20th WCP: The Not-so-trivial Truth of Methodological Individualism

The principle of methodological individualism in the social sciences has its origin in the Austrian school of economics and was introduced into the philosophy of social science in general by Friedrich Hayek and Karl Popper. Hayek was the first to use the term for the view that, in his words, "the concepts and views held by individuals [...] form the elements from which we must build up, as it were, the more complex [social] phenomena" (1942/44, p. 38). Popper used the term for "the quite unassailable doctrine that we must try to understand all collective phenomena as due to the actions, interactions, aims, hopes, and thoughts of individual men, and as due to traditions created and preserved by individual men" (1944/45, pp. 157-158). The issue of the principle's truth was hotly debated in the 1950s, thanks especially to its champion John Watkins, a student of Popper. Since then, arguments defending or attacking methodological individualism occasionally appear in the literature, but the contestants seem to hold on to their initial lines of trenches. Believers express difficulty in understanding how anyone could fail to see that the principle of methodological individualism is correct. Unbelievers aim to show that there always remain 'inherently social' or 'irreducibly social' aspects that a methodological individualist will find impossible to account for in terms of the principle.

The title of my paper reveals my own position. I am a believer. I think methodological individualism is indeed correct, in a strong sense that makes it go beyond a strictly methodological principle. By this I mean that I subscribe to both explanatory individualism and conceptual individualism, as these different forms are distinguished by Tuomela (1984, 1995), although similar distinctions were already introduced during the 1950s debate by Brodbeck (1958) and by Goldstein (1958). I therefore think that the 'irreducibly social' elements brought forward by the unbelievers can be shown to be reducible after all, simply by trying harder than was done by those who claimed they could not. This, however, I will not elaborate in this paper. The point that concerns me here is that it seems to me that the defenders of methodological individualism have not argued for their position as forcefully as they could have done, and have failed to clarify the special nature of social entities to the extent that I think is necessary to show the inescapability of methodological individualism.

Let me illustrate this. Jon Elster has always been a staunch defender of methodological individualism. In his book Nuts and bolts for the social sciences (1989) he calls it 'trivially true'. Elsewhere, however, he likens the relation between the individual and society to the relation between a cell and its composing molecules. This, I think, is badly mistaken. The relation between a social entity and the individuals 'involved' in it is not a part-whole relation as is the relation between a cell and its composing molecules. But the social entities that figure in the description of social phenomena - entities that are generally grouped together under the name of institutions or institutional arrangements - are not mereological sums of individuals (barring some borderline cases such as mobs, queues and crowds). A number of arguments to show this have been given by Ruben in his book The metaphysics of the social world (1986): spatial properties pertaining to individuals do not carry over to the social entities these individuals are involved in, nor is the 'involvement' or 'membership' relation transitive, as should be the case if social entities were mereological sums. But Ruben himself is an unbeliever. He thinks there are irreducibly social entities. Nevertheless, as I will argue, Ruben's insight into the relation between human individuals and social entities is an important step toward recognizing why methodological individualism must be true.

For a start, it seems to me it has to be acknowledged that social science as we conceive it has to adopt the intentional stance. All earlier defenders of methodological individualism have supposed that we must have recourse to such mental entities as beliefs, aims, desires, expectations, etc., in order to analyze social wholes and explain social phenomena. Given that behaviorism is generally thought to have failed even as a purely psychological approach, it is difficult to see how you can succeed as a social scientist if you prefer to avoid all reference to mental terms and to use a behavioristic idiom. The point is not that in that case there is no separation of psychology from sociology - I will argue that this separation as it is usually seen is illusory anyway - but that it seems a hopeless task, from that perspective, to give meaning to the social concepts that figure in human speech in purely nonmental terms. There would be no more meaning in words as 'state' or 'bank' than there is in the words lark, quark or snark. They must remain merely forms of verbal behavior, sound patterns uttered by human individuals in particular circumstances. And the task of teaching us (regarding whom, to be sure, the mentalistic stance is equally mistaken) more on these particular circumstances must fall to the only possible kind of social science, that is, sociobiology.

Notwithstanding its prima facie implausibility, this viewpoint that the scientific inquiry into human behavior, in isolation just as much as in the case of mutual interaction among many individuals, should proceed in complete disregard of the intentional idiom is vehemently defended by Alexander Rosenberg. According to Rosenberg, there simply are no laws that link intentional concepts; sociobiology "employs the narrowest natural kinds under which human behavior falls" (1980, p. 7). Thus there is no room for a science of human affairs that is in between on the one hand sensory and neurophysiology and on the other evolutionary population dynamics. However, this approach has been shown a dead-end street by the detailed critique of some of its by most cherished explanations concerning various phenomena studies in social and cultural anthropology, especially Kitcher (1985). And from a more recent article by Thornhill (1991) it is evident that the hypotheses that match the ethnological data are just the ones that are phrased in the intentional idiom, presenting specific mental states as proximate causes. Mental states, moreover, that are directed to securing wealth and power rather than the reproduction of the subject's own genes.

