Reviews and discusses the modelling of cognition in social simulation with respect to its context-dependency, then proposes a principled way, using cognitive con-text, of integrating machine learning and reasoning processes into a single cognitive model suitable for use in social simulation.
Cite as:
Edmonds, B. (2015) Contextual Cognition in Social Simulation. In Brazillon, P., Gonzalez, A. J. (eds.) Context in Computing – A Cross-Disciplinary Approach for Modeling the Real World. Springer, pp 273-290. DOI: 10.1007/978-1-4939-1887-4)
Abstract: This chapter looks at the modelling of cognition in social simulation with respect to its context-dependency. After making some conceptual clarifications, it briefly reviews existing attempts to include context-like elements into social simulations. It then proposes a principled way, using cognitive con-text, of integrating machine learning and reasoning processes into a single cognitive model suitable for use in social simulation. This approach is not only particularly suitable for social agents and their coordination but solves several problems at once including: the feasibility of learning and reasoning, and avoiding over- and under-determination of practical reasoning. Using an ex-ample model of an artificial stock market it shows how context-dependency can make a substantial difference to the outcomes from such models.