In contrast to van Fraassen’s interpretation, Suppe and Giere’s versions of the se- mantic view suggest that models and theories represent (or have a chance to repre- sent) the way the world works, or the connecting principles of nature and society (see Chapter 5). But we have seen that if we conceive their general claims as laws that hold ceteris paribus, it would be difficult to conceive models of unintended social phenomena as meaningful models. Giere’s notion of similarity, on the other hand, is not well defined, and it does not tell us much about how models explain. An alternative suggestion is that ‘the laws that hold ceteris paribus’, or the results we obtain by way of studying the behaviour of models worlds show us how cer- tain tendencies and capacities are realised in the model world. Or, they inform us about those tendencies and capacities that exist independently of the assumptions of the model. Cartwright (1999) presents a recent defence of this view.
Cartwright conceives models as ‘nomological machines’.38 She argues that the relations that hold in the model only hold under certain conditions (i.e. model’s assumptions), ‘they obtain just when a nomological machine is at work’ (Cart- wright 1999: 25).
Models in economics do not usually begin from a set of fundamental regu- larities from which some further regularity to be explained can be deduced as a special case. Rather they are more appropriately represented as a design for a socio-economic machine which, if implemented, should give rise to the behaviour to be explained.
(Cartwright 1999: 139)
The generalisations (i.e. ‘laws’) that are employed by the model (e.g. every individual would prefer a less-costly to store object), or the ones deduced from the model (e.g. actions of the self-interested economising individuals bring about a medium of exchange) are only valid in the model world. Yet she proposes these ‘laws’ indicate tendencies, or capacities, that work in the real world.39 Particularly, she suggests that the models of economics ‘provide us with a set of components and their arrangement. The theory tells us how capacities are exercised together’ (Cartwright 1999: 53).40 For example, concerning game theory, she argues:
In game theory various concepts of equilibrium describe what is supposed to happen when the capacities of different agents are all deployed at once.
(Cartwright 1999: 55)
Cartwright (1999: 57) argues that to build a theory one needs ‘parts described by special concepts’ and a ‘special arrangement’. Moreover, she suggests that nomological machines (i.e. models) need ‘shielding’ to work, coming close to our argument that models isolate. But do Cartwright’s suggestions answer our main concern about the relation of abstract model worlds to the real world? The suggestion is that in virtue of certain similarities between the model world and the real world, models depict certain tendencies in the real world.
Cartwright (1983, 1989, 1999) suggests that models inform us about certain tendencies, or capacities that exist in the real world even if the assumptions of the model do not hold. To see how models of unintended social phenomena may alert us to real tendencies, we have to realise that there is a two-way relation between them and the real world. To be able to point out real tendencies, models have to utilise some other real tendencies. Models depict or represent certain tendencies (e.g. individual mechanisms) in isolation from other factors, and then suggest possible ways in which these individual tendencies may interact. Considering the (process) models we have examined above, we may say that models alert us to possible aggregate tendencies. For this reason, Cartwright’s suggestion is correct, but somewhat incomplete: Models are based on our knowledge of certain tenden- cies or capacities. For example, we know that people have discriminatory pref- erences, or that individuals try to economise or decrease their transaction costs when possible. Models of unintended social phenomena portray these tendencies in isolation from the factors that may prevent them from being realised, and then demonstrate the ways in which they may interact under certain conditions. They point out possible ‘aggregate’ tendencies: a possible way in which those indi- vidual mechanisms (i.e. tendencies) may interact in bringing about a certain social phenomenon. It is because of our knowledge of, and familiarity with, these ten- dencies, or individual mechanisms, that we consider the model world to be similar to the real world. For this reason, we may argue that models of unintended social phenomena alert us to a certain way in which these mechanisms may interact in the real world.41
