There are two prominent philosophical accounts of theories: the syntactic view and the semantic view. The syntactic view rests on an assumption of well-es- tablished theories that can be presented as axiomatic systems, and emphasises the linguistic and logical structure of theories. The syntactic view does not pay attention to the role of (partial) models in science.28 Thus, it cannot be used as an account of the models of unintended social phenomena.29 The alternative view, the semantic conception of theories,30 emphasises the role of models in science.31 It holds that theories are basically collections of models. The semantic view of theories has three distinct interpretations that differ in the way in which they char- acterise the relation between the models (or theory) and the real system. The first interpretation considers models as isomorphisms, that is, as being isomorphic to the particular part of the real world they represent. In this interpretation there is a one-to-one mapping between the theory and the observable part of the real world it represents (e.g. van Fraassen 1980). The second interpretation suggests that models describe an idealised and abstracted version of the portion of the real world they represent, and hence they are not necessarily isomorphic (e.g. Suppe 1989). The third interpretation conceives the relation between models and the real world in a more flexible manner, in terms of similarity (e.g. Giere 1988a).
According to van Fraassen,
To present a theory is to specify a family of structures, its models, and sec- ondly, to specify certain parts of those models (the empirical substructures) as candidates for the direct representation of observable phenomena. The structures which can be described in experimental and measurement reports we can call appearances: the theory is empirically adequate if it has some model such that all appearances are isomorphic to empirical substructures of that model.
(van Fraassen 1980: 64, last emphasis added)
Van Fraassen holds that if a theory is isomorphic to the observable part of the real world it represents (i.e. empirically adequate), then we may believe in it for it helps us to hold a coherent view of what we have observed.32 In the previous section we have seen that the condition of empirical adequacy (i.e. isomorphism) does not hold. The empirical adequacy condition is too strong and cannot be ex- pected to hold in all models.33 This interpretation considers adequate models as mere descriptions of the real world. This is not the case for many scientific mod- els.34 Other proponents of the semantic view criticise van Fraassen’s interpretation for similar reasons. Particularly, Suppe (1989: 102) argues that the ‘process of abstraction carries no guarantee that any of the theory’s models or substructures will be isomorphic to any actual phenomenal systems; hence there is no guarantee that there will be any models such that all appearances are isomorphic to empirical substructures of the model.’ Or, Giere (1988a, 2000) suggests that isomorphism is a strong relation and usually a weaker relation (e.g. similarity) holds between theories (or models) and real systems.
Suppe’s (1989) alternative is to consider models as descriptions of an abstractly conceived world (i.e. physical systems). He characterises scientific theories as descriptions of ‘the behaviour of physical systems, which are idealised replicas of actual phenomena’ (Suppe 1989: 67). Under this interpretation, a theory is a col- lection (or a cluster) of models, which characterises possible ‘physical systems’ (i.e. possible model worlds). Theories, in this sense, are about the behaviour of abstract systems, ‘what the theory does is to directly describe the behaviour of abstract systems, known as physical systems, whose behaviours depend only on the selected parameters’ (Suppe 1989: 83). For this reason, theories and models suggest how things could be in the real system they represent:
the theory does not characterise the actual phenomena, but rather character- ises the contribution of the selected parameters to the actual phenomena, de- scribing what the phenomena would have been had the abstracted parameters been the only parameters influencing them.
(Suppe 1989: 82–83)
Suppe’s conception of theories suggests that laws only hold ceteris paribus.35 The argument is that if the assumptions of a certain theoretical model hold in the real world then the results derived from the model would hold in the real world. The statement of a model (or a theory) under this conception is that if such and such were the case in the real world such and such results would hold. Yet we have seen that some of the assumptions of the models of unintended social phenomena do not hold in the real world. For example, consider the Kiyotaki–Wright (1989) model. Under this version of the semantic conception, the Kiyotaki–Wright model is supposed to say that if there were three types of rational agents and three differ- ent commodities with different storage costs (and if certain other conditions hold) then money would emerge as an unintended consequence of human action in the real world. Or consider the chequerboard model: if there were two types of agents that can recognise each other’s types, if they all had similar preferences regarding the other type, and if no other factors were influencing their preferences (and if the other assumptions about neighbourhoods hold), then segregation would emerge in a real city (as an unintended consequence).36 We would be doing injustice to the authors of these models if we were to accept this interpretation. The models of unintended social phenomena have more to say. They suggest that the proposed mechanisms may be working in the real world even if the assumptions of these models do not strictly hold in the real world.
A more flexible version of the semantic view suggests that models may tell us something about the real-world systems in virtue of certain similarities between the proposed models and the real-world systems they represent. We have seen that the semantic conception conceives theories as descriptions of abstract systems, thus, they are true about the systems they represent, and they do not claim some- thing about the real world as such. But theories and models make claims about the real world with a theoretical hypothesis (Giere 1988a, 2000). They suggest something about the world through a hypothesis about the real world. But how can we say that the model shows that the hypothesis is defensible, that is, that it holds in the real world? Giere’s argument is that the similarity between the model and the real system suggests that the theoretical hypothesis may hold for the real system. Giere’s account of the relation between models and theories seems to be the best fit among the alternatives we have discussed above. However, there is still an important unresolved issue in his account: similarity is a vague concept and it is not clear what it amounts to in Giere’s account.
Source: Aydinonat N. Emrah (2008), The Invisible Hand in Economics: How Economists Explain Unintended Social Consequences, Routledge; 1st edition.