User:Hopewell Pinstripe/sandbox

A notable proponent of idealization in both the natural sciences and the social sciences was the economist Milton Friedman. In his view, the standard by which we should assess any empirical theory is the accuracy of the predictions that a theory makes. This amounts to an instrumentalist conception of science, including social science. Consistently with this conception, he then argues against the criticism that we should reject an empirical theory if we find that the assumptions of that theory are not realistic, in the sense of being imperfect descriptions of reality. This criticism is wrongheaded, Friedman claims, because the assumptions of any empirical theory are necessarily unrealistic, since such a theory must abstract from the particular details of each instance of the phenomenon that the theory seeks to explain. This leads him to the conclusion that “[t]ruly important and significant hypotheses will be found to have ‘assumptions’ that are wildly inaccurate descriptive representations of reality, and, in general, the more significant the theory, the more unrealistic the assumptions (in this sense).” In this light, he makes the case for seeing the assumptions of neoclassical positive economics as not importantly different from the idealizations that are employed in natural science, drawing a comparison between treating a falling body as if it were falling in a vacuum and viewing firms as if they were rational actors seeking to maximize expected returns.

Against this instrumentalist conception, which judges empirical theories on the basis of their predictive success, the social theorist Jon Elster has argued that an explanation in the social sciences is more convincing when it ‘opens the black box’ — that is to say, when the explanation specifies a chain of events leading from the independent variable to the dependent variable. The more detailed this chain, argues Elster, the less likely the explanation specifying that chain is neglecting a hidden variable that could account for both the independent variable and the dependent variable. Relatedly, he also contends that social-scientific explanations should be formulated in terms of causal mechanisms, which he defines as “frequently occurring and easily recognizable causal patterns that are triggered under generally unknown conditions or with indeterminate consequences.” All this informs Elster’s disagreement with rational-choice theory in general and Friedman in particular. On Elster’s analysis, Friedman is right to argue that criticizing the assumptions of an empirical theory as unrealistic is misguided, but he is mistaken to defend on this basis the value of rational-choice theory in social science (especially economics). Elster presents two reasons for why this is the case: first, because rational-choice theory does not illuminate “a mechanism that brings about non-intentionally the same outcome that a superrational agent could have calculated intentionally”, a mechanism “that would simulate rationality”; and second, because explanations drawing on rational-choice theory do not provide pinpoint predictions, which in certain cases (for instance, he claims, quantum mechanics) would be sufficient to convince one that the theory making these predictions is likely to be true. Accordingly, Elster wonders whether the as-if assumptions of rational-choice theory help explain any social or political phenomenon.

Michael Weisberg has examined these and related questions from the vantage of philosophical reflection on models and idealization. By his lights, we can develop a classification of scientific idealization that picks out three kinds: Galilean idealization, minimalist idealization, and multiple-models idealization. Galilean idealization, in his account, consists in “introducing distortions into models with the goal of simplifying, in order to make them more mathematically or computationally tractable”, whereas minimalist idealization “is the practice of constructing and studying models that include only the core causal factors which give rise to a phenomenon.” Somewhat similar to minimalist idealization is multiple-models idealization, which Weisberg defines as “the practice of building multiple related but incompatible models, each of which makes distinct claims about the nature and causal structure giving rise to a phenomenon.” Further, these kinds of idealization can be differentiated in terms of ‘representational ideals’, which Weisberg sees as “regulat[ing] which factors are to be included in models, set[ting] up the standards theorists use to evaluate their models, and guid[ing] the direction of theoretical inquiry.” Relevant to the debate between Friedman and Elster is the representational ideal of ‘maxout’, according to which the model-builder aims only at maximal predictive accuracy; only this ideal, Weisberg claims, “sanctions black-box models” Moreover, in his view, and contrary to what Friedman’s discussion of the law of falling bodies suggests, Galilean idealization has as its aim not ‘maxout’ but, rather, ‘completeness’ — viz., providing a complete description of a given phenomenon. Relatedly, Weisberg also finds unsatisfying the ideal of ‘maxout’  as a principle for guiding scientific research, inasmuch as this ideal counsels only the development of predictions, thereby neglecting what Weisberg sees as a central part of the scientific enterprise: “[w]hile scientists want to know how a system will behave in the future, they also want an explanation of why it behaves the way that it does.”

The philosopher Kwame Anthony Appiah has defended the value of as-if idealization more broadly, across the sciences as well as the humanities and for ends other than prediction. In short, he argues that such idealization can aid our understanding of a given phenomenon even when that idealization involves claims about that phenomenon that are false. In support of this contention he draws on the thought of Daniel Dennett, in particular his notion, elaborated in The Intentional Stance, that viewing a system as if it were an intentional agent can better our predictions of that system’s behaviour and, moreover, bring to our attention patterns in its behaviour that we would otherwise not notice. But Appiah goes further than this, contending that as-if idealization is an essential feature of several modes of thought. Here, his principal intellectual guide is Hans Vaihinger, whose philosophy he describes in the following way: “[his thought] regards questions concerning our everyday thinking about the world as continuous with our scientific thinking: [b]oth aim, he says, at controlling reality, and both can leave things out in order to make it practicable to represent the world we want to control.” To illustrate his own claims, Appiah describes how the schematic McCulloch-Pitts neuron yielded insights regarding neurophysiology as well as computer science: “a highly idealized model of the brain acquir[ed] independent utility because its simplifying idealizations ended up providing techniques for mimicking the functions rather than the material substrate of the mind.” With respect to social science in particular, Appiah analyses the conception of rationality within rational-choice theory and arrives at the conclusion that this conception assumes perfect computational ability — that is, the ability to process information without error — but is not for that reason useless or inapplicable to the study of human phenomena. In his words:

No actual agents are computationally perfect, but the states that determine their actual behavior can still be characterized by how they would manifest themselves, given computational perfection. Analogously, the actual velocities of real gas molecules, which explain their less-than-ideal actual behavior, may nevertheless be characterized as the velocities that would, if only gas molecules were perfectly inelastic point masses, produce the ideal gas laws predicted by the simplest version of the kinetic theory of gases. (pp.84-85)