit addressed the micro-macro question in an essay called "Structural Explanation in Social Theory" (link), included in Reduction, Explanation, and Realism (1992, ch. 4). Particularly interesting is the brief distinction that they draw between two senses of individualism: atomism versus holism, and individualism versus collectivism. They believe that these two distinctions are quite different, and yet are commonly conflated. 

The first distinction has to do with the logical characteristics of the properties of intentionality attributed to the actor. The atomist holds that the actor's beliefs, rationality, and intentionality are fundamentally independent of the other individuals in the groups to which he or she belongs, whereas the holist maintains that the individual's beliefs, rationality, and intentionality are inextricably connected to the mentality of the group. "How far, for example, do I depend on the convergent responses of my fellows for being able to form concepts and think thoughts involving those concepts? The atomist tradition says that logically I do not depend on this way on the other members of my society. The non-atomist or holist tradition says that I do" (127). I suppose that both Hobbes and Descartes illustrate this sense of atomistic individuals -- Hobbes with his characterization of the mentality of individuals in the state of nature and Descartes in his philosophical method of doubt.

The second distinction has to do with explanatory direction and autonomy for one level or the other. It revolves around the question of whether micro-explanations are more fundamental than macro-explanations. Do individuals in the aggregate determine the properties of the social world, or do the properties of the social world determine the actions of individuals? "The other significant question ... has to do with how far the members of society, whether they are conceived of atomistically or holistically, retain their apparent autonomy in the presence of higher-order social constraints" (127). Strict collectivists maintain that individuals "must" behave in the way prescribed by social structure; strict individualists maintain that individuals retain the ability to act otherwise than as prescribed by social constraint. 

Jackson and Pettit suggest that the first distinction has to do with "horizontal" issues concerning relations among individuals, whereas the second distinction has to do with "vertical" issues having to do with individuals and social structures. Both holism and collectivism are anti-reductionist, in the sense that they each deny that social facts can be derived from some set of purely individual facts; but they amount to different kinds of claims about the priority of the social world over the aggregate of individuals.

It is worth asking whether the concept of "atomistic" offered here is consistent with the way the concept is usually understood. On the common presentation, atomism maintains that there is a core set of purely psychological characteristics possessed by every human being, and that these characteristics are pre-social. These might include a preference for self-interest, a concern for survival, and an ability to calculate risks and benefits of various actions. In his Stanford Encyclopedia of Philosophy article on methodological individualism, Joseph Heath paraphrases atomism in these terms: "The atomistic view is based upon the suggestion that it is possible to develop a complete characterization of individual psychology that is fully pre-social, then deduce what will happen when a group of individuals, so characterized, enter into interaction with one another" (section 1). The Jackson-Pettit account captures the idea that atomism assumes a pre-social individual with basic characteristics that do not logically entail anything about the social environment. But what is somewhat unclear here is their use of "logical" in this discussion. One way of understanding their meaning is that they mean to assert that, for the atomist, the "basic" psychological characteristics can be formulated in a way that does not imply anything about the social world. For example, if our scheme of the basic psychological characteristics included "an impulse towards reciprocity", then atomism would be false -- because reciprocity implies the existence of social relationships. But there are other relationships besides "logical" that might be invoked here: semantic or causal, for example. Are basic individual characteristics definable in terms that do not analytically (semantically) presuppose some facts about the social world? And are basic individual characteristics in principle wholly caused through processes of individual psychological development, or do they unavoidably involve the causal role of social interaction and structure?

Their explication of individualism and collectivism is noteworthy in a different sense: it is primarily framed around the question of the freedom or autonomy of the individual with regard to the social structures he or she inhabits. Individualism entails that individuals create / determine the social world through their actions; whereas collectivism entails that facts about social structure determine individual actions. This issue is intertwined in their discussion with the idea of explanatory primacy: is the micro-level primary with respect to the macro-level, is the macro-level primary with respect to the micro-level, or does causation flow in both directions? But these are really quite different questions. The philosophical question of "freedom of the actor" seems to have little real importance for the sociologist or the philosopher of social science; whereas the question of explanatory primary is indeed important for both.

Jackson and Pettit give some legitimacy to the idea of real causal efficacy for social structures with respect to social change (and individual behavior). Their preferred view of the explanatory force of a structural explanation of an event is what they refer to as the "program model". A structural factor contributes to the explanation of an event, not by identifying the proximate or instigating cause of the event, but by identifying the frame of circumstances in the context of which the event was likely to occur. Prolonged drought is a "program factor" for the occurrence of a forest fire, not because the drought caused the fire, but because it represented a structural circumstance in the context of which many different kinds of events could ignite the tinder. "The program explanation identifies a condition such that its realization is enough to ensure that there will be causes to produce the event explained: if not the actual causes, then some others" (119).

