Tools or Toys?
On Specific Challenges for Modeling and the Epistemology of Models in the Social Sciences

Eckhart Arnold

1 Introduction
2 The role of models in science
3 Why computer simulations are merely models and not experiments
4 The epistemology of simulations at work: How simulations are used to study chemical reactions in the ribosome
5 How do models explain in the social sciences?
6 Common obstacles for modeling in the social sciences
7 Conclusions
    7.1 Consequences for modellers in the social sciences
        7.1.1 Consequences for problem orientated research
        7.1.2 Conseqeunces for method centered research
    7.2 Consequences for philosophers of science

7.1.2 Conseqeunces for method centered research

The method centered approach can be understood as a research strategy where a certain methodology is developed and investigated with regards to its ability to answer different research questions. If a scientist follows a method centered research approach then the method is fixed and the problems are chosen or disposed of according to their suitability for applying the method. The conclusions that can be drawn for method centered research are symmetric to those for problem orientated research:

  1. Chose the right problems for your method, make sure that relevant scientific problems for the method exist: The “right problems” are problems where the success of the models can be tested. A common danger of method-centered research is the irrelevancy of its results (Shapiro 2005). This happens, if problems are chosen only because they fit the method and not because they are relevant problems in any other sense.
  2. Keep in mind that the model needs to be validated: Models and simulations should be designed so that they can be validated. This implies that free parameters should be avoided and measurement inaccuracies should be taken into account. The burden of attuning models to measurement restrictions clearly rests on the shoulders of the modelers and not of the empirical researchers that develop measurement techniques, because the possibilities for developing measurement are limited by the empirical world.
  3. Validate your model, take failures seriously: Models need validation. It is insufficient to base a model - as is often done (see Hegselmann/Krause (2002) for an example) - merely on “plausible assumptions” without either systematically testing the validity of these assumptions nor empirically validating the results. A model that has not been validated does at best have the epistemological strength of a metaphor or a just-so story. Admittedly, this may be sufficient in certain contexts.

    Failures of validation ought to be taken serious: A model that fails validation is a false model. A model that cannot even be validated should be considered as not yet a scientific model in the same sense as an unfalsiafiable theory is considered as unscientific.

t g+ f @