By claiming that social science, from the methodological-individualist point of view, is committed to the intentional stance, I do not at the same time commit myself to a particular position concerning the reducibility or irreducibility of mental states to brain states. This is simply a problem that methodological individualism need not tackle. It must perhaps hold that the eliminativism defended by Stich (1983) and the Churchlands (1981, 1986) is too dismissive concerning the validity of ascribing propositional attitudes to people (as has indeed been forcefully argued by others), but for the time being there is need to choose between the instrumentalist interpretation adopted by Dennett (1987) or the realist position argued for by, for instance, Bishop (1989). But equally rash would be a total dismissal of sociobiology. To explain how the human brain came to function such as it seems to do according the intentional stance is a challenge that one hopes precisely sociobiology can meet.

Nevertheless, it seems to me that a recognition of some intimate relation between mental states and brain states must accompany our adoption of the intentional stance toward the behavior of individual people. No serious social scientist, including Marx and Durkheim, ever held that one could ascribe actions to social entities or 'wholes' as being caused by supra-individual mental states. How one should account for the 'behavior' of social wholes in the description and explanation of social phenomena, or for their nature as entities in the first place, still leaves much to be desired. It is generally agreed that all we observe in the social world are the actions of individual people (barring perhaps the above-mentioned borderline cases of pure aggregates such as crowds and mobs). Social wholes therefore must have a theoretical status. But there must be a difference with the infrahuman theoretical entities such as fields and elementary particles that physics presents us with. Reichenbach (1938) has introduced a distinction between illata and abstracta, which applies exactly here. Illata are entities the existence of which is inferred to explain our sensory experiences; the relation between our sensory evidence and the existence of a particular illatum is probabilistic. Abstracta, on the other hand, are logical constructions of our own making; to settle the question of their existence is a matter of convention.

It is contained in methodological individualism that social entities or wholes are Reichenbachian abstracta. They are open-ended concepts; whether they apply is a matter of judgement, not of truth. They are not natural kinds. They are abbreviations for, mostly, very complex constellations of the corresponding actions and mental states (beliefs, expectations, desires) of many individual people. This is where Elster's above-mentioned mistake comes in. But he is not the only one. It is wrong, as for instance Quinton (1975) has done, to compare social wholes as 'logical constructions' to forest and mountain ranges. Although it remains a difficult question to say when exactly a number of trees become a forest, forests and mountain ranges are objects. It is possible to take a few steps back and observe the forest. Forests and mountain ranges are natural kinds that figure in typical lawlike connections: they block sunlight, they form part of ecosystems, they divert and cool air streams. The relation between a forest and its composing trees is a part-whole relation. None of this is true for social entities and the relation they have with the individuals 'involved' in them or 'participating' in them. We can not take a few steps back and look at society as such, or at a country or a bank or a marriage. There is no higher level from which we can observe institutions as wholes. Nor is the relation between these social wholes and the participating individuals a part-whole relation, as Ruben has emphasized, and accordingly, the individual and social realms of description are not related as different levels comparable to the way the level of organisms is related to the level of cells, or the level of cells is related to the level of molecules. All we can observe in the social world is the behavior of individual people. Through the intentional framework, without which social science can hardly come off the ground, we interpret this behavior as action - that is, as being based on specific beliefs and desires, chief among which are beliefs and desires about the beliefs and desires and corresponding future actions of other people. We simply lack independent measurement procedures for collective concepts. That is, we lack measurement procedures that establish whether or not a group or an institution or any other social 'whole' has a particular property independent of the properties - actions, beliefs, desires - of the relevant individuals.

It might perhaps be argued that one can take a 'Martian point of view' and create a truly 'scientistic' social science - that is, one that deserves the predicate scientism much more than the Hegelian and Comtean examples that Hayek (1942/44) attacked. Suppose that Martians, by carefully monitoring the radio signals they receive from the earth, will be able to calculate certain correlations between specific radio signals emerging from different places on earth at different times and thereby predict patterns in these signals - one such signal being, say, the closing rates of the New York stock exchange and the other signal the closing rates of the Tokyo stock exchange. Or they could measure the correlations between the traffic densities (i.e. vehicle densities) at various places and times and predict from these the moment of the next congestion at a certain spot. However, these predictions would not be a part of any Martian social science about human life on earth; it would just be one of the very many correlations produced by Martian natural science about that strange planet Earth. And correspondingly these Martians would have no clues about how to proceed, that is, where to look for interesting correlations. Any relation might be worth investigating. But as soon as the Martians would hypothesize the existence of intelligent life (a form of life to which the intentional stance is applicable) in order to explain the generation of some of these signals, they would see the sense of a social science, but that science would then be individualistic just as ours is.