Jackson and Pettit concede that micro-level facts are "more fundamental" than macro-level facts; but they also give support to the position that Fodor called "relative explanatory autonomy" of higher-level structures in the social sciences, thought they do not refer to Fodor directly. "The third thing to say on the sider of collectivism is that the program model [their preferred interpretation] forces a break, not just with heuristic individualists, but also with those individualists who tolerate structural explanation but think that the micro-explanation of any social fact is always bound to be of more interest.... On the program model, structural explanation serves a different sort of interest from micro-explanation, as micro-explanation serves a different interest from detailed psychological aetiology, since it gives a different kind of information on the causal history of the event explained" (131). In other words, they support the idea that a structural explanation can be just as "interesting" and informative as a micro-explanation.

The program model works well enough for some kinds of social explanations -- for example, the outbreak of a race riot given the ambient racial disparities and patterns of police abuse. What is least satisfying about this article is the almost complete absence of an account of what "structural factors" look like in the social realm. The examples Jackson and Pettit give are almost always social patterns and correlations -- poverty causes delinquency, economic growth causes urbanization, etc. But there is no discussion of the idea of a social structure, a normative system, or an institution. And the causal mechanism linking micro to macro is also a simple one: "All such [social] facts seem to have at least this in common, that they obtain or largely obtain in virtue of the intentional attitudes -- the beliefs, desires, and the like -- of a number of people, and/or the effects of such attitudes: the actions which the attitudes occasion and the consequences of those actions" (97). This is largely an aggregative conception of the causation of "social facts" based on individual characteristics.

In particular, the analysis provided by Jackson and Pettit misses altogether the active causal powers that social structures can be said to possess. These are social features that are the object of study of organizational sociologists and institutional sociologists. And as researchers like Kathleen Thelen demonstrate (link), these kinds of social features link together into causal mechanisms constituting causal explanations in their own right. Surprisingly, their account is not very interested in "structural" explanation at all. For example, Southwest Airlines underwent a catastrophic failure during the winter holidays in 2022. (Zeynep Tufekci's article in the New York Times provides a valuable post mortem of the failure; link.) Southwest Airlines is a business corporation, an organization with leaders and individuals through whose functioning highly complex processes of coordination take place (assignments of aircraft and crews to airports at the right times and work schedules, management of investments over time). Southwest Airlines exists within a profit-driven environment in which quarterly revenue reports are highly consequential for the top executives. The coordination required of a large airline requires sophisticated computing support. But Tufekci reports that the software systems in use at Southwest were said to be woefully inadequate for several years, according to reports by insiders and labor unions, and the company delayed in upgrading these expensive systems. These are all facts about a social entity, a corporation in a specific environment, and the organizational features it possesses in virtue of which failure is likely. The result was a cascading series of thousands of flight cancellations in December that may cost the airline a billion dollars. Here we have a mechanism-based account of a major social failure that is much more complex and realistic than the context-and-influence model implied by Jackson and Pettit's arguments.

In the end, Jackson and Pettit draw a position that refrains from both methodological individualism and methodological holism, in favor of what we might call methodological pluralism:

The program model of structural explanation is not only a satisfactory account of how such explanation works. It also gives us a nice perspective on the debate between individualists and collectivists. It means that we can embrace the persuasive individualist claim that individuals are agent-autonomous. But it also allows us to understand the collectivist thesis that individuals often make little difference in the course of history and that the best way to study society is often from the top down, not from the bottom up. Those claims constitute the true and attractive core of collectivism. (131)

The conception of methodological localism that I have tried to develop over recent years avoids both atomism and individualism, in the distinctions clarified by Jackson and Pettit. Methodological localism represents the view that structures, institutions, normative schemes, and other social entities are created and embodied by social actors interacting with each other. So the structures of the social world depend on the actions and thoughts of existing individuals. But existing individuals are themselves socially enveloped. Methodological localism proposes a view of the social world in which individuals are "social constituted" and "socially situated"; so from the start, there is no hint of the idea of a pre-social individual. This entails that the view is not atomistic. But likewise, the view is "actor-centered", which implies that individuals act according to their own intentional schemes. These schemes are socially influenced, to be sure, but they are also heterogeneous and diverse; and my premise is that individuals are not determined by the ambient social values, cultures, identities, and institutions in which they live. So the view is not "collectivist" in the strong sense described by Jackson and Pettit. And, finally, the view argues that social arrangements have durable causal powers, both with respect to individuals and to other social arrangements, and that these powers do not need to be reduced to facts about individuals. This is the view of relative explanatory autonomy that Fodor introduced, and that seems very compelling in the context of the micro-macro debate.

So methodological localism is not atomistic; it supports an element of holism, in the limited sense that it asserts that social actors are formed and framed within specific social arrangements and institutions; it is not individualist, in that it does not insist that social outcomes must be explained solely through derivation from facts about individuals; it is individualist in another sense, in that it attributed "freedom" to the social actors; it is not collectivist, in the sense that it denies that social arrangements "determine" individual action; but it is sympathetic to one aspect of the collectivist view, the idea that social structures exercise real causal influence over individuals and other structures. Rather than endorsing a simple relationship between micro and macro, methodological localism posits an iterative relationship over time between "socially constituted individual" and "social structure", a view that has deep parallels with Margaret Archer's conception of morphogenesis (link).