Individual mechanisms that are embodied in a model need not necessarily be common sense elements or things we know from a previous body of (scientific) thought about the subject matter. But good examples of invisible-hand models have this property. The novelty of these models come from showing how these individual mechanisms may interact. And in the best cases (e.g. the chequerboard model) the type of interaction (aggregate mechanism) suggested by the model is a novel one, that is, it is a previously unnoticed (and maybe counterintuitive) type of aggregate mechanism. The way the model connects with the world and with our cognition is the familiarity of the individual mechanisms, and with the help of this familiarity the model may convince us that a previously unnoticed type of interaction among these individual mechanisms is possible, and that this aggre- gate mechanism may be responsible from the phenomenon under investigation. Using familiar elements do not at all prevent novelty. By creatively conjecturing about the way in which those familiar things (e.g. economising action, different goods with different saleability, etc.) may interact in bringing about a certain phe- nomenon, one may find out new (previously unnoticed) aggregate mechanisms. For example, no-one would disagree if you say that ‘some people have mild dis- criminatory preferences’, but one may disagree if you say ‘mild discriminatory preferences may cause residential segregation’. However, if you show how those mild discriminatory preferences are connected to the aggregate phenomenon of segregation in the model world, then you have demonstrated some of the ‘con- necting principles’ and a possible aggregate mechanism that may explain particu- lar cases of residential segregation.
Knowledge of the existence of individual mechanisms that are isolated in the model is an important factor for assessing the similarity between these models and the part of the real world they represent. Yet we cannot accept these models merely because they represent certain tendencies that we know about. The speculative element in the model, the suggestion that those mechanisms may interact in a certain manner, necessitates that certain other external (i.e. external to the model) constraints hold for us to consider their conjectures plausible. As every model is unique, we can only cite some general issues here. The first thing that comes to mind is that if the model contradicts already known facts about the phenomenon, it would be hard to accept. That is, the consistency of the model or the explana- tion with what we already know about the real world is an important criterion. If the model is consistent and coherent with what we already know about the phe- nomenon (e.g. segregation) and about related phenomena (e.g. social psychology) then it is easier to conceive the possibilities that may be generated in the model world as possibilities of the real world. Moreover, complementarity with other available accounts of the same phenomenon is also important and preferable. A partial potential explanation (which is based on the model) that contradicts avail- able accepted explanations would have a limited chance of survival. The reader may object that the requirements of complementarity and coherence are not ‘real’ or firm criteria for evaluating a model. This is true. But the point here is that if a model or explanation that is partial and suggests certain possibilities departs radically from what we know, we would have a hard time fitting it to our world picture. Moreover, as we have seen, models of unintended social phenomena do not generally challenge the previously accepted causal mechanisms as explana- tory factors. They should not be interpreted as rejecting the idea that other mecha-nisms may be at work. Unless one explains a particular case by pointing out that the proposed mechanisms are indeed working in the suggested way, it better not contradict most of what we know about the world. Thus, logical plausibility,42 co- herence and consistency with what we know determine more or less the strength of our beliefs in the suggested possibilities. By virtue of presenting what we know in a novel way, some models of unintended social phenomena suggest new ways to look at the world. Some models, on the other hand, test other models in terms of logical plausibility, coherence and consistency with the facts of the real world.
Nevertheless, it is not possible to define a priori the amount of ‘acceptable real- isticness’ for a model. Concerning theoretical models of the sort we are interested in, the only real ‘hard’ criterion seems to be the success in explaining particular cases. Explanations of unintended social phenomena are rarely singular explana- tions, and their claims should remain claims about what may be possible in the real world, how certain tendencies may possibly interact, etc. The similarity be- tween the model world and the real world is important for it constrains the range of possibilities generated by the model, but we cannot predetermine a degree of acceptable realisticness. This, however, should not prevent us from testing the plausibility of our hypotheses further.
Source: Aydinonat N. Emrah (2008), The Invisible Hand in Economics: How Economists Explain Unintended Social Consequences, Routledge; 1st edition.