INSIGHT: understanding unexpected behaviours in agent-based simulations

Abstract

Unexpected behaviours in simulations require explanation, so that decision-makers and subject matter experts can separate valid behaviours from design or coding errors. Validation of unexpected behaviours requires accumulation of insight into the behaviour and the conditions under which it arises. Agent-based simulations are known for unexpected behaviours that emerge as the simulation executes. To facilitate user exploration, analysis, understanding and insight of unexpected behaviours, we have developed a novel semi-automated methodology, INSIGHT. INSIGHT provides: (1) semi-automatic hypothesis testing for exploring an unexpected behaviour, and (2) automatic identification of statements in an agent-based simulation's source code which have the strongest influence on an unexpected behaviour. INSIGHT is applicable to both deterministic and stochastic agent-based simulations and extends the state of the art in agent-based simulation analysis.

Keywords:

• simulation
• methodology
• computational analysis
• validation
• agent-based modelling
• emergent behaviour