Modelling dynamic normative understanding in agent societies

Abstract

Agent-based Modelling appears as a promising analytical tool when it comes to a lasting question: in how far did different institutions affect the social and economic outcomes of societies? Taking an incremental step to address this question, we present a refined approach that combines existing institution representations (the structure) with a norm identification process to systematically 'grow' normative understanding from the bottom up without relying on any prior knowledge. The proposed mechanism provides agents with the ability a) to detect complex normative behaviour by developing and differentiating stereotypes of social actors, and b) to generalise behaviour beyond observed social entities, giving agents the ability to develop normative understanding as a potential precursor for predicting newcomers' behaviours. We exemplify this approach using a simulated prototypical trader scenario that is evaluated with respect to behavioural diversity (different compositions of non-/cooperative agents) as well as structural diversity (different types of agents). Using the simulation results, we showcase the explanatory power of the derived normative understanding beyond the interpretation of quantitative results, and finally discuss the generalisability of the proposed approach.