The accompanying two pictures illustrate the main points made so far. The whole of the physical world can be described in terms of levels, that are connected by transitive part-whole and space-time relations. For each of these levels it is so that the properties of entities at lower levels explain the behavior of entities at higher levels, but, on the other hand, equally so that for the identification of higher-level entities and their properties no awareness of the existence of lower levels or access to lower-level entities is essential. A plausible candidate for an entity defining a higher 'supra-organismic' level answering to this scheme is the earth as an ecosystem. Social entities, on the other hand, are not 'one level up': they are connected to humans-under-intentional-description in a completely different way. For one thing, they can be identified and investigated only through the individual humans that 'make them up'.

Social scientists have often fought methodological individualism because it implied, in their eyes, an unacceptable reduction of social science to psychology, or the subsumption of social phenomena under psychological theories. Indeed this is how Watkins formulated it in his first characterization of the principle, claiming that "the social scientist can continue searching for explanations of a social phenomenon until he has reduced it to psychological terms" (1952, pp. 28-29). And superficially, if sociology is seen as the science of collective human behavior and psychology as the science of individual human behavior, this might seem correct. However, a closer look reveals that this interpretation of the relation between social science and psychology is seriously mistaken.

On the one hand, when the reduction of a particular branch of science to another one is spoken of, it is supposed that the reduced branch of science existed up till the achievement of the reduction as an autonomous set of laws and regularities, formulated in a vocabulary of its own and each capable of empirical validation (to a certain extent, not absolutely) without any knowledge of the reducing branch of science being a necessary prerequisite for handling it. Examples of such branches of science in-their-own-right, and of their later reductions (or proposed reductions, as all of these examples are still controversial), are chemistry reduced to quantum physics, cytology and physiology reduced to biochemistry and molecular biology, and Mendelian genetics reduced to molecular genetics. As I have argued above, the relation between the social and individual aspects of human existence are not related as different branches of science or as (relatively) macro- and micro-levels similar to the relations between reduced and reducing branches of science or theories in the examples just mentioned. The 'laws' or regularities that link various social wholes or properties of social wholes do not together form an autonomous branch of science, and the explanation of any of these social regularities cannot be considered as a move ahead in the reduction of social science to 'the science of the individual human being'.

On the other hand, it is hardly possible to look upon psychology in its current state as this 'science of the individual human being'. Current psychology is an enormously diverse field of inquiry, covering both sensory physiology and social psychology and pretty much everything in between. In view of the fact that within psychology there is raging a lively debate on the reducibility of mental states to neurophysiological states, any claim that the social sciences were reducible to psychology would be devoid of meaning.

In order to render the relationship of social science to psychology more precise, one should take the intrapsychological debate just mentioned to heart and cut the discipline of psychology in two. On one side of this cleavage fall the approaches whose basic framework is the intentional idiom of so-called folk psychology. This side is occupied largely by social psychology. Here it is assumed that human beings are equipped with beliefs and desires - beliefs both about the behavior of the natural world and about the behavior of their fellows - and that they act upon these beliefs and desires. On the opposing side of the cleavage are the approaches that investigate how humans do the things that folk psychology assumes they do. These approaches form together what is today generally called cognitive psychology or cognitive science. It studies how the human mind does what it does. How the issue of the relation between the two sides along the cleavage is resolved - whether this relation is one of reduction or not - is not of great importance concerning the relation here at issue: the relation between social science and psychology. Opinions on whether or not there is a type-type identity relation between brain states and mental or intentional states may differ, but they must converge on the need to do justice to the fact that it certainly looks as if human beings have beliefs and that it looks as if these can explain their actions.

The relation between cognitive psychology and intentional psychology is then indeed one of different levels. But as I hope to have made abundantly clear by now, my claim is that intentional psychology and social science form in fact one single scientific realm, and that there is no justifiable distinction to be made between the two, certainly not one in terms of different levels. This realm is the science of what human beings do, viewed from the intentional stance. There is no special place in this realm for what human beings do 'in social conditions' or 'when together with other people'. Any human being is permanently aware of the existence of other people; his or her set of beliefs is fully shaped by the fact of the existence of other people. A non-social human being does not exist; it would have to be produced artificially.

Thus in a sense it is true to say that intentional psychology or 'folk psychology' - the framework of beliefs, desires and actions - is the underlying theory of social science, but this is something completely different from the reducibility of social science to intentional psychology. The role of intentional psychology toward social science should be compared rather to the role of classical atomic theory, or corpuscular theory, to Newtonian mechanics. The framework of small mass-bearing particles - idealized to mass points - is the world picture assumed by Newtonian mechanics; it would be nonsensical to say that Newtonian mechanics reduces physics to atomic theory. The classical mechanistic world picture can rather be said to unify the application of the theory of mechanics to the world. Similarly the intentional stance functions as a unifying framework to social reality. Whether there is a single theory of action corresponding to mechanics - some form of rational-choice theory being the most frequently proposed candidate - is quite another matter, which I must leave for a different occasion.