The transformation of science through computer simulation is often considered to be a methodological one. A lot of literature has been dedicated to determining the relationship between computer simulation, experiments or theories as the classical sources of knowledge. This relation is both methodologically and technically complex. This minisymposium addresses these problems by offering two lectures on computer simulation methods.ABS - An epistemological analysis from study cases
Agent-Based Simulation (ABS) has become a common experimental methodology in fields like Sociology, Politics, Economics and Geography. Research in these areas investigate properties of dynamic systems characterised by their global behaviour being obtained by locally defined rules. This makes ABS greatly apt to both predict and understand data extracted from - and applied in - these contexts. More recently, the same approach has been used in the field of Autonomous Agents Systems and Robotics to anticipate expected system behaviours in the absence of global controllers, e.g. for network theory models applied to human agents and for swarm robotics. In this lecture we shall present two study cases that we have recently developed in these areas and use them as explanatory tools to question some foundational aspects of ABS. In particular, we will focus on aspects that have been investigated in the literature for computer simulation at large, but less so specifically for ABS, including:
Computer simulations and experiments
- the relation between reality, formal model and ABS implementation;
- the role of code and its execution in relation to the semantic of intention in the predictive and explanatory power of ABS;
- the practice of experimental check and formal testing of code on the design and re-formulation of the intended model.
It is undoubtedly evident that science has entered what has been called the ‘age of computer simulations’. If computer simulations were adopted at the beginning to build tractable models to solve the equations provided by theories, nowadays their role expanded and, besides dealing with the construction of models of greater and greater complexity, they are used to augment the exploration opportunities. In the last years, the experimental capabilities of computer simulations have been put under attention, and philosophers have begun to consider in what sense, if any, computer simulations are experiments. They can be used as experiments in different ways: ‘just’ as techniques to derive numerical solutions to systems of differential equations with non-analytical solutions, but also as explorations to develop new hypotheses, models, and hints. The use of computer simulations as experiments, thus, opens a wide range of possibilities for investigating a number of unexplored, and sometimes even unimaginable, phenomena and processes. By moving from experiment as a pure controlled experience, where simulations implement another way (with respect to real experiments) of choosing and controlling experimental factors, to experiment as exploration, where the theoretical model under construction is shaped by simulation results, the reliability of simulation results becomes urgent.