How Models Fail
A Critical Look at the History of Computer Simulations of the Evolution of Cooperation

Eckhart Arnold

1 Introduction
2 The empirical failure of simulations of the evolution of cooperation
3 Justificatory narratives
    3.1 Axelrod's narrative
    3.2 Schüßler's narrative
    3.3 The story of “slip stream altruism”
    3.4 The social learning strategies tournament
4 Bad excuses for bad methods and why they are wrong
5 History repeats itself: Comparison with similar criticisms of naturalistic or scientistic approaches

3.1 Axelrod's narrative

Axelrod motivated the use of the Prisoner's Dilemma mostly by the fact that it already was an extremely popular game theoretical model that had already been used in experimental economic research. He compares the Prisoner's Dilemma to the E.coli in biological research. Comparisons to the ever successful natural sciences are quite typical for the justificatory discourse of the modeling approaches in the social sciences. With the benefit of hindsight it can, however, be said that this comparison was slightly misleading. E.coli is a great object of study in biology, because what one learns when studying E.coli can often directly be transferred to other bacteria. Many bacteria are similar to E.coli in important respects. The same is unfortunately not true for the RPD model, which is not at all a robust model (Arnold 2013b, 127f.). Change the parameters of the simulation, the initial set of participating strategies or other aspects of the model only a bit and you can get qualitatively different results. Most likely, another strategy than TIT FOR TAT would turn out as winner, and maybe not even a friendly or cooperative strategy (Binmore 1994, 315).

One part of Axelrod's motivation is also a supposed advantage of the simulation approach to experimental approaches. Axelrod relates to the notorious problem of economic experimental research that the laboratory setting is usually highly artificial and that, therefore, any obtained results cannot easily be transferred to real life situations. He omits to mention, however, that computer simulations based on highly stylized models like the RPD share the same problem.

t g+ f @