For many years, there have been papers in JASSS and talks at ESSA conferences about the need engage with psychological theory in order to enhance the realism of social simulations (e.g., Jager 2017 in JASSS).
At the same time, cognitive scientists have solved many of these problems, coming up with psychologically realistic and experimentally well-tested agent architectures at the individual level. These models include rule-based approaches (e.g. Anderson’s ACT-R), coherence-based neural network models (e.g. Thagard’s work or more generally connectionism in social psychology), newer neurobiological models (e.g. Eliasmiths spiking-neural-network brain model), and even sophisticated models that implement cultural conceptual structures (e.g., Heise’s affect control theory).
It is therefore odd that the calls for more psychologically grounded social simulation keep being repeated while the uptake from cognitive-science models is so slow. We conjecture that there is a quite pragmatic reason for that; namely, that the different research communities rely on their specific tools. In the social simulation community, the Netlogo platform is by far the most influential tool, but it contains no standard way of implementing state-of-the art cognitive-science models.
As if we needed a reminder from reality for how important psychology indeed is for understanding society, the current political crisis in many democratic societies defies assumptions about human communication that are implicit in many social simulations. This is true for simple spread models (in reality, people can be quite resistant to new information, if it does not suit their pre-existing belief systems), but also of course for rational “economic man”-type models (in reality, people lack full information and cognitive capacity to optimise decisions).
Therefore, we aim to start a conversation of how we can jointly produce a “cognitive science” module (or set of modules) for Netlogo, enabling a quick and easy uptake and integration of relevant approaches into typical social simulation models. In this way, we aim to start a collaboration that will (eventually) provide the cognitive modelling tools that the community needs.
We will spend time in the workshop to assemble
a) the most relevant cognitive-science modelling approaches;
b) the most typical social simulation scenarios for which a cognitive-science module might prove useful; and
c) a roadmap for developing such a model, including prospects for funding such an endeavour.
We propose that this meeting happen on Monday the 23rdth of September 2019. We suggest a length of 120 minutes + 20 min coffee break. If accepted and if possible, we would ask that this is not scheduled opposite the Qual2Rule workshop, since there is a substantial overlap of interest.
Bruce Edmonds and Tobias Schroeder