© Copyright JASSS

JASSS logo ------

Simulation and Similarity: Using Models to Understand the World (Oxford Studies in Philosophy of Science)

Weisberg, Michael
Oxford University Press: Oxford, 2012
ISBN 978-0199933662 (pb)

Order this book

Reviewed by Andreas Koch
University of Salzburg

Cover of book Modelling and simulation as a distinct way of doing research gain increased attention not only in those scientific communities who are applying models in their disciplinary contexts but also in those communities who are investigating models’ theoretical, methodological, and epistemological traits – thus the meta-modelling level (see, for example, Christie et al. (2011), Mitchell (2009)). Michael Weisberg with his book Simulation and Similarity makes no exception in this respect. As a philosopher of science he is interested in the value of models (i.) to progress in reasoning, (ii.) to represent adequately the subject matter (the target system in his words), (iii.) to map the actual or imagined original from which a model is being derived, and (iv.) to yield interpretative profit. In fact, Weisberg explicitly takes the modeller in his/her role as theorist into consideration. Recent literature, he criticises, “has not fully explored the role of theorists’ intentions in all aspects of modeling, including the individuation of models, the coordination of models to real-world systems, and the evaluation of the goodness of fit between models and the world”.

Based on this assessment he structures the book accordingly. First he distinguishes three basic types of modelling and models – concrete, mathematical, and computational modelling (Chapter 2). Even one can imagine different modes of structuring model types these three types are to be understood as epistemologically different in terms of model intention, procedure, relation to the target system, and philosophy. This is followed by a definition of models as “interpreted structures” whose consequences are delineated in Chapter 3. Next he criticises the implementation of fiction into the scientific process of reasoning about modelling and model result interpretation (see also Derman (2011), Morrison (2015), p. 85ff), because of “inter-scientist variation, the limited representational capacity of fictions, the inability … to account for the practice of modelling, and variation in the face-value” (p. 56). The remaining chapters follow a pattern of relaxation by starting with the narrow concept of target-directed modelling, through idealisation to ideas and concepts about modelling without a specific target. This ultimately leads Weisberg to a discussion about the notion and meaning of “similarity” between the model and the target (the world).

A first account must be made of the title: it is only to a minor degree about simulation, but much more about models and modelling. The alliteration seems to address a connection between the two words that goes beyond the semantic beauty, but it is the connection between model and target as well as modelling and modeller that Weisberg is focusing on in fact.

The model-target and modelling-modeller relations are actually the two core accounts that leave the reader both convinced and puzzled. Weisberg, on the one hand, explicitly appreciates the theorists’ and modellers’ intentions in the process of modelling as an inevitable subjectivization (individuation) of models and as an intrinsic motivation in the process of scientific progress. The influence of a constructivist approach can be clearly recognised. This leads to an open-minded debate about the diversity of models, their utility, purposes, aims, and contextualisation. A good example for this assessment is given with Chapter 7 where he advocates modeling without a specific target, and urging the reader to think about generalized modeling, hypothetical modeling, and targetless modeling more thoroughly.

On the other hand, it remains an open question to me why he put so much emphasis on the target system as an absolutely different reference against models created by individuals or collectives. He admits that modellers never take the total state of a target system into consideration; indeed, they create their own, abstract target systems (p. 91). Consequently, “fitting models to the world does not depend on the total states of phenomena, but rather on abstracted target systems” (p. 93). This in turn means that both the model and the target system are constituted, constructed by the modeller, and both sides of the relation are demarcated artificially (i.e., not naturally predetermined). Therefore I think it is misleading to draw a distinction between model and target, as has been done throughout the book. To give an example at page 95: while “the structure of a mathematical model is a free choice of the theorist, … the mathematical representation of the target is a highly constrained representation of the way that part of the world is”. Insofar the modeller is free to create the subset of structures, properties, etc. of the total system then there are no absolute exogenous constraints the modeller is forced to take into account. In other words, the target itself is a model created by the researcher by setting the elements, relations, and boundaries of the target system subjectively.

This leads to at least two further conclusions. First, all types of idealization mentioned in Chapter 6 – the ‘Galilean idealization’ towards simplification, the ‘minimalist idealization’ towards core causality, and ‘multiple models idealization’ towards analytical subsets and synthetic relations thereafter – are manifestations of gaps between a model and its supposed target within the same do-main of models and modelling. Second, though the domain may be the same, model and target are not synonymous, however. I therefore disagree with the total conflation Weisberg outlined with respect to Schelling’s segregation model (on page 30 and page 118). He writes: “Say we find a pattern of segregation in a city that looks much like what Schelling predicted. If his model is an explanation of that pattern of segregation, then the model’s algorithm must be similar to what is happening in the city” (p. 30). In fact, the algorithm uncovers a fascinating mechanism of the relationship between individual motives and social behaviour; but it does so due to extreme simplifications which are necessary to detect the underlying mechanism in principle, but will never have anything in common with the empirical complexity of any city. The computational model is like a laboratory (a sterilised environment that helps focusing on the object of interest) and the target system might be a city, say Philadelphia, which has been artificially demarcated (and there are many boundaries that separate Philadelphia from its environment).

In conclusion, I think the book is worth reading, because it allows to critically rethinking concepts of models and modelling from a philosophical point of view. Weisberg has some good arguments to do exactly this.

* References

CHRISTIE, M., Cliffe, A., Dawid, P. and S. Senn (eds.) (2011). Simplicity, Complexity and Modelling. Wiley & Sons, Chichester.

DERMAN, E. (2011). Models.Behaving.Badly. free press, New York.

MITCHELL, M. (2009). Complexity. A Guided Tour. Oxford University Press, Oxford, New York.

MORRISON, M. (2015). Reconstructing Reality. Oxford University Press, Oxford, New York.


ButtonReturn to Contents of this issue